The Delivery Truth Engine

The Delivery Truth Engine

Lighthouse reconciles your delivery signals into one defensible truth — and shows you the working.

Most delivery tools give you one confidence number and ask you to trust it. Lighthouse cross-checks nine independent lenses, catches their contradictions, and won’t sign off on a steering pack that doesn’t add up. You see what we saw, why we said it, and what to fix.

Reconciles confidence

Every initiative gets nine confidence lenses, calculated independently from your data. Lighthouse reconciles them into one defensible number — and shows you exactly which lens disagreed and why.

Delivery confidence

How well the plan is hitting its dates, based on milestones and progress.

Capacity confidence

Whether the squads doing the work actually have the hours to do it.

Data confidence

How fresh and complete the underlying numbers are — stale data is loud.

Financial confidence

Is the spend tracking to plan, or is the burn rate quietly running hot?

Governance confidence

Stage gates passed, approvals recorded, dependencies acknowledged.

Scenario confidence

Best alternate plan from your scenario library — what if we rebalanced?

Pre-mortem confidence

Lighty imagines this initiative failing and tells you the top 3 reasons.

Monte-Carlo confidence

Simulates thousands of versions of the plan to find the realistic finish.

Reported confidence

What the PM said in their last update — usually the most generous lens.

Real example. In our sample-data demo, the “Apollo” initiative reports 78% delivery confidence. Reconciliation drops it to 52% — capacity confidence flagged the squad at 145% utilisation and pre-mortem confidence flagged a vendor risk the PM hadn’t recorded. Both are visible in the breakdown.

Catches contradictions

The Coherence Gate runs nine cross-checks against every steering pack. Plain-English rules — not regex on a database table.

  • If you say it’s healthy but capacity is at 145%, we’ll catch it.

  • If a 28-week initiative has 4 weeks of plan, we’ll catch it.

  • If a milestone is overdue but confidence is still green, we’ll catch it.

  • If you’re reporting on people who aren’t actually allocated, we’ll catch it.

  • If a dependency on another initiative is silently broken, we’ll catch it.

  • If a steering pack is using a baseline older than the last reset, we’ll catch it.

  • If scenario confidence is wildly better than the live plan, we’ll catch it.

  • If two confidence lenses disagree by more than 30 points, we’ll catch it.

  • If a stage-gate decision wasn’t actually recorded, we’ll catch it.

Before

Apollo: 78% confidence, on track

Capacity is at 145%, but nobody flagged it. The PM’s update wins by default.

After

Apollo: 52% reconciled, capacity contradicted reported

Lighthouse caught it. The pack lists the contradiction next to the number.

Decision audit trail

Every recommendation Lighty made, every call you took, and every outcome that followed lands in an append-only register per initiative. The board can audit the path — not just the outcome.

When
Lighty recommended
You decided
What happened
2026-04-08
Pause Apollo, free 2 engineers for Checkout
Hold scope on Apollo, defer Checkout one sprint
Checkout slipped one sprint; Apollo capacity stayed at 145%
2026-04-22
Promote scenario S-7 (rebalanced) to live
Promoted with caveat: Range tribe vacancy still open
pending — outcome captured at the next gate

Why this matters. When the post-mortem asks “why did we choose that path”, you have a row, not a guess. When auditors ask about a decision two quarters back, you have a row, not a screenshot of a Slack thread. Decisions are append-only — the register can be reviewed, but not rewritten.

Won’t sign off on bad packs

Lighthouse refuses to mark a steering pack as “board-safe” if the numbers don’t add up. Three states, no overrides.

Board-safe

Numbers add up across every lens. Take this pack to your board.

With caveats

Mostly aligned, one or two coherence rules flagged. We’ll print the caveats so you can pre-empt the question.

Not board-safe

The numbers contradict each other materially. Fix the contradictions, then re-run.

Why this matters. The cost of a bad board meeting isn’t the meeting — it’s the next three weeks of recovery. The Board-Safe Gate stops bad packs going out, full stop.

See the Delivery Truth Engine in your own data.

Reconciles delivery signals into board-safe decisions. Live in every steering pack.