Skip to content

Lesson 1 of 114 min video5 question quiz

Find the automation rules eating your step allowance

A fourteen-minute lesson that finds the automation flow spending the most steps on a Jira demo site and cuts it down four ways: a tighter trigger, a JQL condition first, the right scope and fewer branches. On the way it explains how Atlassian counts steps and why branches multiply them, and shows the same twelve edits going from 84 steps to 18.

Level
Intermediate
Products
Jira
Format
On demand, free, no sign-up

Watch the lesson Take the quiz

How steps are counted

Watch this from 1:03

Every part that runs is a step: trigger, condition, action and branch, with Atlassian's example of a trigger, a condition and two actions as four steps

An automation step is any part of a flow that runs: the trigger, each condition, each action and each branch. Atlassian's own example is a trigger, one condition and two actions, which is four steps.

Part of the flowWhat it costs
TriggerOne step every time it fires, even when it finds nothing to do
ConditionOne step. When it fails, nothing after it runs, so nothing after it counts
ActionOne step each time it runs
BranchOne step, plus everything inside it once for every work item it finds

Branches are where steps pile up. A branch over five child work items, with three actions inside, is fifteen action steps plus the branch itself.

Watch out. Before, a run only cost something if it did something. Now every step that runs counts, so a flow that starts on every edit and then finds nothing still spends its trigger and every branch it enters.

The allowance is pooled across the whole organization: every app and every site draws from one monthly pool, and unused steps do not roll over. Paid Jira plans include 400, 750 or 1,000 steps per user per month, by plan. Extra steps are charged per 1,000, extra usage is on by default, and admins get alerts at 80 and 100 percent. Billing for extra steps starts on 3 December 2026, or at your first renewal after that date. Sources: Atlassian, How is my usage calculated? and usage-based pricing.

Find the noisy flow

Watch this from 2:37

The Usage tab on the upcoming usage model: 109 steps this month, and high usage flows sorted by automation steps, the noisy flow first at 13 runs and 103 steps
  1. Open Jira settings, then System. Under Automation, choose Global automation.
  2. Open the Usage tab and switch to the upcoming usage model. That is the steps view.
  3. Read the two cards: automation steps this month, split by app, and Rovo credits used by automation. Both are estimates.
  4. Under high usage flows, sort by automation steps, highest first. The flow at the top is where to start.
Flow on the demo siteFlow runsAutomation stepsSteps a run
Keep child work items in sync with their epic13103Almost 8
The other flows1 each2 to 42 to 4

Watch out. The current usage model counts only runs that did something. On it, the same flow in the same month showed four runs, which is why it never looked like a problem. Switch to the upcoming model to see steps.

To see what one run spent, open the flow, then its audit log, and expand a run. A story edit fired the trigger, then three branches ran and found nothing to do: four steps for an edit the flow never needed to see.

Read the flow

Watch this from 4:16

One edit counted: an epic edit with four children costs 16 steps, a story edit with no children costs 4

Read a flow from the top, one step at a time, and ask what each part lets in.

PartOn the demo flowWhat it means
TriggerWork item updatedAn edit to any field, on any work item, starts the flow
ConditionNoneEvery edit goes straight into the branches
Branch 1For: Children, copy priority from the epicRuns once for every child
Branch 2For: Children, copy due dateLoops over the same children again
Branch 3For: Children, add a commentLoops over them a third time
ScopeGlobalAn edit in any space starts the flow, even spaces with no epics
One editTriggerBranchesActionsSteps
An epic with four children1312, three per child16
A story, no children13, empty04, and nothing changes

The scope is under the menu at the top right of the flow, then Settings.

A tighter trigger

Watch this from 5:53

Pick the narrowest trigger: work item updated listens to any field, field value changed listens to the fields you name
  1. Select the trigger. The panel on the right shows what it listens to: every update.
  2. Click the delete icon on the trigger. The list of triggers opens.
  3. Search for field value changed and select it.
  4. In the fields to watch, type priority and select it, then type due date and select that too.
  5. Leave change type on any changes to the field value. That covers a value being added, removed or changed.
  6. Under For, choose edit work item. Creating a work item, or moving it through the workflow, no longer starts the flow.
  7. Click next. The trigger card reads: value changes for priority, due date.
TriggerListens to
Work item updatedAny field, any edit
Field value changedThe fields you name, for the operations you choose

Rule of thumb. Pick the narrowest trigger that still catches the change you care about. For scheduled flows, the same idea: put the JQL in the trigger, run hourly instead of every five minutes, and include only work items that changed since the last run.

