What each update to Stintwise changed. Stintwise is not yet listed on the Atlassian Marketplace, so no customer's site runs it yet: each date is the day the change reached the app's production environment. Checked on the test site means checked on a Jira site the developer keeps for testing, and not yet checked names what has not been checked there.
One update is not listed yet. On 1 October 2026 the same chart was added as a widget for Jira's new dashboards; it has its entry here once its checks on a real new dashboard are done, and until then the status page lists what of it is not yet checked.
When an entry says charts are rebuilt once, the first view of the tile after that update builds its chart afresh (how a chart is built). Nothing else is needed from you.
3 November 2026: a chart's next build after a weekend
- What a chart's next build starts from is kept 97 hours after it is written, where it was 49, so that the first view after a weekend, like the next morning's, starts from the last chart built before it and searches Jira again only for what changed since, after checking with counts, as you, that the rest still holds (how a chart is built). The app can no longer read it 145 hours after it is written, at the latest (how long data is kept).
- What was kept before this update still expires 49 hours after it was written.
- No chart's numbers change.
- Not yet checked on the test site: a chart built on a Friday whose first view on the Monday after starts from the Friday's build.
1 November 2026: a large chart's first view, read from both ends at once
- A chart's build reads, as you, a first search the view's count expects to hold more than 200 work items from its oldest and its newest work items at once, until the two meet, beside the work in progress, and then the status histories up to three blocks of 1,000 at once (how a chart is built). The check that you can still open every work item a chart lists, made before the chart is shown, counts up to three lists of 500 at once.
- A tile that has just started its chart asks about it 3 seconds later, and from then on as before.
- Each build still sends at most ten requests a second, now also while its reads overlap, and when Jira asks a build to slow down it reads one thing at a time for the rest of it (the time limit). The app's logs say so, and say when Jira's search returns work items out of the order a build asked for (logs).
- The same work items are read and kept as before, so no chart's numbers change. Charts are rebuilt once.
- Checked on the test site, in the app's development environment: three first views of about 450 work items, each asked about 3 seconds after its chart was queued and ready within 10 seconds of it; and a first view of 6,000 work items and one of 10,500, which took about 39 and 46 seconds from the page's reload to the chart, of which the build took 25 and 35, where first views of charts of the same work items took about 62 and 72 seconds earlier that day, the build 45 and 59, before this update's reads at once. Jira answered the earlier 6,000-item view's searches more slowly than the 10,500-item one's an hour later, so part of the 6,000-item difference is Jira's pace that hour. Both charted and excluded as many work items as the earlier ones, 5,000 and 1,000, and 5,000 and 5,500, and used about as many of Jira's hourly points as the app counts them: about 6,200 and 10,700, where the earlier ones used about 6,100 and 10,600. The 6,000-item chart's ledger exported after the update is byte for byte the one exported before it.
- Not yet checked on the test site: in the app's logs, the build's times as wall time and its requests a tenth of a second apart.
1 November 2026: a summary on a schedule, through Jira Automation
- Two actions for a Jira Automation rule, Write a Stintwise summary and Write a Stintwise comparison, write the summary Copy text and Copy Markdown give for a tile set up in the action's form, counted as the person whose connection the action uses, for the rule's later steps to send (how). The app sends nothing; your rule sends it. A saved filter or a project can be a source; a JQL query cannot, nor can a saved filter whose query names another saved filter or uses a function an app adds to Jira.
- Whoever the rule sends the summary to can read what it states, whether or not Jira lets them open its sources, and only settings saved by the person whose connection the action uses can run (who can read it).
- Each run counts as steps in your site's Automation usage, and every summary needs a warm-up rule earlier the same day (the recipe).
- A table whose charts list more work items than one run of the action can check in time is never ready: the form
says so under the sources, and a run says so in its
note. - No chart's numbers change.
- Checked on the test site, in the app's development environment, on 1 and 2 November 2026: Write a Stintwise
summary in a rule with a Scheduled trigger, its form inside Jira Automation's editor set to a saved filter, Last
30 days and every hour counted; its outputs offered as the smart values
{{fetchedStintwiseSummary.ready}},{{fetchedStintwiseSummary.text}},{{fetchedStintwiseSummary.markdown}}and{{fetchedStintwiseSummary.note}}; the first run not ready, itsnote, This chart is being built., in the rule's audit log through Log action; the next run's summary posted as a comment on a work item, with no work item key in it; the next day's comment byte for byte the text Copy text gives on a tile set up the same way; and the audit log recording that the action ran, and none of its outputs. On 2 November 2026, Write a Stintwise comparison in a rule with a Scheduled trigger, its form set to six saved filters, Last full quarter and weekdays counted; its outputs offered as the smart values{{fetchedStintwiseComparison.ready}},{{fetchedStintwiseComparison.text}},{{fetchedStintwiseComparison.markdown}}and{{fetchedStintwiseComparison.note}}; and its warm-ups counted for those six, whose charts each charted 27 to 149 work items, in runs about a minute apart: four runs not ready, each starting three charts, each period before started after its source's chart, and the fifth ready. - Not yet checked on the test site: the text a comparison writes, Write a Stintwise summary of several sources, a run that fails and its sentence in the rule's audit log, the form's refusals and the largest settings a rule keeps, how many warm-up runs any table other than the one counted needs, and any table when they are 15 minutes apart or in an hour when people at their tiles start charts, and a summary sent by email, to Microsoft Teams, to Slack, or as a comment naming a source whose name holds formatting; and, until the developer has a second account on the test site, an action whose settings someone else saves, and a saved filter whose query someone else changes.
26 October 2026: full screen
- Full screen, the last action in a chart's or a table's footer, fills your screen with the tile for a retro or a review, enlarged to fit, and Exit full screen or Escape leaves it (how). Going into or out of it stores nothing and records no event.
- Full screen answers the roadmap's a full-page view, which leaves the roadmap: no page of its own is added to Jira's Apps menu.
- Checked on the test site: in Chrome, Full screen offered inside Jira's frame and reached by keyboard; a refused request's line and its link; the tile's settings in Jira the same after as before.
- Not yet checked on the test site: full screen inside Jira's frame pressed by a person, in Chrome and then Firefox (the browser the checks are run in refuses full screen on every page): the face filling the screen at the width and enlargement the docs give, the Flow view's and cumulative flow's days pointed at and pressed, Copy text, Escape, the dashboard's scroll after leaving, a source of a table pressed in full screen, the dark theme, and whether the tile's hourly reload ends it; and on the widget for Jira's new dashboards.
26 October 2026: lead time on the chart
- On Histogram, Scatter and Trend, Measure under the chart chooses Cycle time or Lead time. In lead time the view draws lead time, from creation, over the work items with a lead time: the charted ones and those that never entered a Started status, each with a lead time. The headline, the line under it, the reasons under the chart, the key, the values to the minute, Show, Compare with … and Image or summary all state lead time, with cycle time's 85th beside it as the reference, and without a lead time lists the work items it leaves out for want of one (lead time on the chart). Your choice is remembered for the gadget in your browser (privacy); Show a target and Show natural process limits are not offered in lead time, and stay ticked for cycle time.
- Under Show, the key now states the group's own lead time, over its charted work items and those of it that never entered a Started status, where it said lead time was not split.
- Export data (CSV) and every list's export close on two more rows,
lead_timep50andp95: lead time's 50th and 95th percentiles, over the same work items as its 85th and with its counts. They come after the Aging view's three lines, so no row a spreadsheet already refers to moves; the file now ends with nine closing rows (the CSV format). Each figure lead time states is reproduced from the file by the recipe. - Every chart states the same numbers as before, but for a group's lead time under Show. Charts are rebuilt once.
- Checked on the test site: a chart of a project from 26 September to today, weekends not counted, with two work
items added the same day, one moved straight from To Do to Done and one through In Progress: with Include today,
cycle time charting 8 and lead time 9, 9 work items with a lead time · 3 excluded from lead time: 3 counted to zero,
the never-started one in the lead-time bar's list; the export's three
lead_timerows with 9 and 3 and its 50th, 85th and 95th equal to the 5th, 8th and 9th of the nine lead times, the 85th the headline's 2 minutes; Lead time still chosen after a reload; the summary and the image in lead time, the image's file named…-lead-time-…; and Scatter and Trend in lead time. - Not yet checked on the test site: the choice not carried to another browser; and the views in lead time at 304 px and in the dark theme.
26 October 2026: a calendar for each source side by side
- On a tile of 2 to 20 sources side by side, Public holidays, in the settings, offers The same for every saved filter and Each saved filter's own: with the second, each source has a holiday list of its own, a country or one of its states or regions, chosen in a select named for it. A source added starts on the list chosen above, one moved keeps its own, and down to one source the list chosen above takes the remaining one's (the setting).
- Such a row is counted in its list's time zone where the list names one: a state's or region's own, its main one where it spans several, such as America/Chicago for Texas, or a country's of one time zone, such as Asia/Tokyo for Japan. Every choice that sets a zone names it, as Texas, United States (America/Chicago); a country of several time zones, such as the United States' national holidays only, counts in the gadget's. Its weekends, holidays, days the whole team is off and working hours then fall by that zone's clock, while the Date range, its days and every time stay in the gadget's; a team in another of a region's zones, such as El Paso's in Texas, is counted on the main zone's clock (saved filters side by side).
