Cost and value
Redline could prove review works and could not say what it cost. Cost per BLOCKER caught is the number that closes that, and no other tool in the chain can compute it.
Why nobody else can compute it
A cost-management tool knows spend and has no findings, so it cannot compute value. A DORA tool has neither. Redline knows which findings were acted on. Joining the two is the whole argument.
$ npx redlinegate metrics roi --data data --days 90 \ --spend-total 850 --spend-grain org
It writes roi.html — one page a finance stakeholder can read, sourced entirely from collected data.
Value is what was acted on
Not what was reported. A finding nobody acted on caught nothing, and counting it would let the return be inflated by producing more noise — precisely the behaviour the guardrails exist to prevent.
The refusals are the design
Every figure here will be quoted at someone who acts on it, so each one that cannot be sourced says so instead of being approximated.
- An org-level spend figure will not answer a per-repository question. Org spend divided by repository count and presented as per-repo cost looks precise, is invented, and is exactly the number a stakeholder would act on. Redline says to publish org-level cost against org-level value instead.
- Deployment frequency is unknown, not zero, where a repository does not use the deployments API. Assuming one deploy per merge reports a trunk-based team and a quarterly-release team as identical — the exact distinction the metric draws.
- MTTR is refused by name, so nobody wonders whether it was forgotten. It needs an incident feed Redline does not have and should not acquire.
- Change failure rate travels with its caveat. It is a floor, not the true rate: a revert or hotfix is evidence of a failed change rather than proof, and a team that fixes forward without saying “hotfix” scores better than one that labels honestly.
The spend figure is yours to supply
No script here reads a vendor's billing API, and nothing on the page is estimated. You pass the figure from the assistant vendor's own usage reporting, and Redline records its grain — per repository or org-wide — alongside it, so a later reader knows what it can and cannot answer.
Delivery metrics, from data already collected
Lead time and change failure rate come from the merged pull requests the nightly collector already walks — see Telemetry. Nothing new is instrumented in a product repository to produce them.