A JQL condition first

Watch this from 7:34

Order matters: a JQL condition first stops a story edit at 2 steps, the same condition at the bottom costs 5 steps or more
  1. Click the plus just below the trigger card.
  2. Choose condition, then JQL condition.
  3. Enter the query for the work items this flow is for. On the demo site, open epics only: issuetype = Epic AND statusCategory != Done
  4. Click validate query. Jira confirms the query works and shows how many work items match.
  5. Click next to keep it.
Where the condition sitsA story edit costs
Right after the trigger2 steps: the trigger and the condition. Nothing below it runs
At the bottom of the flow5 steps or more: every branch runs first

One JQL condition can check type, status and anything else JQL knows, in a single step.

The right scope

Watch this from 8:40

Scope to the spaces that use it: single space, multiple spaces or space type, instead of global
  1. Open the menu at the top right of the flow, then Settings.
  2. Open the scope list and choose single space.
  3. Choose the space: type its name and pick it. On the demo site, Customer onboarding.

An edit in any other space now never reaches the flow. Atlassian's own advice is to avoid global scope wherever you can.

Who uses the flowScope
One spaceSingle space
A few named spacesMultiple spaces
Every space of one kindSpace type

Fewer branches

Watch this from 9:30

One loop, not three: three branches with three actions per child become one branch with one action that sets both fields
  1. Zoom out on the canvas until all three branches fit.
  2. Open the edit action in the first branch. It sets the priority.
  3. Open choose fields to set and add due date.
  4. Open the due date's menu, choose copy, then pick trigger work item: the epic that changed.
  5. Click next. The one action now copies both fields.
  6. Point at the second branch and click its delete icon.
  7. Delete the third branch, the comment, the same way. The change already shows in each child's history.
BranchesActions for every child
Before33
After11, setting both fields

Rule of thumb. When several actions loop over the same work items, put them in one branch. When one action can set several fields, use one action.

Save and measure

Watch this from 10:50

The same 12 edits before and after the fix: 84 steps before, 18 after
  1. Click save. The fixed flow is live.
  2. Test it: open an epic in the space the flow is scoped to and change its priority.
  3. Back in Global automation, open the flow, then its audit log, and open the new run.
  4. Expand each step: the trigger, the condition, the branch and the edit inside it.
The test editSteps
Trigger: field value changed, on the epic1
JQL condition, passed1
Branch1
One edit for each of the four children4
Total7, where the same edit cost 16 before

Then the same twelve edits were replayed on the demo site. Before the fix they cost 84 steps; after it, 18. The epic edits still did their job. The rest stopped at the condition or never started the flow at all.

Make it a habit. Run the same check on your top five flows every month.

Check what stuck

5 questions on this lesson. Pick an answer to see whether it is right, and why. Nothing is sent anywhere and there is nothing to sign up for.

  1. Question 1 of 5A flow has a trigger, one condition and two actions, and everything runs. How many steps?

    The answer: 4.

    Every part that runs is one step: the trigger, the condition and each of the two actions.

  2. Question 2 of 5A condition fails halfway down a flow. Do the steps after it count?

    The answer: No.

    They never run, so they never count. The trigger and the condition that failed still do.

  3. Question 3 of 5A branch finds 6 child work items, with 2 actions inside. How many action steps?

    The answer: 12.

    Everything inside a branch runs once per work item it finds: two actions, six times. The branch itself adds one more.

  4. Question 4 of 5Your flow only cares about priority. Which trigger?

    The answer: Field value changed.

    Field value changed, watching priority, stays quiet for every other edit. Work item updated starts the flow on any field.

  5. Question 5 of 5Where do you see which flows spend the most steps?

    The answer: Global automation, Usage tab, upcoming usage model.

    The upcoming usage model counts steps. Sort high usage flows by automation steps. The current model counts only runs that did something.

Who teaches it

Riley Venable

Riley Venable

VP of Services, Atlas Bench

Full-stack engineer, Atlassian Certified Instructor, and Atlassian Community Champion who works on the layer underneath the tools: identity, infrastructure, and the systems that decide whether AI can be trusted with them.

Build this on your own data.

The report in this lesson runs on demo data. An Atlas Bench architect can build the same reporting on your Focus, Talent and Jira data, with each connection scoped and each dashboard shared on purpose.