- Each row states its own list on its line under the table, as Germany (Sachsen) holidays · counted in
Europe/Berlin, in its name as a screen reader reads it, and in the table's image and summary, and the table names
none of its own: weekends and each saved filter's own holidays not counted.
A row pressed says, on its footer, its drill panel, its image and its summary, which zone its calendar is counted in
beside the zone its times are in, and its ledger export names that zone in a
calendar_zonerow beforeworking_day(the CSV format). - The privacy page and the security page say that everyone who can see the gadget is sent each source's holiday list with its settings, also for a source they cannot open.
- A dashboard open since before this update shows This tile's settings were changed while it was open. and Reload the dashboard, or, for a tile of one source, that the chart could not be loaded, until it is reloaded, wherever a tile's sources each have their own list: it never states such a table as if it had one calendar.
- Every other chart states the same numbers as before. Charts are rebuilt once.
- Checked on the test site: a table of two saved filters over 12 March 2026, counted 09:00–17:00 on weekdays, the
first on Sachsen's list and the second on Texas's: the same work item, in progress from Wednesday 4 March at 11:15 to
Thursday 12 March at 16:30 Belgrade time, charted within 6.7 working days on the first row and 6.2 on the second,
counted in America/Chicago; each row's line under the table; the second row pressed, its footer, image and summary
naming America/Chicago, and its export's In Progress row with 49.5 hours counted and its
calendar_zone,America/Chicagorow; Export saved filters (CSV) with 53.25 and 49.5 hours; and, in the settings, The same for every saved filter saved, both rows then within 6.7 working days, and Each saved filter's own chosen again with Texas typed for the second row and saved. - Not yet checked on the test site: a typed query's row; twenty rows at 304 px by keyboard, light and dark; and a table cut to one source with its alert.
26 October 2026: days the whole team is off
- Public holidays, in the settings, has a field under its select, Days the whole team is off: up to 60 weekdays from 2015 to 2032, one date per line, such as 2026-12-24. Its note says that everyone who can see the gadget is sent these dates, so the list is for days the whole team does not work, never one person's leave, and a date that cannot be saved says why, as 2026-12-26 is a Saturday, and weekends are already left out. (the setting).
- Each listed day is left out of the chart whole, as a public holiday is, while weekends are not counted, on every
row of sources side by side too; a day that is also a public holiday of the chart's list is that holiday. The chart
says so, as weekends, Germany holidays and team days off not counted, never how many, and the drill panel and the
ledger export name each listed day that removed time, as team day off 24 Dec 2026 and
team-day-off:2026-12-24(the CSV format). - The privacy page names the days the whole team is off among the settings everyone who can see the gadget receives, and says that the app cannot tell a team's days off from one person's, and names each day the whole team was off that removed time among the exclusion codes the app stores; the security page says the settings hold them.
- A dashboard open since before this update shows This tile's settings were changed while it was open. and Reload the dashboard, or, for a tile of one source, that the chart could not be loaded, until it is reloaded, wherever a tile's settings list days the whole team is off: it never states such a chart without saying so.
- On a day whose clocks go back across midnight, as they do in a few time zones such as the Azores' and Cuba's, the day now always starts at the first of its two midnights, whatever the season the chart is built in: a chart built in another season could start that day an hour or more later. A day whose clocks go forward over midnight starts, as before, when they do.
- Apart from those days, every other chart states the same numbers as before. Charts are rebuilt once.
- Checked on the test site: a work item in progress from Wednesday 4 to Thursday 12 March 2026, with weekends not
counted and Monday 9 March listed as a day the whole team is off, charted within 5.3 days, against 6.3 without the
list, the face saying weekends and team days off not counted; its export's In Progress row with 72 hours excluded
as
weekend + team-day-off:2026-03-09and 125.25 counted; and the drill panel's 3 d excluded (weekend, team day off 9 Mar 2026). - Not yet checked on the test site: the field by keyboard, a Saturday and 61 dates refused with their lines, and the note under it.
23 October 2026: working hours that differ by weekday
- Time counted, in the settings, has a fourth choice, Working hours that differ by weekday: Monday to Friday each with its own From and To, on the hour or the half hour, each at least an hour long. Choosing it fills every weekday with the third choice's times, or 09:00 and 17:00, and a refused time names its weekday, as Friday's working day must end at least an hour after it starts, on the same day. (the setting).
- A day on such a chart is the average working day, the week's working hours divided by five: 7.2 hours for Monday
to Thursday 09:00 to 17:00 and Friday 09:00 to 13:00. The chart says it counts working hours that differ by
weekday, the footer and the method line name the week, the drill panel says what d is, and the ledger export's
working_dayrow names the week, asMon-Thu 09:00 to 17:00 and Fri 09:00 to 13:00, with the day's hours,7.2000(one day, the CSV format). On such a chart a target in days is whole days. - The recipes that check a verdict against the exports, faster, slower or about the same, write the two printed figures in whole seconds rather than whole minutes, which is exact on every chart: a tenth of an average working day of 7.2 hours is 43.2 minutes (the comparison's recipe).
- The privacy page names each weekday's own working hours among the settings.
- A dashboard open since before this update shows This tile's settings were changed while it was open. and Reload the dashboard, or, for a tile of one source, that the chart could not be loaded, until it is reloaded, wherever a tile's settings are changed to working hours that differ by weekday: it never states such a chart in a day it does not know.
- Every other chart states the same numbers as before. Charts are rebuilt once.
- Checked on the test site: a work item in progress from Wednesday 4 March 2026 at 11:15 to Thursday 12 March at 16:30 charted on Monday to Thursday 09:00 to 17:00 and Friday 09:00 to 13:00 within 6.9 working days, 49.25 hours on a working day of 7.2 hours, the footer naming the week; the fourth choice in the settings by keyboard, a Friday ending at 08:00 refused with its line and the hint's average withdrawn until it was put back, the saved settings then holding the week and no single window; and a save back to Weekdays storing no week, the same work item then within 6.3 days.
- Not yet checked on the test site: that chart's export's
working_dayrow; the settings at 304 px; and a dashboard held open across the update showing the reload line, never a figure.
23 October 2026: public holidays of a state or region
- Public holidays, in the settings, offers 219 states and regions of 12 of the 20 countries besides the countries
themselves: each under its country, after the country's own list and by its own name first, as Bayern, Germany
after Germany: national holidays only, so typing its name finds it (with the comma after Sachsen, Tasman or
Mato Grosso, which begin other names). A state's or region's list is the public holidays the open-source
date-holidays library lists for it, which can differ from local law, with a few of its country's holidays kept: a
day of Switzerland's that a canton marks only as optional or as a bank holiday, such as Easter Monday in Lucerne,
since most people there are off; Brazil's 20 November in its states from 2024; and Tasmania's New Year's Day on the
Monday after a Sunday and its Easter Sunday from 2027. Known differences from the law include Washington's and Oregon's lists keeping Columbus Day,
Ontario's adding Easter Monday, British Columbia's missing 30 September, the National Day for Truth and
Reconciliation, and Washington's missing the Friday after Thanksgiving, Native American Heritage Day there;
public holidays names each one known. The drill panel and the
export name each holiday that removed time. The chart names a state or region after its country, as
weekends and Germany (Bayern) holidays not counted, and the ledger export writes each of its holidays under its
code, as
holiday:DE-BY:Fronleichnam(the CSV format). - A public holiday that starts partway through a day, such as Christmas Eve evening in South Australia, is in no list, so its day counts as any weekday, and the settings and the documentation say so.
- The privacy page names a state or region among the settings, and says the exclusion codes the app stores name each public holiday with its country or region.
- A chart of a country's public holidays states the same numbers as before. Charts are rebuilt once.
- Checked on the test site: a work item in progress from Wednesday 4 to Thursday 12 March 2026, with weekends not counted, charted within 5.3 days on Victoria's list, which removes its Labour Day, Monday 9 March, and within 6.3 days on Australia's alone, the face naming Australia (Victoria) holidays and Australia holidays.
- Not yet checked on the test site: that chart's export naming the holiday; and the settings by keyboard at 304 px, in light and dark, in Chrome and Firefox, typing a region's name to choose it.
18 October 2026: changing a tile's settings always takes effect
- A classic Jira dashboard keeps any setting a save leaves out, and a save used to leave out each setting the choices made had no use for, so a choice undone could stay in force: a tile of several sources cut to one stayed a table of them all, working hours turned off with Weekdays were still counted, and a Date range moved from custom dates, a calendar period or Last closed sprint to Last 30 days, Last 90 days, Last 180 days or Last 365 days kept the range before it. Each save now writes every setting, one the choices have no use for as empty, so a table cut to one source becomes that source's chart, and turning working hours off counts the time chosen.
- A tile whose settings were changed that way before this update shows the change once its settings are saved again.
- Charts are not rebuilt.
- Checked on the test site: a table of two saved filters cut to one in the settings and saved, the tile then that filter's chart, its stored settings listing no second filter and every setting the choices had no use for stored empty.
- Not yet checked on the test site: working hours turned to Weekdays and saved, the chart then counting whole weekdays; and custom dates moved to Last 30 days and saved.
