Stop guessing how the team delivers. Measure it.

Q: Where does the data come from - do I install an agent?

No agent. Data is read from whatever you already connected: your Git provider, your Jira link, your deploy and incident integrations. Analytics builds on the sources that are already there.

Q: Is this a developer surveillance tool?

No. The default views are team- and project-level; the per-person Developers view is opt-in per project and service accounts are filtered out automatically. The goal is finding the jam in the flow, not grading people.

Q: Why do CFR and MTTR show empty?

Because your incident source is not connected yet. These two metrics need real incident records; they start filling once Dynatrace or New Relic is linked. Showing empty beats showing an invented number - that is our honest-reporting rule.

Q: What does AI Impact actually measure?

Two things, side by side. First, what the installation's AI key was used for: fix suggestions, assistant answers and comment classification, with input and output tokens per person, per model and per week. Second, the quality signals over the same window: the ArchScan score, the share of merged MRs that are fixes, and average merge time, first half against second half. It does not guess which lines an AI tool wrote.

Q: Which plans include it, and does history backfill?

Analytics comes with every licence; no module is gated. Whatever history your sources expose (MRs, deployments, Jira changelog) is pulled in on first sync - you do not start day one with an empty board.

Engineering Analytics rolls MR flow, DORA metrics, delivery flow and AI impact into one board - the data comes from the Git, Jira and incident sources you already connected, no extra setup.

Screens from a running installation

What is on the board?

Seven reports, one data model. All of them feed off sources you already connected; no separate agent or plugin to install.

MR Insights

Review time, MR size, merge velocity and bottlenecks - broken down by team and project.

DORA

The four key metrics, measured from your deploy and incident sources; an incident within an hour of a deploy counts against it.

AI Impact

What the AI key was spent on, per person and per model, next to how the score, fix share and merge time moved over the same weeks.

Delivery flow

Work-item flow read from Jira's lifecycle: wait times, backflows and seven ready-made alarms.

Findings funnel

Every module's findings in one funnel from opened to resolved: what piles up, what gets fixed, what gets ignored.

Portfolio

Every project in one table: score, open findings, trend. The screen a lead opens on Monday morning.

Developers

Git members and MR authors sync automatically; profile, contribution and an ArchPilot-based score. Opt-in per project.

DORA, not the marketing version

All four metrics are measured from real sources - no surveys, no estimates, no hand-typed numbers.

Deploy frequency

Counted from your deployment records (Argo CD / Kubernetes).

Lead time for changes

From MR opened to deployed, split into waiting for review, review and waiting to ship.

CFR

Change failure rate

Fills in once your incident source (Dynatrace / New Relic) is connected.

MTTR

Mean time to recovery

From incident open to close, attributed by time correlation.

The honest-reporting rule

If there is no data to feed a metric, that metric shows empty - it is never padded with an invented average or an industry guess. Every number you take to management has a real record behind it.

The thresholds behind every alarm

Merge request data is synced every 15 minutes on all four Git providers. An alarm turns amber, then red, at fixed and visible thresholds.

Alarm	Warning	Critical
Large MRs (over 400 lines)	20% share	40% share
Self-merged	20% share	40% share
Merged without approval	- share	30% share
Merged without review	25% share	50% share
Time to first review	24 h	72 h
Time to merge	72 h	168 h
Open longer than 14 days	listed

Risky merges, scored

Each reason adds points; a merge is listed at 4 or more, and the ten riskiest are shown.

Over 400 lines+2
Over 1000 lines+3
Self-merged+3
No approval+2
No review+2
Critical finding+3
Error finding+1

How DORA bands are drawn

Lead time is cut into three waits so you can see which one grows. It shows a number only once at least five changes in the window have been measured.

	Elite	High	Medium	Low
Deploys per week	≥ 7	≥ 1	≥ 0.23	< 0.23
Lead time	< 24 h	< 168 h	< 730 h	≥ 730 h

Change failure rate and time to restore count an incident against a deploy when it opens within an hour of it on the same service; restore time uses resolved incidents only. Incidents are read live and never stored; an unreachable source shows as unreachable, never as zero incidents.

Whose question does this board answer?

Three meetings, three concrete moments.

Sprint retro

Why did review climb to three days?

MR Insights shows in which week, in which project, at which MR size the waiting started. The conversation moves from feelings to data.

Quarterly review

Is the AI investment paying off?

AI Impact shows what the installation's AI key was spent on - suggestions, assistant answers, tokens, per person and per model - next to how the score, the share of fix MRs and merge time moved over the same weeks. The answer is a table, not a slogan.

Board update

Four DORA metrics, one slide

DF, LT, CFR and MTTR come from real deploy and incident records; every number drills down to its source. When an auditor asks, the answer is already there.

Frequently asked

Where does the data come from - do I install an agent?

No agent. Data is read from whatever you already connected: your Git provider, your Jira link, your deploy and incident integrations. Analytics builds on the sources that are already there.

Is this a developer surveillance tool?

No. The default views are team- and project-level; the per-person Developers view is opt-in per project and service accounts are filtered out automatically. The goal is finding the jam in the flow, not grading people.

Why do CFR and MTTR show empty?

Because your incident source is not connected yet. These two metrics need real incident records; they start filling once Dynatrace or New Relic is linked. Showing empty beats showing an invented number - that is our honest-reporting rule.

What does AI Impact actually measure?

Two things, side by side. First, what the installation's AI key was used for: fix suggestions, assistant answers and comment classification, with input and output tokens per person, per model and per week. Second, the quality signals over the same window: the ArchScan score, the share of merged MRs that are fixes, and average merge time, first half against second half. It does not guess which lines an AI tool wrote.

Which plans include it, and does history backfill?

Analytics comes with every licence; no module is gated. Whatever history your sources expose (MRs, deployments, Jira changelog) is pulled in on first sync - you do not start day one with an empty board.