Engineering insights

Engineering insights shows what your teams ship from GitHub: how many pull requests you merge and what each one cost, how quickly work moves from first commit to merge, and how much of it was written with AI. Go to Insights → Engineering insights.

It's the delivery half of what a project returns. Read it next to the project's Value tab, which shows what the work cost and what it moved. To set both up, follow Measure what delivery returns. To see whether AI is changing how your teams deliver, see Compare delivery with and without AI.

You need access to effort reporting to open it, and access to financial figures to see costs. Ask your Flowstate admin. Without cost access, you see days of effort instead of money.

What each tab tells you

Tab	Questions it answers
Overview	How many pull requests did we merge, and what did each one cost? How much of our code comes from bots (Automated %)? Where did the spend go across initiatives, and how is AI cost trending?
Pull requests	Which pull requests shipped, who wrote them, which initiative they were for and what each cost? Which ones aren't linked to any work? See Link pull requests to tickets.
Flow	How long do pull requests take to merge? How long do they wait for a first review, and how long does review take? How big are they, and how much changes after they're opened?
AI impact	How much of what we merge is AI-assisted or written by agents, what does that AI cost per pull request, and how has that changed since the previous period? See AI impact.

Tip

• Flow times are medians, so one very slow pull request doesn't skew them. Bots are left out.
• In a very busy period, the Overview totals cover the most recent pull requests, and a note says so.
• Only AI impact compares each figure with the previous period. On Overview and Flow, read the monthly charts to see change over time.

Read the Overview tab

The Overview covers pull requests opened or merged in the period you choose.

The figures across the top:

Figure	What it tells you
Merged	How many of those pull requests have merged.
Cost / merged PR	What a merged pull request costs on average: the people who worked on it plus the AI they used. If you can't see financial figures, you see Days / merged PR, the average days of effort, instead.
AI cost	What AI cost across those pull requests. Shown only if you can see financial figures.
Automated %	The share of changed lines that came from automated accounts, such as bots, rather than people.

The charts below them:

Chart	What it shows
PR throughput	How many pull requests were opened and merged each month.
Cost per merged PR	The average cost of a merged pull request, month by month. Without access to financial figures it's Effort per merged PR (days).
Spend across initiatives	What the pull requests for each initiative cost each month, stacked so you can compare initiatives. Without access to financial figures it shows days of effort.
AI cost trend	What AI cost each month. Shown only if you can see financial figures.

To see why a pull request's cost isn't the figure finance uses, see Why pull request cost differs from finance figures.

Read the Flow tab

Each figure across the top is the middle value (the median) for the pull requests in the period, leaving out bots.

Figure	What it measures
Median cycle time (h)	Hours from a pull request opening to merging.
Median coding time (h)	Hours from the first commit to the pull request opening.
Median pickup time (h)	Hours from opening to the first review.
Median review time (h)	Hours from the first review to merging.
Median PR size (lines)	Lines added plus lines removed.
Median post-open churn %	The share of a pull request's commits made after it was opened. The higher it is, the more work carried on after opening, for example in response to review.

Read coding, pickup and review time side by side to see where work waits: before a pull request is opened, before someone reviews it, or during review.

The charts below them:

Chart	What it shows
PR throughput	How many pull requests were opened and merged each month, including bots.
Cycle time (open → merge)	The median cycle time for the pull requests that merged in each month.
PR size distribution	How many pull requests fall into each size: XS under 10 lines, S under 50, M under 250, L under 1,000, and XL 1,000 or more.

Narrow a tab to teams, repositories or initiatives

On Overview, Flow or AI impact:

1. Select Filter.
2. Choose Team, Repository or Initiative, then tick the ones you want. Only choices with pull requests in the period are listed.
3. Choose the dates with the period control. It starts on the current financial quarter.

The dates you choose apply to every tab and are remembered in this browser. Each tab keeps its own filters. To reset the filters, select Clear all.

The Pull requests tab has its own filters. See Link pull requests to tickets.

Save a view

Save the filters you use often, so you can come back to them in one click.

1. Set the filters you want.
2. Open the views menu in the toolbar. It shows Default view until you choose a saved view.
3. Select Save current view and enter a View name.
4. To let everyone in your organisation use it, tick Share with organisation, if you see it.
5. Select Save.

Your views are listed under Saved views on that tab. Select one to apply it, or Default view to go back. To rename or delete a view you saved, point to it in the list and use the buttons that appear.

Put delivery next to what it returned

• For one project or initiative: filter a tab by Initiative, then open a project in it and select Value. See Show what a project returns.
• For cost per pull request by team and delivery kind: add the AI Delivery — Cost vs Human Baseline template to a dashboard. See Compare delivery with and without AI.
• For your own charts of merged pull requests, cost per pull request or lines changed: build a report from the Code Delivery measures.

Switching it on

1. Ask your Flowstate contact to switch on Engineering insights. Until then you see a holding screen, and Flow and AI impact show a sample with Talk to us to turn on Engineering insights.
2. Connect GitHub. Until you do, each tab shows Connect GitHub. If you can't connect it, ask your Flowstate admin.

What you get from each connection

You connect	You get
GitHub	Pull request volume, flow times, lines changed and cost per pull request. Code activity can also shape the weekly effort estimate.
GitHub plus Linear or Azure DevOps	Pull requests tied to tickets, and through them to projects and initiatives, so Spend across initiatives fills in.
GitHub plus Jira	Everything GitHub gives on its own. Pull requests aren't tied to Jira tickets automatically — link them to a project or initiative by hand.
AI sessions from the Cloud Proxy	Which pull requests were AI-assisted, AI cost per pull request, and how many lines AI wrote.

GitHub Copilot seats and spend are a separate connection, shown in Agent insights rather than here. See GitHub Copilot.

Why pull request cost differs from finance figures

The cost of a pull request is based on the days the people who committed to it spent on it. It shows as soon as the code arrives: it doesn't wait for weeks to be submitted, and the pull request doesn't need to be on a project.

Capitalisation and R&D claims don't use these figures. They use the effort team leads submit each week. See Which numbers finance uses.

If something's not right

I see a holding screen. Engineering insights isn't switched on for your organisation. Ask your Flowstate contact.

Every tab says Connect GitHub. GitHub isn't connected yet. See Connect GitHub.

Spend across initiatives is empty. No pull requests in the period are tied to an initiative. See Link pull requests to tickets.

I see days instead of money. You don't have access to financial figures. Ask your Flowstate admin.

A filter choice I expected isn't listed. It has no pull requests in the dates you've chosen. Widen the period.