18 October 2026: four quarters side by side, and the next build of a last full month's or quarter's period before
- On a tile of several sources side by side whose Date range is Last full quarter, Show four quarters under Compare with … sets each row's 85% in each of the three quarters before beside its own, oldest first, in the table's unit and in place of its 50%, 95% and strip. Each quarter is that source's own chart of the quarter, built for you once you tick the box, the nearest first, three at a time, or one at a time while one holds more than 1,000 work items, unless Jira confirms, as you, that a colleague's holds the same work items. The box is offered while the rows' searches find 2,000 work items or fewer between them in a quarter, and the tick is not kept. Pressing a row while four quarters are shown offers the whole ledgers of its second and third quarters before in its footer, each once that quarter is in; the nearest is the period before its own Compare with exports (saved filters side by side). Once every row's quarters are in, the table's image and summary state them too, and Export saved filters (CSV) adds each quarter's columns and ends its name with the first quarter's first day (the file).
- The period before a chart of Last full month or Last full quarter, which Compare with sets beside it, starts its next build only from earlier builds of that same month or quarter, never from another's, so at the turn of a month or quarter it asks Jira for no checks that cannot hold (how a chart is built).
- The privacy page says that the period before of each source on a tile of several side by side is searched and built for a project or a typed JQL query as for a saved filter, as it already was; and it names what the app searches to show each source over four quarters, the three calendar quarters before a last full quarter.
- On a tile of several sources set beside the period before, the comparison no longer stops, until its box is ticked again, when Jira says the app's hourly limit is used up: it says until when, asks again by itself once the limit resets, and a period before that was being built then says it is taking too long only five minutes after that (why).
- No figure the tile states changes. Charts are rebuilt once.
- Checked on the test site: the first view after the update building each chart afresh; four quarters on a table of three saved filters, ticked by keyboard, each quarter built nearest first and each quarter with nothing finished in it showing a dash; a row's two quarter exports reproducing its cells, one of them a single work item shown as Cycle time within 197.3 h, and Export saved filters (CSV) with four quarters reproducing them too; the table's image and summary stating the quarters; the quarters read, not built, when ticked again after a reload; and, on the development site with its part of the hour set low, the quarters still to build each saying Waits until and the time, with no build started, then building by themselves a minute after the hour turned.
- Not yet checked on the test site: the quarter before a chart of Last full quarter, built on two days in a row, the second built from the first; a tile of saved filters side by side compared with the period before, each row's period before built and shown as before; that comparison going on by itself once Jira's hourly limit for the app, used up, resets; Compare with ticked after four quarters building nothing new; the table's image and summary, opened before the tick, gaining the quarters once they are in and losing them, still open, on an untick; the table at a 304 px column and in Jira's dark theme; a project's and a JQL query's quarters; and a second account shown a colleague's quarters once Jira confirms its own.
18 October 2026: each site's new charts kept to their part of Jira's hourly limit
- The app keeps, for your site, how much of Jira's hourly limit for this app your site's new charts have used this hour, as the app counts it from what Jira's searches returned, and, while a chart is being built, what its build is expected to use. The count starts again at 0 each hour, names no work item, filter, query or person, and expires as the app's other records do (privacy).
- Once your site's new charts have used a fifth of that limit this hour, as the app counts it, or Jira says less than a tenth of it is left, a chart not yet built waits until the hour turns, unless its own build is already running, when it waits its turn; and once what your site's new charts have used and what those still being built are expected to use come to that fifth, the next ones wait their turn while they are built. Charts already built, their lists and their exports are served as usual. The app adds a random amount of up to 100 points of that limit to every chart it counts, so that one such wait cannot reliably tell whether a colleague's search found a few work items or none; what a wait can tell you about other people's new charts is on the privacy page, and the app's logs say when it holds a new chart back (logs).
- The tile says why and until when, the time in the zone of the person who set it up, with that zone named, says that it asks again a few minutes after, and does, by itself. A chart waiting its turn says so. A tile of several sources draws each row already built, and a row waiting for the hour says Waits until and the time; its answer, its file, image and summary and its period befores wait for every row, as they do for any row not yet built. Compare with, Include today and Measure ages each say that they wait, and stay ticked (why).
- No figure the tile states changes. Charts are not rebuilt.
- Checked on the test site, with the site's part of the hour set low for the check: a first view of 445 work items counted at 494 of the hour, where Jira's own reading of such a view was 474 on 11 October, the difference within the random amount; a new chart then waiting until the hour turned, with its log line, while the built chart's views, its list of work items and its file were served; a tile of two saved filters drawing the built row and saying Waits until and the time for the other, its file not offered, then asking again a minute after the hour turned and building by itself; and, with the site told that little of Jira's limit is left, a new chart waiting, and Compare with waiting beside a chart already built, which stayed.
- Not yet checked on the test site: a build delivered again, counted once for each delivery and held once; a chart waiting its turn while other new charts are built; Jira itself saying less than a tenth of the limit is left.
17 October 2026: Aging and By status grouped by a board's columns
- On a chart of a saved filter, Aging and By status can be grouped by the columns of the boards that use that saved filter: Group by: Status · Board column, under the view's answer. Aging's columns become the board's, each holding its statuses' work items, a column's count the sum of its statuses' counts; By status's rows are grouped under their column's heading, whose figure is the sum of the figures printed in its rows. A status of the chart on no column is under Not on the board. The lines, the stays and the sentences are as they were. The choice is remembered in your browser. A chart of a project or a typed JQL query offers no such row (grouped by board column).
- The columns are read as you, when you choose them and each time the tile opens either view with them remembered, from up to five boards you can open, and used only when those boards set out their columns alike; otherwise the view stays by status and says why. Only the chart's own statuses are named. They are never stored, and never shown to anyone but you (privacy).
- This update asks your site's administrator to approve two more permissions,
read:board-scope.admin:jira-softwareandread:project:jira, with which the app reads the boards that use a saved filter and their columns; neither lets it write anything in Jira (security). - A file exported while either view is grouped by board column reads the columns again, as you, and closes with a
board_columnrow for each of the chart's statuses a column holds, before thestarted_statusrows, so each column's count is reproduced from it (the CSV format). Where the board's columns changed since the tile drew them, the tile draws them again and says so. - No figure the tile already stated changes, and every row the file already had is the same.
- Charts are rebuilt once.
- Checked on the test site: until the site approved the two permissions, it went on running the version before this one, so this version never runs without them; once approved, Board column on a saved filter no board uses said so and stayed by status, which Jira's own list of that filter's boards confirmed; and on the sprint project's board filter, Aging grouped as the board's three columns, To Do, In Progress and Done, each holding the statuses the board maps to it, and By status under the same headings, each heading the sum of its rows; no board's name reached the app's log.
- Not yet checked on the test site: the words Atlassian shows an administrator for the two permissions; a viewer who cannot open one of the boards; a team-managed project's chart offering no Group by row; a grouped export against the face; a column read by VoiceOver; what one choice costs of Jira's hourly limit; and a board's columns renamed while the tile is open.
13 October 2026: cumulative flow in the Flow view, flow efficiency on By status, natural process limits on the Scatter, and the statuses of work finished after a chart's range and your Started statuses in the ledger export
- The Flow view can draw cumulative flow: Chart: Run charts · Cumulative flow, under its warnings, draws its days as one panel, the work finished from its first day at the bottom and above it, at each day's end, the work in progress in each status, adding up to the run charts' work in progress on every day, or the run charts are drawn with a line saying so. A day's list of work in progress, in either chart, states the status each work item was in at the day's end. The choice is remembered in your browser. Every band is reproduced from the ledger export (cumulative flow), and Image or summary draws it and states the last day's bands. The Flow view's recipe now takes each of the six kinds of group Show offers.
- By status states, under its answer, the share of the charted cycle time spent in your Started statuses and the share spent in other statuses between start and finish, which always add up to 100%, as in Flow efficiency: 62% of cycle time was in your Started statuses, marked below, and 38% in other statuses between start and finish. Each Started status's row is marked Started. There is no new setting: the Started statuses you chose are the ones counted as working time, so a queue status among them, such as In Review, counts as working time until you untick it. With one group shown, the line says the figure is the whole chart's only. The figure is reproduced from the ledger export (flow efficiency), and Image or summary states it too.
- Scatter can show natural process limits: Show natural process limits, under the chart, draws the average of the cycle times shown and their limits by ActionableAgile's formula, taken in the order they finished, printed outward on the steps every figure takes, and rings each dot beyond them, as in Natural process limits of the 287 cycle times shown, taken in the order they finished: average 4.2 d, upper limit 13.6 d. 6 took longer than it., with a list of them. It needs 20 cycle times, and is not offered while the tile compares the chart with the period before. The box is remembered in your browser. Every figure is reproduced from the ledger export (natural process limits), and Image or summary draws and states them too.
- Export data (CSV) writes, for each work item that finished after the chart's range and before it was built,
a
since_segmentrow for each status it was in from its first move into a Started status until the range ended, before itssince_totalrow: which status it was in at the end of each day of the range (the CSV format). - Every file a chart gives, a list's too, but for Export ages now (CSV) and its lists, closes with a
started_statusrow for each of your Started statuses that its charted work items spent cycle time in, after the other closing rows and before a working-hours chart'sworking_dayrow. A Started status removed from your site's workflows since is named by the name its history gives. - Each chart keeps, for each status, the counted time charted work items spent in it and whether it is one of your Started statuses, and, for each day the Flow view draws work in progress, how many work items in progress were in each status at its end; a chart whose counts by status would be too large keeps none of them (privacy).
- No figure the tile already stated changes, and every row the file already had is the same.
- Charts are rebuilt once.
