Outlier — what is unusual on your Jira board this morning

A Jira dashboard gadget. It raises a few things worth mentioning at standup, measured against your board's own history rather than a model of a healthy team. Measured on six boards: 27 mornings in 46 on the one running two-week sprints, and 5 in 209 on the five long-running ones, where it is quiet for weeks and says what it looked at instead. Six boards is what we have, not a study — thirty days is enough mornings to find your own number.

What it does

Up to three things worth raising at standup, and says so when there are none. Why

How it decides

From what your own board has finished. No benchmark, no target. The arithmetic

Where it arrives

A dashboard card, or your chat thread on a schedule. The route

What it reads

Three fields and how tickets moved — never the words in a ticket, and nothing leaves your site. The policy

What it costs

Free while your site has ten people or fewer, then per licensed user on your Jira site, not per person who looks at the card: fifty pay $22.50, a thousand $195. The bands

What it takes

An administrator installs once, then anyone adds a card. Thirty days free. Setting it up

A morning on a board that runs sprints#

Two words that card uses. A stay is one run of a ticket through one column, from the day it arrived to the day it left, counted in working days — the norms are built from the ones that finished. Watching 4 of 5 columns means the fifth has no norm yet, usually because too few tickets have finished in it; the card names which, and what it is still waiting for, under how this was counted.

Two examples, from two boards that work differently, because the same arithmetic reads differently on each. The one above is a team on two-week sprints: short columns, work that moves in days.

And the same board on a morning with nothing new#

What it is for#

Two things happen on most boards. Somebody reads the board aloud at standup, row by row, because nobody can tell which rows are worth the room's ten minutes. And a ticket sits in review for a week with nobody noticing, because a week does not look like anything on a board — it looks like a card in a column.

Outlier answers both from one place: it knows what a usual wait looks like here, on this board, from what this board has already finished. Four days in one column is ordinary and in another is three weeks late, and no benchmark knows which is which.

The same product on a board that does not run sprints#

A long-running board — support, maintenance, work that can sit for months. Nothing about the method changes; the numbers are simply bigger, and a queue that has stood for eighty working days is a different conversation from one that has stood for nine.

One card watches one board. It is an ordinary dashboard gadget, so it takes a column of a dashboard and sits beside whatever else is there — and a board carrying two streams takes two cards, each narrowed to its own labels, which is also how two teams on one board keep separate norms.

It can reach your channel without anybody opening a dashboard#

The cards above sit on a Jira dashboard, and a dashboard has to be opened. Ask the people on your team how often they open one and you have the honest answer, which is why this is on the front page rather than in the setup notes: for most teams it decides whether the thing is a tool or a tab nobody visits.

So the same morning can arrive in the channel where your standup already happens. In Jira Automation, a rule with a scheduled trigger adds the step Read this morning's questions, picks which card to read, and posts what it gets. Nobody presses anything.

We do not deliver it, and that is a design decision rather than a gap. Your rule asks, on your schedule, into your channel, under your credentials — an app that posted to Slack itself would need an address outside Atlassian, and this one deliberately has none. The rule can also be told to stay silent on mornings with nothing to raise, or to post the quiet card so the channel shows the board was read.

One thing that follows, said here rather than discovered later: a channel that goes quiet looks the same whether the board was calm or the collection broke. The setup page has the rule, the smart values it provides, and the two ways to tell those apart.

What the first morning looks like#

There is no watching period to sit through. The board's history is read when the gadget is configured — minutes on a small board, up to an hour on a large one — and the norms come out of that history, so the first morning's card is the whole product: the findings, the columns behind them, and the footer saying how much of the board it looked at. A column that has not finished enough work to have a norm is the exception, and the card names it and says how far along it is.

What time adds is the card's own record: a row of squares for which of the last ten mornings it had something to raise. Until that exists the card also says what it would have said over the board's last quarter — on a morning with nothing new to raise, where there is room to read it — reconstructed from the same history and measured by the norms in use now. It is labelled that way every time it appears — there was no card on those mornings — and it stops being shown once ten mornings of squares can make the point from evidence instead. On a board where no column has a norm yet there is nothing to reconstruct, and the card says that rather than reporting a quiet quarter it cannot see.

Jira already has reports#

It does, and they answer a different question. A control chart or a cycle time report is a place you go: somebody opens it, chooses a window, and reads a distribution. It tells you how the board has behaved over a quarter, and it tells you the same thing whether or not anything happened last night.

