Sample deliverable · engineering capital statement

A real assessment.

Not a list of what's missing — anyone can produce that. Where the engineering time actually went, and what we told the client to stop.

Engineering capital statementPublished with client consent

Engineering time assessed

72 days

Measured from commit history

Spent on the launch blocker

0 days

Out of 72

Headline decision

Stop

You need less, not more

01 Where the time went

Measured, not asserted.

Derived from commit history and code volume, not from the client's self-report. Proxies have weaknesses, so the first figure is a range rather than a point.

Verifying the code32–53%

1,117 of 3,471 commits touch tests. Test code is 53% of everything written.

Shipping, running, recovering it0.63%

22 of 3,471 commits touch CI, deploy config or security rules.

The thing actually blocking launch0%

No content reviewer hired. The corpus file is still named 'for-review'.

Coverage is 99.18% — world-class — on a product that has never met a user. That is not an achievement with some gaps beside it. It is a misallocation, and it is invisible precisely because the over-investment is the thing the team is proudest of.

02 Findings

Two. Each one produces a decision.

A finding a competent engineer could produce in an hour with repo access is not intelligence, and we should not be paid for it.

F-01

72 days deployed · 0 days to the constraintSTOP · HIREConfidence: High

The constraint is not engineering — and it is being solved with engineering

The launch slipped by two months. In the period since it slipped, engineering velocity rose to ~99 commits/day. Meanwhile the product's core content remains unreviewed, because the reviewer has never been hired. Test code cannot produce a corpus. Refactoring cannot produce a reviewer.

Reallocation — From: further engineering. To: a content reviewer — the only line item that moves the launch date.

F-02

~50:1 · verification over operabilitySTOP · INVESTConfidence: Medium

Effort is allocated roughly 50:1 toward proving the code over being able to run it

Coverage stands at 99.18% — genuinely world-class — on a product that has never met a user, has no alerting, no tested restore, no security scanning and zero release tags. The marginal test is worth approximately nothing. The first tested restore is worth the product.

Reallocation — From: test and coverage work. To: an operational baseline — three days of work, against the many weeks already spent on tests.

03 Decisions

Ranked by capacity moved, not by severity.

STOPCease test-coverage work. You are at 99.18%.Releases the largest single block of capacity

HIREA content reviewer. This is the binding constraint.The only spend that moves the launch

PROVERestore from a backup, into a scratch project. Today.1 day

INVESTOperational baseline — alerting, uptime, restore.3 days

04 The inputs

Every number, and where it came from.

Shown to the client for correction before the findings were written, not after.

Engineering days in current sprint72Git historyMEASURED

Test share of codebase53.1%216,705 vs 191,167 linesMEASURED

Commits touching tests32%git logMEASURED

Commits touching CI/deploy0.63%git logMEASURED

Test coverage99.18%coverage-summary.jsonMEASURED

Effort before this sprint—Never establishedGAP

Allocation is stated in engineering-days rather than currency. Converting days to money requires the client's cost basis, which is theirs and is withheld here. In the delivered report it is present, and the same ratios carry a figure.

05 Limits

What we couldn't see, and what we got wrong.

This section ships in every report. A gap we hide is a gap the client pays for later.

Effort before the current sprintMissingIntensity of the earlier phase was never established, so total effort to date cannot be stated honestly.

Business model, funding, runwayNot askedWe ran out of engagement before asking. Our omission, not a client refusal.

Production status and alertingMissingSentry is installed; whether alerting is configured was never confirmed.

331 commits mentioning monitoring/backupUnresolvedNo runbook or incident record exists in the repo. Either the work left no artifact, or the keyword proxy is matching unrelated commits. Not resolved, and not used as a finding.

Two claims we withdrewCorrectedWe claimed the content corpus was unrecoverable, and we put a currency figure on total capital. The client corrected both: the corpus does not substantially exist yet, and the currency figure rested on an assumed day-rate. Both were withdrawn before this report was finalised.

This is the shape of what you get.

Where your engineering capacity is going, what to stop, every assumption on the table, and an honest account of the limits.

Request an assessment →Run an instant Delivery Snapshot