- Checked on the test site, on its 450-item filter, by keyboard: the first view at this update built again,
its export closing with a
started_statusrow; flow efficiency against that export by the page's steps, under 1% and over 99% on the Started statuses as preselected, with the Started mark on its row; the Scatter's natural process limits against the page's script, average 1.8 h and upper limit 6.6 h of 445 cycle times, the one work item longer than it ringed, beside a target, each rule in a stroke and tag of its own; cumulative flow against the page's script on three days, each day's bands and finished count the same, and a day's list opened with Enter with its At the day's end column; each part loading inside Jira's frame. - Not yet checked on the test site:
since_segmentrows, which need work finished after a chart's range; three work items' statuses at a day's end traced in their change history; what the build costs of Jira's hourly limit; flow efficiency with a queue status unticked and with one label shown; the limits with one label shown and on the sprint filter's Last closed sprint; the bands on five days against the upper run chart, beside Jira's own cumulative flow diagram, with one label shown, and no Chart row on Last closed sprint; and the Started rows, a ringed dot and the slider read by VoiceOver.
12 October 2026: a chart's next build reads again only what changed
- A chart built again within 49 hours of an earlier build of the same settings, such as the next quarter hour of a chart that includes today, a colleague's next chart of the day or the next morning's first, no longer searches Jira for every work item again where checking what changed uses fewer of Jira's hourly points. It counts, as you, the work items updated since about an hour before that build and searches only for those; it checks with counts, as you, that each block of up to 1,000 earlier work items is still returned by your search, and with searches, as you, that no work item joined it without being updated; it searches a block again where a count falls short; and it reads again one work item for each project, status, issue type, sprint, parent, component and team, whose keys, names and sprints' dates change without the work item being updated (how a chart is built).
- A search these checks cannot settle, such as one where more than half the work items changed or a sprint closed since, is read in full, as before. So is one whose checks would use more of Jira's hourly points than reading it again, at a point for each search and for each work item it returns and two for each count, such as a search of seven work items or fewer; one whose changes, once counted, would take its checks past that; and one whose checks would ask Jira more than 100 times, such as one spread over 50 projects or more. Once the checks have taken a minute, every search is read in full; and every status history is read again at each build, as before.
- The chart is the one a full search would build at that moment.
- What a chart's next build starts from is kept for 49 hours: every work item its searches returned, with its id, key, creation date, the status it is in now, its issue type by id and name, its sprints, parent, labels, components and team, and when it was last updated (privacy).
- No figure on the tile changes.
- Charts are rebuilt once.
- Checked on the test site: a chart of 450 work items that includes today, built in full at its first quarter hour and from that build at the next, which searched 6 times where a full search reads 456 work items; a twentieth of its work items given a new label and one work item's only label removed, after which the next quarter hour's build found the 23 changed and the 1 that left, searching one block again for it, and its Export data (CSV) held the 23 new labels and not the work item that left; that file against the same chart built by a full search: every row of finished work the same, and the rows of work in progress the same work items, differing only in the hours an open stay had run, since the two were built to different quarter hours.
- Not yet checked on the test site: what such a build costs of Jira's hourly limit; an issue type renamed; a sprint closed; and how long Jira's search takes to show an edit, which sets how far before an earlier build the next one searches. The checks that need a second account with narrower access wait for one.
12 October 2026: one chart for colleagues whose own searches return the same work items
- A tile's chart is still built at the first view of the day, from that person's own search. A colleague who opens the same tile later is shown that chart only when Jira, asked as them a moment before, confirms that their own searches return exactly its work items, that they can open every one, and that its issue types, sprints and teams read to them as the chart states them; otherwise a chart is built from their own search, and colleagues whose searches return the same work items as theirs are shown that one (who sees what).
- Each row of a tile of several saved filters, projects or queries side by side, and each row's period before, is built and shown the same way: from your own search, unless a colleague's chart of it on that tile, built that day, holds exactly the work items your search returns, as Jira confirms, as you, a moment before (saved filters side by side). Pressing a row shows the same chart its row states, and its lists and file are of that chart.
- A chart you are shown is the one you are shown that day. Its figures are as of when it was built, which the chart shows, and that can be earlier than your first view.
- Whoever can edit the tile's saved filter or its dashboard is always shown a chart of their own search, and theirs is shown to nobody else; so is everyone when the saved filter or the JQL query names another saved filter or uses a function an app adds to Jira, and on Jira's new dashboards. On a tile side by side this holds row by row: a row whose saved filter you can edit, every row when you can edit the dashboard, and a row whose saved filter or query names another saved filter or uses an app's function. To tell, the app reads, as you, whether you can edit the saved filter, in the read it already makes, or, for a row, before it is first checked or built for you that day; and the dashboard once a day, or, on a tile side by side, at most once each time the tile asks after its rows. Neither answer is stored (privacy).
- While a colleague's chart of the same settings is being built, the tile says This chart is being built., with no count, and asks again within half a minute (how a chart is built). A row waiting so says Waiting its turn, with no count, and does not hold up the tile's other rows; a row being checked says Checking you can still open its work items.
- What a chart shown this way can tell you about someone else's search, and the limits of the check, are on the privacy page.
- No figure on the tile changes.
- Charts are rebuilt once.
- Checked on the test site: the saved filter's and the dashboard's editor kept to a chart of their own, and a saved filter naming another kept to one too; with the development switch that lets the owner's own test filter be shared, a chart built once as a shared chart, then shown again at the next view without being built again, and its Export data (CSV) downloaded whole (450 work items, 445 charted); a tile of two saved filters side by side, one row shown the chart already built for its filter and the other building its own, a row opened with the same figures as its line in the table, and no row's build dropped because its saved filter changed in between.
- Not yet checked on the test site: a chart built by one account and shown to another whose searches return the same work items, with their ledger exports compared byte for byte; a tile of two projects built again for an account that cannot open one of them, its export holding no key of that project; a work item moved into a Started status, after which the next view builds a new chart; two views within a second building one chart, the other tile saying This chart is being built. with no count; rows a second account is shown from the first one's charts, and a row waiting for a colleague's chart showing no figures; and what each check of a shared chart costs of Jira's hourly limit. These need a second account with narrower access, and wait for one.
12 October 2026: a project or a typed JQL query as a chart's source, a chart split by parent, label, component or team, and a target of your own
- The settings' first control is now Saved filters, projects or JQL. Its box lists the projects you can browse beside your saved filters, and takes a JQL query typed into it, which Jira checks as you; Use this JQL adds a query Jira accepts with no error and no warning, and otherwise the settings quote Jira's first error or warning to you alone (configuration). A project is read as you each time the tile is shown, and a project you cannot browse is said so (that project is not available).
- A typed query is kept in the dashboard's settings sealed with a key only the app holds, and is shown only to people Jira accepts it from, checked as them each time the tile is shown. Jira's check does not look at the people a query names, its labels, its text searches or what it passes to a function, so those words are shown to everyone it runs for (who sees what). A tile whose query Jira does not accept from you says so (why), and so does one whose query can no longer be read after the app is reinstalled (why).
- The footer names a project as Project and its name, and a query as JQL: and its first 80 characters; Time zone, build time, method and query opens it whole, and every image and summary states it whole. A tile with a project or a query among the sources it shows side by side says sources, and counts those you cannot open by kind (side by side).
- The ledger export has four more columns, after
sprints:parent, the key of the work item's parent, blank where it has none or where you cannot open it in Jira;labels;components, by name; andteam, the name in Jira's Team field. Several labels or components share one cell, each on a line of its own. The first 19 columns are as they were (the columns). - To fill them, the app reads each work item's parent, labels, components and team in the reads it already makes of each work item, with no other request: the build's own search as you, or, when Jira does not let the app act as you, its lookup by id (how a chart is built). It finds the site's Team field in the field list it already reads for the Sprint field. It keeps the parent's id and key, the labels, the components' names and the team's name, and nothing else Jira sends with them (privacy). A team is named as the Team field gives it to the app; whether Jira would show you that team is not checked (who can see a chart).
- Show lists a chart's groups by parent, label, component and team after its issue types and projects, and Compare by, now a list, offers every kind the work items shown split into (one group on its own, side by side). The words for these four load the first time a chart has two groups of one of them.
- A work item with two labels, or two components, is in each one's group, so by label or by component Compare's rows add up to more than the work items shown, and a line under its answer says by how many.
- A parent is named only to someone who can open it: each time the chart is shown to you, and when you export it, the app asks Jira, as you, which of its parents you can open, and a work item whose parent you cannot open is under No parent you can open. Parents are named by their keys; the first time you choose one, or Parent under Compare by, the app reads, as you, the summaries of up to 200 parents you can open and names each with its summary. The summaries are shown, and never stored (privacy).
- Show a target, under the chart after Show the 70% line, takes a target you type after within, in days of the chart's day to one decimal or in whole hours, and draws it: a vertical rule on Histogram, a horizontal one on Scatter, Trend and Aging. The key states how many charted work items finished within it, of how many, with the share rounded down, as in Target 5 d: 18 of 23 within it (78%), and on Aging how many in progress are past it (a target).