This is the other half. It comes to you, every morning, and its subject is this morning: which few things are unusual for this board right now, put as a question, in a card small enough to read before a call — and a footer that says what it looked at, so a morning with nothing on it is an answer rather than an empty chart. A report will not tell you that a review queue has not cleared since Tuesday unless somebody thinks to go and look.

They are not alternatives, and if the reports already get read every morning then this is not for you.

What it costs#

Free while your Jira site has ten people or fewer. Past that every licensed user is charged, the ten included, on a graduated scale — each band pays for the people inside it, so a large site pays the low rate on most of its headcount rather than the entry rate on all of it. Thirty days to try it, and no permanent free mode: what is free is the entry tier and the trial.

Ten users or fewer pay nothing at all — that is a flat price for the whole site rather than the first row of this scale. From the eleventh person on, the scale is charged on every user, those ten included.

So a site of 15 people pays $6.75 a month, one of 50 pays $22.50, one of 120 pays $51, and one of 1,000 pays $195 — for the whole site, however many boards carry a card. Atlassian bills it with the rest of your apps.

The count is your Jira site's licensed users — the number Atlassian already bills your Jira on, which includes everybody who will never open a dashboard. It is not the size of the team using the card, and it cannot be bought for one team: this is billed through Atlassian, against the site.

So a large site with one interested team pays on its headcount rather than on its use, and that is the reason the bands above stop rising after 250 people. Worth knowing before you ask anybody for a budget, because it is the difference between quoting the rate and quoting the bill.

Atlassian sells it on two cycles, and they are not the same arithmetic. Paid monthly, it is the scale above on the users your site has. Paid annually, it is a fixed tier rounded up: a site of 100 pays $450 a year, and a site of 120 pays $750, because the tier above 100 runs to 200 and the whole tier costs one price. So the rate here answers a monthly bill exactly, and an annual one only for a site that lands on a tier boundary — the calculator on the Marketplace listing is what answers for your own number.

What it will not do#

• It measures no person. Nothing it derives or stores is about one, checked on every record before it is stored — Jira's changelog hands over the name of whoever made each change, and it is dropped where the response is read. The board's saved filter is not kept either: the board, the filter and the labels are enough to compose the query when it is needed, and a filter quite ordinarily names people.
• It gives no verdicts. Every finding is put as a question, because a queue standing still can be a problem or one large ticket, and only the people in the room know which. A question two findings would ask word for word is printed once on the card — and on every line of the text you paste, where the lines are read apart.
• It says how much it looked at. Three findings at most, always with a denominator — so a morning with nothing on it is a result rather than an absence.

Who is behind it, and what happens if that stops#

One person. That is the whole answer, and the rest of this section is what it does and does not mean for a board that has this on it.

Nothing here runs on anything of ours. There is no vendor backend, no service to keep paid for and nothing that calls home, so there is no switch that gets thrown when somebody stops working on it — the app runs inside your Atlassian site and keeps doing so. Atlassian's own rule for an app taken off the Marketplace is that it stops being findable by new customers while installations that already exist keep working and keep renewing. And what it has read is in your site's storage, where removing the gadget removes it.

What would stop is new versions, fixes and answered email. We make no promise beyond that today — not an escrow, not a successor, not a date. If that is a condition for you, it is a fair one to have, and this is not yet the product that meets it.

What it reads#

Three fields per ticket and their changelog: when things moved, and between which columns. Plus a board's labels while somebody is configuring the gadget, and the name of the open sprint where the card is framed by sprints — Method lists every field. Nothing it reads leaves your Atlassian site — there is no vendor backend, no analytics and no model in the pipeline. Collection runs on the app's own hourly timer rather than when somebody opens a dashboard. Cloud only, built on Atlassian Forge.

One control is outside that sentence and belongs beside it: Ask Rovo why hands a single question to your site's own Rovo, on your credits, if your site has Rovo at all. Somebody presses it, the answer goes to them, and nothing about it comes back to us.

The documents#

All of them are public, and Method is public on purpose: a method nobody can read is indistinguishable from a model nobody can audit.

• Getting started — what the setup looks like, and how to read the card.
• Method — the arithmetic behind every number on it.
• Security — how the app is built and operated, and how to report a flaw.
• Privacy — what is read, stored and kept, and for how long.

The listing is on the Atlassian Marketplace, which is where the app installs from. Anything you would rather ask first goes to support@chmut.com.