Metrics you did not have to collect.
GroundTruth is already reading every branch, pull request, review and merge to keep your tracker true. The same ground truth produces your DORA metrics and your AI spend, with no survey, no instrumentation, and nothing self-reported.
The four keys, from the code itself.
Not typed into a form at the end of a sprint. Derived from what actually merged.
Merged pull requests, read straight from source control.
Pull request opened to merged, median.
Reverted merges over merges.
High-severity findings, raised to resolved, median.
Every metric states its basis. Deployment frequency counts merges as the deploy proxy, so trunk-based teams read true and teams batching releases read conservative. We would rather show you the definition than a flattering number.
Know what the agents cost you.
In the ADLC your model bill becomes a real line item. GroundTruth ledgers every model call it makes on your behalf: which feature, which model, which ticket, how many tokens, and what it cost. Attributable spend, not one opaque invoice.
- Per feature, per model, per day
- Attributed to the ticket that triggered it
- Projected monthly run rate
- Cached by commit, so re-runs cost nothing
| Semantic acceptance criteria | 1,284 calls | $18.42 |
| Acceptance-criteria drafting | 206 calls | $4.930 |
| Total this month | 1,490 calls | $23.35 |
Daily spend, last 12 days
Reports that assemble themselves.
Delivery risk
Every open ticket scored for slippage from unmet criteria, cross-team dependencies and stalls, ranked worst first.
Portfolio roll-up
Initiative to epic to story health with true completion numbers, drillable to the story that is dragging it.
Release readiness
One go or no-go verdict: what claims done but is not, what merged without review, what is still at risk.
Guarantee scorecard
Live measurement of the accuracy and write-safety guarantees your plan is backed by.
Exec digest
Weekly and monthly summaries delivered to Slack or Teams on your cadence, with nobody assembling them.
Ask the ground truth
"What is blocking us?", "Are we ready to release?", answered in channel from the same data.