- Each Trend period states its own count in its tooltip and its name; by sprint, the Trend view draws no
target yet, and its key states the whole range's count. Compare, saved filters side by side and
the comparison with the period before add a Within 5 d column or row, a period before's counted on its
charted cycle times in whole minutes; Export saved filters (CSV) ends with
target_hoursandwithin_target; every image and summary states the target it was drawn with. Each count can be checked against the export (the recipe). - A chart of a project or a typed query is built as a saved filter's is, by the build searching, as you, the project or the query; a query goes with the build's queued job sealed with a key only the app holds, as your account id does (privacy).
- The target is yours alone: remembered for the gadget in your browser, never a setting and never sent to the app or to Jira (privacy).
- No figure on the tile changes.
- Charts are rebuilt once.
- Checked on the test site: the test tile set to the project SPR, with the same figures as filter 10104's
tile over the same 30 days (10 charted, 85% within 1 minute) and no error in its logs; its ledger export, taken
from the keyboard, ending with the four new columns, SPR-1 to SPR-3 holding the parent SPR-10, the labels
apiandmobileeach on a line of its own and the component Backend; from the keyboard, the queryproject = SPR AND labels = apityped into the settings, added with Use this JQL and saved, the tile then charting its 3 work items with the line JQL: project = SPR AND labels = api, and the dashboard's copy of the setting holding none of the query's words;project = NOPEtyped and Jira's own error quoted; Show with the parent SPR-10, named with its summary, and with the labelapi; Compare by Parent, Label and Component, and the overlap line under labels ("the rows add up to 13, more than the 10 shown"); Show a target ticked with Space and 5 typed, the key's "10 of 10 within it (100%)", 3 of 3 under the labelapi, and a week's "10 of 10 within the 5 d target" against the ledger export, and Compare's Within 5 d column. - Not yet checked on the test site: the build's log line counting no value left out; a team-managed board's
work against its project's (board 68 and project SPT); twenty sealed queries saved in the test tile's settings
and read back; the whole query under the footer's switch; Image or summary opened while a group is shown;
side by side's Within column beside the period before; a work item given a team in Jira's Team field, once
the test site's organisation has a team, and its name in the
teamcolumn and under Show; and, with a second account, a project and a query it cannot open, each refused with nothing of the query shown, a parent it cannot open, its work items under No parent you can open and its cell blank in the export, and whether the tile names a team that account cannot see in Jira.
11 October 2026: each chart built by reading Jira as the person viewing it
- A chart's build now searches Jira as you, through Atlassian's offline user impersonation, with the fields the app used to look up by id, and reads each work item's history as you, so each work item is read once where it was read twice: by your search, then by the app's lookup (how a chart is built). The site's status list, its Sprint fields and whether Jira keeps content from apps are still read as the app.
- When you view a chart that is not built yet, the app reads your saved filter and counts, as you, the work items its search will find: the tile's counter starts at that count, and moves to how many the search found once the build has searched. Your account id goes with the build's queued job, sealed with a key only the app holds; the app never stores it or writes it to its logs (privacy).
- When Jira does not let the app act as you, as for an account that is not licensed, the build stops, and your next view of the chart runs your searches as you while you look at the tile, as before, the app reading the rest as itself, until the chart's next day.
- A work item your search returns that an organisation's app access rule keeps from the app is now charted from its history, where it was counted as the app is not allowed to read. Work items a chart excludes that were last updated in the same millisecond may be listed in another order in the drill panel and the ledger export; and when your search returns more than 10,000 work items, of those last updated in the same millisecond as the 10,000th, another may be the one read. Work finished after the chart's range is read up to the moment the build reads it, not the moment you viewed the tile. No other figure changes.
- Charts are rebuilt once.
- Checked on the test site: the three test charts of about 450, 6,000 and 10,500 work items built this way, each charting and excluding the same work items as before, each work item read once and none looked up again; a first view of 450 work items used 474 points of Jira's hourly limit and one of 6,000 used 6,113, where they used about 936 and 12,300; and an hour of seven large first views and four of 450 used about 52,000 of the hour's 65,000 points with no pause, where the same views used more than the whole hour before.
- Not yet checked on the test site: the scope approved on a production installation (none exists yet), and what the approval shows; each chart's ledger export against the same chart built by reading as the app; a tile of saved filters side by side whose rows are built this way; and a person Jira will not act as, whose chart is built as before.
10 October 2026: the Trend view by sprint
- Sprint joins the Trend view's Period, after Quarter: one column for each sprint your chart keeps whose time overlaps its range, in the order they started, then None for the work that finished outside every sprint it was in, so the bars still add up to the charted count (how). Each column is labelled by the day its sprint started; pointed at, it names its sprint as Jira names it to you, with its days, and it opens its work items. Sprint is offered where its 3 to 27 columns, None included, fit the gadget's width, never as a range's own size, and is remembered as every size is.
- To name the columns, the app reads as you, once, when the view is first divided by sprint, the Sprint field of one work item of your chart for each sprint it draws. The names are shown while the tile is open and never stored (privacy). A sprint Jira does not name for you is said so, with its days; if the read fails, each column is named by the day its sprint started, with Try again under the chart.
- Every other figure and file is as it was, and no chart is rebuilt.
- Checked on the test site: the Trend view by sprint on a 30-day chart of a Scrum project's saved filter, by
keyboard: Sprint offered, four columns (the first sprint, cut by the range and hatched, with 3; the next
with 4; the third with 2; None with 1) adding to the 10 charted, and a column naming its sprint as Jira
names it, with the part of it in the range; a column's list, opened with Enter, holding its 2 work items; and
Export data (CSV) against the recipe: its
sprintrows naming each sprint with its days, and its work items counted by the sprint they finished in, 3, 4, 2 and 1.
10 October 2026: the app may read Jira as the person viewing a chart, in its development environment only
- The app's scope to read Jira work,
read:jira-work, is now declared for Atlassian's offline user impersonation, which lets the app read Jira as any person of a site who has access to the app, even while that person is not using it; Jira then shows it only what that person may see. No other scope is declared for it (security). - The app reads that way only in its development environment, for one request it has priced there since 8 October 2026: a lookup of a chart's work items made as the person viewing it (privacy). Charts are still built by reading as the app.
- Nothing on the tile changes, every figure and file is as it was, and the app's production environment sends nothing new.
- Checked on the test site: Forge took the new form of the app's scopes when the app was deployed, and the test site's installation, updated, asked for no new approval, since the app asks for no scope it did not already have; and Jira answered a lookup made that way, from the queue that builds charts.
8 October 2026: fewer requests to Jira for each chart
- To build a chart, the app looks up the work items your searches return, 100 at a time. Jira answers a lookup that fills a page of 100 with a pointer to a next page even when nothing follows, and the app asked for that empty page too. It now stops once every work item it asked for has come back, which removes about half the lookups of a large chart. A lookup still missing a work item asks on, as before, so one the app is not allowed to read is still found and counted (the app is not allowed to read).
- A work item that one of your searches returns as finished in the chart's range and another as in progress, because it was reopened, or as finished after the range, now has its history read once a build, not once for each search, unless the first read was cut short or the site's status list was read again in between; and it is looked up once where the app is allowed to read it (how a chart is built).
- No figure changes, except on a chart whose build reached its time limit, or for which Jira did not return a full change history: since it no longer reads again a history it has read in full, such a chart may now list fewer work items in progress, or finished after the range, as not read or as missing part of their history.
- Charts are rebuilt once.
- Checked on the test site: the three test charts of about 450, 6,000 and 10,500 work items built again with no lookup that returned nothing: 6, 60 and 100 lookups where there were 10, 120 and 200, each chart charting as many work items as before, and the large charts' lookups taking 22 s and 32 s where they took 29 s and 48 s.
- Not yet checked on the test site: each chart's ledger export the same byte for byte as one taken before this update.
8 October 2026: the app's development environment prices one kind of Jira request at a time
- Jira gives this app one hourly limit, which every site with the app shares, and says how much of it is left only several seconds after the requests that used it. So that charts can be built using less of it, the app's development environment can now price one kind of request at a time: each is sent on its own and priced from what Jira says is left before it and, after half a minute in which nothing else is asked, after it. It runs only when the developer turns it on there, starts only while Jira says at least 8,000 points of the limit are left, and sends nothing more once Jira says fewer than 4,000 are.
- Five kinds are priced: a search made as the person viewing a chart, the app's lookups of work items, the same lookups made as that person, a count, and a search that finds nothing. The person's account goes with the job that makes them, sealed with a key only the app holds; the app never stores it or writes it to its logs. Each kind's log line says how far apart Jira's figures for what was left lay, and whether that was close enough to price it (privacy).
- Nothing on the tile changes, every figure and file is as it was, and the app's production environment sends nothing new.
- Checked on the test site: the five kinds priced over one quiet hour: about one point for each work item a search returns, whether the search is made as the person viewing, as the app, or as that person by the app's queue; about two points a count; and about one point a search that finds nothing. The lookups made as the person viewing were answered by Jira.
- Not yet checked on the test site: the cost of a first view of 450 work items measured again with a quiet minute on each side.
8 October 2026: the last closed sprint, the sprint before it, and sprints in the ledger export
- Last closed sprint joins the Date range, in a group of its own, Sprints: the sprint that closed last before today, of the sprints your saved filter's work was in, found among the work it finished in the 90 days before today and its work in progress. Its chart runs from the instant the sprint started to the instant it closed, holds the work that counts in it, and names the sprint as Jira names it to you (how). Where there is no sprint to chart, the tile says why in its place. It charts one saved filter: the settings will not save it beside others.
