Appearance

Measure what delivery returns ​

When you've finished this checklist, you can answer three questions for any project:

  • What did it cost, in people and in AI?
  • Did it move what it set out to move?
  • Is delivery changing as your teams use more AI?

You read the answers on three screens:

  • a project's Value tab;
  • Insights → Engineering insights;
  • Insights → Analytics & reporting.

A head of delivery or engineering lead owns this checklist, with a finance partner. A GitHub organisation owner and a Flowstate admin help with the engineering steps.

This isn't only for engineering. A support team's AI assistant project can have drivers such as contact deflection or handle time, and its Value tab works the same way. Steps 3 to 5 are only for teams whose work lands in GitHub.

Before you start ​

1. Put projects and initiatives in place ​

What: Every project you want to measure has an owner, start and end dates, and belongs to an initiative. Who: Engineering or delivery lead, with project owners. Where: Delivery → Roadmap → Projects and Initiatives. See Set up your projects, steps 2 and 3. Done when: Each project you'll report on shows an Owner, Start Date, End Date and an initiative, not No initiative.

2. Report effort, so projects carry actual cost ​

What: Team leads confirm each week where their team's time went, so each project's cost comes from real effort and not only from the plan. Who: Whoever runs effort reporting, with team leads. Where: Set up weekly effort reporting. Done when: Weeks are submitted for the teams on these projects. Each project's Value tab shows Cost to build, not "No actual or forecast cost is on record for this project yet".

3. Connect GitHub and match contributors ​

What: Install Flowstate's GitHub App, then match each GitHub contributor to a Flowstate employee or contractor. Who: A GitHub organisation owner, or someone allowed to install apps there, with a Flowstate admin. Where: Settings → Integrations → GitHub. Use step Connection → Connect GitHub, then step Map contributors. See Connect GitHub. Done when: Verify health shows Healthy with Backfill completed, and Map contributors says "Every active contributor is mapped".

  • You need access to integrations to connect it. Ask your Flowstate admin.
  • The App can only read from GitHub. You can install it on all repositories or only the ones you choose.
  • After you connect, Flowstate brings in the last 8 weeks of pull requests and commits. Run backfill under Verify health does it again.
  • Flowstate matches contributors to people by email address. Bots and automated accounts aren't counted. A contributor who isn't matched still appears in pull request counts, and matching them later also updates their past activity.

4. Link pull requests to tickets ​

What: Let Flowstate tie each pull request to its ticket, and through the ticket to a project and initiative. Who: Engineering lead, with your engineers. Where: Connect your project tool and link its projects. Engineers mention the ticket in the branch name, pull request title or description, or a commit message. Done when: Most merged pull requests on Insights → Engineering insights → Pull requests show an initiative. The Linkage → Unlinked filter only holds work you expect.

ToolWhat links a pull request to a ticket
LinearThe issue key, such as FLO-123, in the branch, title, description or a commit message. Linear's own pull request links count too.
Azure DevOpsAB# followed by the work item number anywhere, or the work item number at the start of the branch name.

You can't map a whole repository to a project. See Link pull requests to tickets.

Not available yet: Jira ticket links

Pull requests aren't linked to Jira issues automatically. With Jira connected, Engineering insights still shows pull requests, their cost, flow and AI impact. Link each pull request to its project or initiative by hand.

5. Capture AI sessions, for AI impact ​

What: Record your engineers' AI sessions, so Flowstate can match them to the pull requests they helped with. Who: IT and security, with an engineering lead. Where: Track AI in your workflows. Done when: Insights → Engineering insights → AI impact shows AI-assisted pull requests.

Without AI sessions, AI impact shows only pull requests written by agents.

6. Set each project's value: its drivers and targets ​

What: Decide what each project is meant to move. Add those drivers to the project and get each one's starting reading, target and target date recorded. Who: Engineering or delivery lead and project owners choose the drivers. A Flowstate admin sets up driver types. Where: Settings → Delivery → Driver Types. Then, on the project's full page, open the Driver picker on Overview. See Show what a project returns. Done when: Each project's Value tab lists its drivers, with a Before and Target on each.

  • Pair every speed or cost driver with a quality one that should stay where it started, such as satisfaction or incidents. This is a guardrail: its verdict is Held or Regressed.
  • If a business metric covers the project's team, the Business metric column links to it. See Business metrics.

Not available yet

You can't yet set a driver's reading before the work, its target, its target date or which way is better in the app. Send them to your Flowstate contact, who sets them for you. Until a driver has readings from before and after the work, it shows Not read.

7. Give people access ​

What: Decide who can open each screen, and who sees cost. Who: Flowstate admin. Where: Settings → Users & Access → Roles. See Roles and permissions. Done when: An engineering manager can open every Engineering insights tab and the projects they own. They see cost only if you meant them to.

  • People whose role can open effort reporting can open Engineering insights.
  • People who can't see financial figures see days of effort instead of money.
  • Recording an outcome on the Value tab needs Update Drivers. Adding or removing drivers also needs access to manage drivers.

8. Read delivery and value together ​

What: For each project, read what it cost and what it moved. Then read how its initiative's pull requests flowed over the same months. Who: Head of delivery or CTO, with the finance partner. Where:

  • The project's Value tab.
  • Insights → Engineering insights, filtered by Initiative.
  • A dashboard with the AI Delivery — Cost vs Human Baseline template. See Compare delivery with and without AI.

Done when: For a finished project you can say what it cost, what share was AI, and which drivers beat, held or missed their targets. You can also see how its initiative's pull requests moved in the same period.

When a driver's target date passes with no reading after the work, the Value tab asks you to Record an outcome →.

Read the result carefully

  • Capacity isn't cash. A driver that improved has freed up time. It only becomes a saving when headcount comes down or the team takes on more work. More AI on top of the same team adds cost.
  • Empty isn't zero. Not read means nobody recorded a reading. It doesn't mean nothing moved.
  • Don't read speed alone. More merged pull requests can simply mean more, smaller pull requests. Read throughput with cycle time, the project's verdicts and its cost.

You're set up when ​

  • Projects you report on have an owner, dates and an initiative, and effort is submitted for their teams.
  • Verify health on Settings → Integrations → GitHub shows Healthy, and every active contributor is mapped.
  • With Linear or Azure DevOps connected, most merged pull requests show an initiative.
  • Flow and AI impact show your own data, including AI-assisted pull requests.
  • Each project's Value tab lists its drivers with a Before and Target.

To have GitHub shape effort as well, see step 10 of weekly effort reporting.

Next: day-to-day guides ​

Flowstate Documentation

Copyright 2026 Flowstate.