Skip to content

Briefing · Decision systems

· 3 MIN

Delivered is not verified: count outcomes only when measured

Completion is a date. Verification is a measurement against a threshold that was set before the decision.

Delivered is an event.

A delivery recorded is a milestone closed. A benefit recorded is a claim about the future made in the present tense. The two are often treated as the same thing, and that difference is where value quietly disappears.

Verified outcomes are counted only when measured. Forecast is not realised, and delivered is not verified.

The distinction matters most where delivery is a proxy for something else: a platform meant to reduce cost, a control meant to reduce risk, a capability meant to change a decision. The milestone closes; the effect may never arrive.

This is not a semantic preference. What gets counted determines what gets funded next, and an organisation that counts completion will keep funding completion.

What delivered leaves out.

A system can be installed and unused. A capability can exist and never be invoked. A saving can be booked before it is banked. None of that is necessarily dishonest, and all of it is common: completion measures activity, not effect.

The corrective is not more reporting. It is deciding, at the point of decision, which measure would show whether this worked and what value of that measure would count.

It matters commercially too. A contract can be complete and a benefit unrealised, and the people closest to the work often know that long before the reporting does.

Choose the measure before the decision.

A threshold chosen afterwards is a description, not a test. Decision and governance work best when the measure, the baseline and the review date are recorded with the decision itself, so that verification has something to be verified against.

That is a discipline before it is a technology. The engine holds it: the position, the trade-off accepted, the person who accepted it and the date the claim is tested.

Written that way, the threshold is also a test of the decision. If no measure would count, the decision was a preference rather than an investment.

  1. 01Decision
  2. 02Threshold set
  3. 03Delivered
  4. 04Measured
  5. ↻ measured outcomes return as evidence
ONE DECISIONMEASURED

Verified outcomes return as evidence.

Verification is not an audit endpoint. A measured outcome is the strongest evidence the next decision can have, because it is the only evidence that survived contact with the operating record.

Outcomes return as evidence and change the position the next time it is formed. That loop is what makes an intelligence system improve rather than merely accumulate.

It also makes the system's own claims falsifiable, which is the point. An intelligence system that cannot be shown to have got something wrong has not earned the right to be trusted.

It changes how the system describes itself as well: not as a tool that produced a recommendation, but as one that can show whether the recommendation held.

The cost of measuring.

Not everything can be measured quickly. Some outcomes take a year, and some are contested by the people who own the number. Verification is slower and less flattering than completion, and it is the only basis on which a second decision can be made honestly.

Where measurement is genuinely impossible, the honest move is to say so, name the proxy being used and state what it cannot show.

The compromise most organisations reach is to measure what is available rather than what matters, then treat the proxy as the outcome. That is how a verified process ends up verifying the wrong thing.

The alternative is uncomfortable and shorter: name the effect you expect, name the measure that would show it, and accept that the answer may arrive after the people who made the decision have moved on.

In short

Delivered is a fact about activity. Verified is a fact about effect. A system that counts the second will look slower this quarter and be more reliable across the decade.

All insights

Research

The research behind the briefing.

turns research, industry expertise and verified outcomes into methods the engine can run.

Explore Research