- Compare with the sprint before sets that chart beside the sprint before it on its board, which the chart's own build picks, named once its chart is in; its rule on the histogram and the scatter reads Sprint before 85% (how).
- On such a chart the Flow view says its first and last days are parts of days, with the times the sprint started and closed, and the Forecast view says the range ended on the day the sprint closed.
- If the app were rolled back to a version before this one, a tile on Last closed sprint would ask for its settings until saved with another Date range; nothing would be charted under a sprint's name.
- Each finished work item's row in the ledger export names the sprint it finished in, in a new
sprintcolumn, and its sprints the chart keeps, in a newsprintscolumn, afterproject; the seventeen columns before them keep their places. A work item counts in the sprint it finished in: of its sprints, one that had started when it finished and had not yet closed, or is still running; where two hold it, the one that closed first (how). Each sprint the rows name has asprintrow of its own, before the closing rows, with when it started and closed, its state and its name as Jira names it to you when you export, or a blank name where Jira does not (why). A file whose sprints could not be read, or whose chart holds more than 100, says so in one row instead (the recipe). - To do so the app reads each work item's Sprint field with the lookups it already makes, finds the field by its type in your site's field list, which it reads about once an hour, and keeps of each sprint its id, board, state and dates; it never keeps a sprint's name or goal (privacy).
- Charts are rebuilt once.
- Checked on the test site: a saved filter of a Scrum project charted as Last closed sprint: its last
closed sprint named on the tile as Jira names it, its four finished work items charted and the one still to
do left out, and the sprint's start and close in the footer as Jira gives them; a team-managed project's
saved filter the same, with no sprint before it on its board; the settings' Last closed sprint beside a
second saved filter refused before any search; and Compare with the sprint before, its three work items
charted and both ledger exports, pressed by keyboard, with each sprint's
sprintrow as Jira names and dates it. - Not yet checked on the test site: a sprint's name for someone who cannot see its board, and how long a build takes with the Sprint field read beside the same chart without it.
8 October 2026: the app's logs say how much of Jira's hourly limit each kind of request uses
- Jira gives this app one hourly limit, which every site with the app shares. So that charts can be built using less of it, the lines the app writes to Atlassian's Forge logs now say how many work items each kind of request returned: when a chart's build is queued, how many the searches made as you returned; when it finishes, how many its lookups of work items returned, how many of those lookups returned none, and how many status changes its reads of history returned. When Jira says how much of the limit is left, which it does once less than about a fifth of it is, the lines also say how much of it the searches, the lookups, the reads of history and the check of what you can open used, and which limit (privacy).
- The line written when little of the limit is left now reads it from the header Jira sends once the limit is enforced, and from the hourly limit alone. It took the least that any limit Jira named had left, usually its per-second limit's, and read only the header Jira sends before a limit is enforced.
- Nothing on the tile changes, and every figure and file is as it was.
- Checked on the test site: the new counts, over an hour of first views of charts of 450, 6,000 and 10,500 work items, and the line that says how much of the limit is left, naming the hourly limit, once less than a fifth of it was. A first view of 450 work items used about 936 of the hour's 65,000 points.
- Not yet checked on the test site: how much each kind of request uses on its own. Jira reports what is left several seconds late, so requests made close together cannot be told apart; and only one of the three kinds of request priced in the development environment ran before the hour ran low.
8 October 2026: the app's logs say where a chart's first view spends its time
- So that a chart's first view of the day can be made faster, the lines the app writes to Atlassian's Forge logs now say where its time goes. When a chart's build is queued: how long the searches made as you and the queuing took. When it finishes: how long it waited in the queue, how many lookups of work items and batches of histories it read and how long each took, and how long writing its records took. When a chart is served after asking Jira whether you can still open its work items: how long reading the chart and that check took. Each line holds these durations and counts and the hash that names the chart, and nothing that names a saved filter, a query, a work item or you (privacy).
- Nothing on the tile changes, and every figure and file is as it was.
- Checked on the test site: three first views each of a chart of 445 work items, of 6,000 and of 10,500. Each wrote its three new lines with its chart's hash and nothing else. From the page's reload to the chart took 14 s, 44 s and 69 s, of which the build took 5 s, 37 s and 59 s, mostly waiting on Jira.
- Not yet checked on the test site: a tile of three saved filters side by side.
8 October 2026: saved filters side by side whose settings change while the tile is open, a status added to the site counted at once, and Jira's hourly limit for the app
- Every request a tile of several saved filters makes names the saved filters its rows were drawn from, in order, and the app answers it only while the tile's settings list the same ones. When someone else saves others while you have the tile open, the tile shows no row and draws the table of the new settings, or, where Jira has not yet handed the tile the new settings, says This tile's settings were changed while it was open. with a button, Reload the dashboard (how). This closes the known issue the status page listed, that such settings could leave the rows' names, and the count of saved filters you cannot open, out of step until a reload.
- A status added to your site, such as a new team-managed project's own Done, counts in every chart built once work has moved into or out of it: a build that meets a status the app's copy of the site's status list does not hold reads the list again. Until this update such a status counted only in charts built from an hour after it was added, and a chart built within that hour showed that project's finished work as unfinished for the rest of its day, as the status page listed among its known issues. A status renamed to a name you chose counts in the next chart built once work has moved into or out of it since the rename, or once a work item the chart reads is in it, and otherwise, as before, from an hour after the rename (how).
- When Jira answers that the app's hourly limit is used up, the tile says until when, in the gadget's time zone, and the app asks Jira nothing more for your site until the limit resets; the tile then asks again within five minutes of it (why). Until this update the tile said only that Jira was rate limiting, and asked again when the dashboard was next reloaded.
- Charts are rebuilt once.
- Checked on the test site:
- The test tile set to three saved filters, then its saved filters reordered while it was open. A classic dashboard showed nothing new until a reload, and a row's press still opened that row's own saved filter, named in its footer, so its rows did not go out of step.
- Two work items moved into a new team-managed project's own In Progress, and one more in a second new project an hour inside the app's copy of the status list. The next chart's build read the list again (its log line says so), and Aging counted all three among nine in progress.
- Not yet checked on the test site: the line with Reload the dashboard, which a classic dashboard did not show (a new dashboard's widget may); and the history read one work item at a time, forced on the test site's development environment, charting the same work items as the batch read. Jira's hourly limit used up is tested only: forcing it would spend the limit every site shares.
8 October 2026: the chosen view under forced colours, separators on every axis, and whole rows in the status lists
- Under forced colours, such as a Windows contrast theme, the chosen view's tab stands in a frame of the text colour, so you can tell which view is showing. In Jira's light and dark themes the tabs look as before, and no tab moves when you change view.
- Every chart's axis writes 1,000 or more with a thousands separator, as
5,000and1,500 d, as every other count on the tile does: the Histogram's counts and the days under its bars, the Scatter's days, the Trend view's counts and days, the Aging view's ages, and the strips of Compare and of saved filters side by side. The Histogram, the Trend view and the Aging view make room for the wider numbers, so none of theirs runs into an axis title or off the chart, and a strip writes fewer of them where two would meet. - In the settings, the Started statuses and Done statuses lists are as tall as their statuses, up to 300 px. Their box was 208 px, which cut the last of five statuses in half. A longer list is cut between two rows, never through one, and a line under it says so, as in 30 statuses; the list scrolls.
- Checked on the test site:
1,000to5,000on the axis of a chart of 5,000 work items, clear of the axis title; and the settings' status lists on the test site's own statuses, with no row cut in half. - Not yet checked on the test site: inside Jira's frame under a Windows contrast theme or Chrome's forced-colours emulation, the chosen view's frame, with the samples in each key kept; and the line under a status list taller than 300 px.
8 October 2026: history that came from an import
- A work item whose status changes came into Jira through its JSON importer is measured from them. Jira keeps such changes without the field id its bulk history read filters on, so until this update a work item imported with all of its history was counted under never entered a started status, and one imported part-way and then moved on in Jira was measured from its creation or from its first move made in Jira, or counted under never entered a started status, depending on the status the import left it in (how). Not yet checked with a move from Jira Data Center.
- Once a chart shows history that came from an import, every work item of it is read whole, so one whose imported start shows no sign of its own is measured from it too (how).
- Jira's own search does not see status changes that came from an import, so a chart also finds a finished work item by its resolution date in the range. Until this update, a work item created before a chart's range whose move into Done came from an import was not on the chart (how).
- Charts are rebuilt once.
- Checked on the test site: a work item imported with its dated history, charted with those dates: 8 days 5 hours from its imported start to its imported finish, on a chart of eight work items with none excluded.
- Not yet checked on the test site: that work item's rows in Export data (CSV); a chart of work items made in Jira with the same counts as before the update; a work item imported in progress and finished in Jira, measured from its imported start; work items imported, moved back to To Do and then started again or finished in Jira, measured as the documentation says, alone and beside a work item whose history shows the import; a work item imported in progress, aged in Aging from its imported start, with its list; and a work item created before a chart's range whose imported move into Done is inside it, found by its resolution date.
8 October 2026: a tile's settings no longer carry a saved filter's name
- A tile's settings, which Jira hands everyone who can see the dashboard, no longer hold the first saved filter's name; they keep its id. The tile names a saved filter only to someone who can open it, by the name Jira gives them, as before (who sees what). Settings saved before this update hold the name until they are next saved. Every figure, image, summary and export is as it was.
- The security page answers the questions a vendor review usually asks, in one place, and says what the developer has not yet decided.
- Checked on the test site: a tile's settings, saved again, read back with no saved filter's name and the same filter id, and the tile naming its saved filter as before.
- Not yet checked on the test site: an image downloaded before and after the update, compared byte for byte.
7 October 2026: the tile follows a switch of Jira's theme
- Switching Jira between its light and dark themes with a dashboard open now switches the tile too. Before, a switch from dark to light left the tile's background dark under dark text, unreadable until the page was reloaded.
- Nothing else changed: every figure, image, summary and export is as it was.
- Checked on the test site: a switch from light to dark and back again with the dashboard open, on a classic dashboard in Chrome; and, in the dark theme, the Forecast view, the comparison with the period before and the footer.
6 October 2026: the chart higher on the tile, its time zone, build time and method one press away, and the time counted asked
- The footer's first line names the filter, the range and its days, the period compared with, which days are counted and the holidays, and ends in Time zone, build time and method. Press it for the chart's time zone and when the chart was built, and for the method line, as in 50th / 85th / 95th percentile (nearest-rank), rounded up · 1 d = 24 h of counted time, or, on a chart counting working hours, 1 d = 8 h of counted time, the working day 09:00–17:00 (how). It opens closed each time the tile loads.
- While a chart is drawn, Compare with …, Include today and Show the 70% line stand under it, side by side where they fit, above its key, so the chart starts higher on the tile (how). By keyboard they come after the chart's bars and dots. With nothing charted yet, Include today stays under the line that says why.
- Image or summary states the time zone, when the chart was built and the method line, as it did, and its panel now names the time zone among what both state. Every image, summary and export is as it was, byte for byte, and so is every figure on the tile.
- A new tile's settings ask which time is counted, with none of Every hour, Weekdays and Working hours on weekdays chosen, and save once one is (how); until then the field says Choose which time is counted, then save. A tile already set up keeps what it counted, and its settings open on that choice.
- Not yet checked on the test site: none of this update has been checked there yet. To check: the switch opened and closed inside Jira's frame, in the light and dark themes; the chart's top and the keyboard's way to the boxes under it; an image downloaded before and after the update, byte for byte; and a new tile's settings, asking which time is counted, then saved and opened again.
6 October 2026: working hours
- In the settings, Time counted takes the place of Exclude weekends, still one of six settings: Every hour, the default as before; Weekdays, as Exclude weekends was; or Working hours on weekdays, which counts only the time from From to To on Monday to Friday, in the gadget's time zone, one window the same every working day, on the hour or the half hour (how). Gadgets keep what they counted before.
- On a chart counting working hours, a day is one working day, 8 hours for 09:00 to 17:00, in every
figure, bar and line, and the chart says so: its headline and footer say counted 09:00–17:00 on
weekdays, and its method line 1 d = 8 h of counted time, the working day 09:00–17:00. The time
outside the hours is excluded, named
off-hoursin Export data (CSV) and outside working hours in the drill panel (how). A work item started and finished outside the hours, such as an evening fix, is counted to zero and not charted (how). - Such a chart's Export data (CSV) ends with a
working_dayrow, and its Export saved filters (CSV) with aworking_dayline, holding the hours its day counts, and the documentation's recipes divide by those hours where they divide by 24 (how). A chart counting every hour or weekdays states the same figures as before, and its exports are as they were, byte for byte. - The recipes that turn the CSV's hours into days round the quotient to six decimals before rounding
it up, as in
ROUNDUP(ROUND(hours / 24, 6), 1). Without that, a spreadsheet's arithmetic can put a figure exactly on a tenth a hair over it:146.4 / 24rounded up read 6.2 where the chart states 6.1 (how). - Charts are rebuilt once.
- Not yet checked on the test site: none of this update has been checked there yet. To check: the
rebuilt charts' figures and Export data (CSV) against those before the update, byte for byte; a
chart switched to Working hours on weekdays, one work item's drill panel against its Jira history
and its export against the recipes, and back to Every hour; the time fields in Safari and Firefox;
and
=ROUNDUP(ROUND(146.4/24, 6), 1)in a spreadsheet.
6 October 2026: the Flow view, time in status, ages measured when the Aging view opens, and a 70% line
- A new view, Flow, after Aging, shows how many work items were in progress at the end of each day of the chart's range and how many finished each day, as two run charts, with what counts as in progress, the averages, and the lowest and highest day (how). Each day lists its work items, and a day that could count low is not drawn, with the reason above the charts. Every figure can be checked against the export (how).
- In Aging, every list of the work in progress it plots adds In it for, how long each work item has been in the status it is in, and Order lists them longest in status first (how). A sentence under the view's answer, and a table under its key, compare that time with how long charted work items stayed in each status: within what 85% of their stays ended, from 7 stays, and how many work items in progress have been in it longer, each count opening its list (how).
- Measure ages when this view opens, under the Aging view's note, has the ages of your work in progress measured again, as you, each time the view opens, on any date range and at most once a quarter hour, against your chart's own lines. It is remembered for the gadget in your browser, and Export ages now (CSV) gives their file (how; privacy).
- Show the 70% line, beside Show on Histogram, Scatter and Aging, adds a rule at the 70th percentile of cycle time, tagged 70%, with its figure in the key, as in 70% within 4.1 d, and the footer's legend names it while it is shown. The headline and every sentence stay on the 85th. It is remembered for the gadget in your browser, and the export reproduces it (how; privacy).
- When a chart's date range runs to the day it is built, as rolling ranges, This quarter so far,
custom dates to today and Include today do, the app also asks Jira, as you, for the work items
your filter returns for you that moved into one of your Done statuses after the range ended and
before the chart was built, having been in one of your Started statuses, and reads up to 1,000 of
them. Export data (CSV) lists them after the work in progress, one
since_totalrow each, with when each first started and when it last finished (how; privacy). The Flow view counts them in progress on the days before they finished. - Each chart also keeps when each charted work item first started, which the Flow view counts from (privacy).
- Every figure the tile stated before states what it did, and the export's other rows are as they were.
- Charts are rebuilt once.
- Not yet checked on the test site: none of this update has been checked there yet. To check:
that Jira accepts the new search as a viewer and returns a work item moved to Done that morning;
its
since_totalrow in Export data (CSV); the rebuilt charts' counts, charted and excluded, against those before the update; the stays table against the export; ages measured when the Aging view opens, timed, with a second opening in the same quarter hour served without reading again; one day's work in progress on the Flow view against the export, with three of its work items traced in Jira; and the 70% line's figure against the export.
6 October 2026: saved filters side by side set beside the period before, the table as an image or a summary, and each one's files named by its place
- The table of saved filters side by side has Compare with … of its own once every row is ready, as in Compare with Q1 2026: each saved filter's row is set beside its own period before, built for you after the table, three at a time, or one at a time while one holds more than 1,000 work items, and none beside a saved filter that charts nothing (how; privacy). The strips give way to the period before's 85% and, under Against Q1 2026, Faster, Slower, About the same, Too few or No verdict, by the rule a chart's own comparison states, or See its chart where the table's unit would show the two figures another way. Nothing about the comparison is stated across saved filters, and the tick is not kept. Once every row's period before is in, Export saved filters (CSV) adds six columns and the period before's days to its name, and the table's image and summary state the comparison too.
- The table of saved filters side by side has Image or summary in its footer once every row is ready, as a chart has: a PNG of the table with every row, whatever the tile's width, its strips, the lines under it and its footer, and a summary with a line for each saved filter, as text or Markdown (how). Both name each saved filter you can open and count those you cannot, and neither names a work item (privacy). The image of Compare's rows now paints its dots the same way, each where it was before.
- On a tile of several saved filters, a saved filter shown on its own, by pressing its row, adds its
place in the settings to the name of every file it gives: its Export data (CSV), as
cycle-time-ledger-filter-3-2026-04-01-to-2026-06-30.csv, a list exported from a panel, the period before's ledger and the image from Image or summary. The place is thefilter_placeof Export saved filters (CSV), so files downloaded one after another are told apart, and each is matched to its line (how). A tile of one saved filter names its files as before, and no file is named after a saved filter. - The test that decides faster, slower or about the same against the period before is now one piece of the app, shared by every comparison it states.
- Nothing else you see changes: every chart, and the table of saved filters side by side and Export saved filters (CSV) without the comparison, state the same figures as before.
- Charts are rebuilt once.
- Checked on the test site: a tile of two saved filters rebuilt both rows, 445 and 455 work items charted, with no error logged, and drew the sentence across them, how many work items are in more than one, Compare with … above the table with what it does, and Image or summary beside Export saved filters (CSV). Not yet checked there: the rebuilt charts' counts, charted and excluded, against those before the update; a saved filter's files, downloaded from Jira's frame, named by its place; the table's Image or summary opened in Jira's frame, its download and its copies pasted; and the table set beside the period before once Compare with … is ticked, each row's verdict against its own chart's, its file and its copies.
5 October 2026: charts stored more compactly
- Each chart the app keeps for you is written in a more compact form. Nothing you see changes: the numbers, the lists a bar or row opens, and the exported files are as they were.
- Nothing new is stored, and what is stored is kept for as long as before (privacy).
- Charts are rebuilt once.
- Checked on the test site: the first view after the update built the chart and its period before from the new records, with the same counts, charted and excluded, as before, and no error was logged. Not yet checked there: a bar's list exported before and after the update, compared byte for byte.
5 October 2026: several saved filters side by side
- A tile can chart up to 20 saved filters. In the settings, type in Add a saved filter and pick one from the list under the box; ↑ moves a saved filter up the list and ✕ removes it. One saved filter is a chart of that filter, as before. Two to 20 are shown side by side, one row each, in the order the list holds them, all on the same Started and Done statuses, date range and calendar (how).
- Each row is that saved filter's own chart of the work items it returns for you: how many it charts, its 50%, 85% and 95% where it has enough work items, as Compare's rows state them, and a strip with a dot for each work item. To compare teams, components, labels or boards, save one filter for each.
- Nothing is stated across the saved filters until every one you can open is ready. Then the sentence above the table ranges over those with 20 or more work items, the line under the table says how many work items are in more than one of them, and Export saved filters (CSV) gives a line for each row. Nothing is added up across rows.
- Pressing a row shows that saved filter's own chart, with every view, Export data (CSV) and Image or summary, and a button back to the table, such as Back to all 12 side by side.
- A saved filter you cannot open has no row and no name: the line under the table says only how many there are.
- Your first view each day waits for every row to be built for you: eighteen small teams' filters take about a minute or two, and twenty filters of 5,000 work items each can take a quarter of an hour or more.
- Charts are rebuilt once.
- Checked on the test site: a tile of one saved filter rebuilt its chart with the same counts, charted and excluded, as before, and its face unchanged; and the one search that reads a tile's saved filters for you returned five of them exactly as reading each alone does. Later the same day, a tile of three saved filters showed them side by side: each row as it was built, the sentence across them once all three were ready, how many work items are in more than one, and a row over the item cap with its note. Not yet checked there: pressing a row, Export saved filters (CSV), and Add a saved filter itself.
5 October 2026: larger targets, separators and exact By status rows
- Outside the chart itself, the tile's buttons, boxes, tabs and options are at least 24 by 24 pixels, and no two overlap where a row wraps. A view's tab, and an option such as Month, lights up under the pointer, as the other buttons do.
- The tile and its drill panel add no header or footer landmark beside Jira's own.
- Under forced colours, such as a Windows contrast theme, the colour samples in each key, the dots beside the notes and the progress bar keep their colours.
- The tile's page reads Loading… before its script has run.
- Counts of 1,000 or more have a thousands separator, as in 5,000 work items charted, everywhere but the count axes of Histogram, Trend and Forecast. A count of one takes a singular verb, as in 1 has incomplete history in Jira.
- The Scatter view's date labels no longer run into each other: on a quarter, in a tile 640 pixels wide, they step by fortnight.
- By status rounds its rows exactly as the documentation states, in whole seconds, so where two rows lose the same in rounding the row listed first takes the step, and statuses that fall in the same place run by name in one order, whatever language your browser or Jira is in (how). No mean, count or percentile changes.
- A list of names keeps each name whole: a holiday whose name holds a comma or an and is quoted.
- Charts are rebuilt once.
- Checked on the test site: the rebuilt chart has the same counts, charted and excluded, as before, with the count under the headline and the reasons under the chart. Not yet checked there: the targets, the landmarks and the forced colours inside Jira's frame, and whether the frame shows Loading… before the script runs.
5 October 2026: a plainer tile
- By status runs in the order work moves, as far as the app can tell without reading your workflow: To Do statuses first, your Done statuses last. Its rows are rounded in that order, by a rule the documentation states (how).
- The count of work items charted and excluded sits under the headline, and every reason and note is on a line under the chart and its key, always shown. Press the words after the count to go straight to them (the reconciliation line).
- A Compare table states every figure in one unit, its scale's (how).
- The drill panel says excluded time in words, as in 3 d excluded (weekend, public holiday Tag der Deutschen Einheit), where it showed the CSV's code.
- With one issue type or project shown, the Trend view's notes and count title name it.
- The Scatter view's vertical axis has a title.
- Export ledger (CSV) is now Export data (CSV). The file it gives is unchanged.
- Charts are rebuilt once.
- Checked on the test site: the rebuilt chart has the same counts, charted and excluded, as before, with the count under the headline and its reason under the chart's key. Not yet checked there: By status in workflow order, the button on the count line, where each panel opens, Export data (CSV) against a file exported before the update, and Jira's dark theme.
5 October 2026: the Forecast view
- Forecast, a seventh view and the last of the chart views, answers How many by a date and When will this many be done, for a number you type, each at a 50%, 85% and 95% chance, from how many charted work items finished on each day of a history the chart holds (how).
- Nothing in it is random, so everyone who views the chart sees the same figures, and each can be reproduced from the ledger CSV (the recipe).
- A forecast, not a promise: the view says so under every forecast.
- Charts are rebuilt once.
- Checked on the test site: the view is offered after Aging, and the rebuilt chart has the same counts, charted and excluded, as before. Later the same day its face, on a chart whose 445 work items all finished on one of the 90 days sampled: by 1 Nov, not even 1 at 50%, 85% and 95%, which is what a chance of about 27% that any of those 28 days goes like the busy one gives; the pace warning; the reach of 45 days; and Use the 6 in progress. Not yet checked there: its controls, the list a day of its strip opens, and its image.
2 October 2026: the chart as an image or a summary
- Image or summary, in the tile's footer beside Export ledger (CSV), opens a panel with the chart as a PNG 1,920 pixels wide, always on a light background, and its answer as a few lines of plain text or Markdown (how).
- Your browser makes both from what the tile shows. Nothing is asked of Jira or the app, and nothing is sent or stored.
- Checked on the test site: the panel and the image inside Jira, in Chrome.
- Not yet checked there: the download and copy buttons, Safari and Firefox, and Jira's dark theme.
1 October 2026: quarters, custom dates and the period before
- Date range offers Last full month, Last full quarter, This quarter so far and custom dates, beside the last 30, 90, 180 and 365 days. Move both dates on steps custom dates forward by their own length, as from one sprint to the next (date range).
- Compare with …, a box on the tile, sets your chart beside the period before it, with both periods' figures in one table. It says faster or slower only when each period has at least 20 charted work items (how).
- Include today appears on a range that runs into today, and charts up to the current quarter hour instead of the end of yesterday (how).
- A calendar period or custom dates lead the headline, as in In Q2 2026, 85% ….
- Charts are rebuilt once.
- Checked on the test site: a chart saved before the update keeps its range, and the comparison agrees with Jira's own counts. Not yet checked there: a comparison with figures in both periods, because the test site's work items were created from 8 September.
1 October 2026: the Aging view
- Aging, a sixth view, shows the work that has started and not finished: each work item in the status it is in, at the age it has reached, against the 50%, 85% and 95% lines of the finished work. One sentence says how many are older than the 85% line, and every mark and column opens its list (how).
- The ledger CSV holds work in progress in rows of its own, so the view can be reproduced in a spreadsheet.
- Charts are rebuilt once.
- Checked on the test site: the count of work in progress is Jira's own count.
30 September 2026: one issue type or project, or all of them side by side
- Show narrows the whole tile to one issue type or one project, where the chart has more than one (how).
- Compare, a new view, puts one row per issue type or project on one scale, each with its count, its 50th, 85th and 95th percentiles and a dot for every work item (how).
- Every list shows each work item's issue type after its key, and the ledger CSV has two more
columns,
issue_typeandproject. - Charts are rebuilt once.
- Not yet checked on the test site: Show and Compare themselves. Every work item there is a Task, and each filter holds one project, so the tile offers neither.
30 September 2026: the Trend view
- Trend answers "are we getting faster?" in one sentence, comparing the later half of the range with the earlier. It draws the 85th and 50th percentiles of each week, fortnight, month or quarter over a count of the work items finished in each, and every period opens its list (how).
- Charts are rebuilt once.
- Checked on the test site, whose few weeks of history make no realistic trend.
30 September 2026: a clearer chart face
- The tile opens with a one-sentence headline, as in 85% of charted work items finished within 6.2 days of starting, rounded up so that it stays true (why).
- Labelled axes; values when you point at or tab to a bar or dot; By status as one row per status; the drill panel as a table; and a progress bar while a chart builds.
- The tile remembers which view you last chose, in your browser.
- The settings refuse a status ticked as both Started and Done.
- The ledger CSV closes with the four figures the tile states.
- Fixed: the dashboard's frame cut the tile's last line in half.
- Charts are rebuilt once.
27 September 2026: a new name and a new home
- The app is now Stintwise, because two Marketplace apps already used the names Dwell and Dwel.
- The documentation moved to stintwise.adigium.com, and support to stintwise@adigium.com. Links in tiles from before the move still land on the right section.
- Checked on the test site: three tiles opened together, on filters of 10,500 and 6,000 work items, each charting 5,000 and built in 42 to 67 seconds, with no error.
26 September 2026: your chart, from your own search
- Each person's chart is built from their own Jira search, run as them, and is shown only while they can still open what it lists (who sees what). Not yet checked with two accounts of different access.
- A work item imported or created straight into a Done status is counted, under never entered a started status. Before, it was in neither count.
- Everything the app writes for a chart expires on its own (how long data is kept, under How long data is kept).
- Charts are rebuilt once.
8 September 2026: the first version
- First installed on the test site. Until 25 September it was reviewed and corrected many times: on the numbers, on who can see what, on accessibility and contrast, and on the CSV export.