How to Map a Commitment Score to a Pilot Go/No-Go Decision

Take the simplest possible tally. Each signed commitment is worth one point. A commitment counts only if it meets four conditions: it is signed by someone who can authorize spending, it names a start window, it names a success measure, and it is not conditional on the pilot being free of charge. Anything else goes in a separate "interest" column that never enters the score.

Your published rule might read:

• Score of 8 or more on the review date: go. Book the build and name the pilot owner.
• Score of 5 to 7: hold. The pilot does not start; the team gets a fixed extension (for example, four weeks) to convert interest into signatures, then the rule is applied again with no further extensions.
• Score of 4 or below: no-go. The pilot is cancelled, the evidence is filed, and the team moves to the next hypothesis.

The numbers themselves are a business judgment, and you should be able to defend them. A useful sanity check is to ask what the score would have been for the last three things you built and whether that order matches how they actually turned out. If your scoring would have rated a failure above a success, the definitions are wrong, not the threshold.

Here is the same logic as a comparison table you can adapt. The middle band is what most teams skip, and it is the band that prevents the two bad outcomes: launching on hope and killing on a technicality.

Keep the tally mechanical. If two people can look at the same evidence and get different totals, the definitions are ambiguous and the decision will be contested. The discipline of tagging raw notes with themes and counting patterns, such as noticing that seven of ten interviews show the same behavior, is a known technique for turning qualitative work into something countable per this guide to actioning discovery interviews.

A total score hides concentration. Eight points from eight different buyers is a different situation from eight points from one enthusiastic buyer who signed eight line items. This is where a second threshold, a stop trigger, earns its place: it is a condition that forces a no-go regardless of how high the total is.

Good stop triggers are rare, objective, and cheap to check. Examples:

• Fewer than three distinct signing organizations, whatever the point total.
• Zero commitments from the segment named in the pilot plan, meaning all signatures came from adjacent, easier buyers.
• The review date passes with the tally incomplete, and the rule says an incomplete tally counts as the score achieved so far.
• A dependency the pilot cannot survive (a data access approval, a security review, a partner contract) is still unsigned on the review date.

The third item is the one teams most often forget. An unscored commitment is not a neutral object. Allowing "we are still waiting on four signatures" to postpone the decision indefinitely is how a hold becomes an accidental go. Write the rule so that the deadline binds: on the review date, whatever is signed is the score.

This mirrors the "committed to stop" logic in aviation, where operators are advised to define a point after touchdown beyond which a go-around should not be attempted, so that the decision is not made under pressure in the moment per this report on committed-to-stop points. The transferable idea is that the point of no return should be named in advance and briefed, not discovered.

Stage gates work the same way in product organizations: each gate evaluates the project against predefined criteria and the outcome is approve, rework, or kill per this description of the Stage-Gate process. The value is not the gate itself but the mandatory demonstration that the criteria are still met, which stops projects drifting forward on momentum.

Real tallies are messy. Someone signs a letter of intent with no budget. Two reviewers disagree about whether a verbal pledge counts. A procurement contact says "we would definitely buy" and then stops replying.

Three treatment rules cover most of it, and they should be written on the same page as the bands:

• Half credit for partial commitments. A document that names a start window but not a success measure scores 0.5. Half credit exists so the tally reflects momentum without letting weak evidence equal strong evidence. It also keeps totals from being rounded up informally.
• Excluded for unverifiable commitments. Any commitment you cannot show to a third party (a verbal yes, a hallway promise, a "sounds great" email) is excluded from the score and noted in the interest column. If it cannot be produced in an audit, it cannot move a decision.
• Pending on the review date counts as zero. Pending is not a state that changes a total. It is a state that describes evidence you do not have.

Disputes need a tiebreak that is not seniority. A practical rule: disputes about whether a commitment meets a definition are resolved by the person who signed the rule sheet, not by the pilot owner, and resolved before the tally is finalized. If the disagreement is about the definition itself rather than the evidence, the definition is amended for the next pilot only, never mid-tally.

Self-reported intent deserves special suspicion. A click is not a commitment, and product teams now routinely note that click-through measures curiosity rather than commitment, since a click does not prove a user would change a workflow or pay per this analysis of fake door testing. The same caution applies to surveys. Asking people to reconstruct a recent event, what triggered action and what the current workaround costs, produces far more decision-grade evidence than asking whether they would probably use something per this set of customer discovery questions.

If your commitment score is built from interviews rather than signed documents, treat each interview as evidence for a definition, not as a point in itself. Narrow the segment before you recruit: asking people who share the role, environment or use case you are testing makes a pattern interpretable, whereas a mixed panel produces averages that describe nobody per this discussion of customer discovery interview practice. And keep the base rate in view: a large share of new product ideas fail, with no market need among the leading causes, which is the reason the score exists at all per this summary of validation research.

An auditable rule has four parts, and all four belong in a single document that gets a version identifier and a date.

Two habits make the difference between a file that survives scrutiny and one that does not. First, timestamp the rule version independently of the tally, for example by sending it to a shared channel or filing it with an outside reviewer, so nobody can argue it was backdated. Second, keep the evidence register in a form that can be handed over, because the ability to export records and the terms on which you can do so is exactly the kind of thing worth checking before you depend on a tool to hold your evidence per this vendor evaluation guide.

The principle behind both habits is that a public commitment without verifiable disclosure of how it was met is worth very little. That is the finding of a study that scored sixteen AI companies against their voluntary White House commitments and found wide variation, with an average score of 54 percent and one commitment area averaging 17 percent; the authors' central recommendation was that commitments should come with proactive, verifiable disclosure per this AAAI paper on voluntary commitments. Your pilot decision deserves the same standard: not "we decided this was enough" but "here is the evidence, here is the rule, here is who applied it."

A minimal decision record in a flat file is enough for most teams:

rule_version: commitment-threshold-v3
rule_signed_on: 2026-02-10
signed_by: [pilot_owner, finance_reviewer, independent_reviewer]
review_date: 2026-04-15
tally:
  full_commitments: 6
  half_commitments: 2      # 2 * 0.5
  excluded_unverifiable: 3
  pending_at_review_date: 1
score: 7.0
band: hold
extension_days: 28
next_review_date: 2026-05-13
applied_by: pilot_owner
notes: "All half credit from letters of intent without success measures."

The meeting where the rule is applied should feel boring. The work that matters happened when the rule was written. Three practices keep it honest.

First, count before you discuss. Open the tally sheet, compute the score in the room, and read the band and the resulting action aloud before anyone offers an opinion. If the rule says hold, the meeting's purpose shifts immediately from "should we launch" to "what is the conversion plan for the extension," which is a much narrower and more productive conversation.

Second, separate the evidence question from the value question. Whether a commitment meets the definition is a factual dispute resolved against the written rule. Whether the pilot is worth doing at a given score was decided when the threshold was set. Mixing the two is how thresholds quietly move.

Third, log the override path even when you do not use it. Knowing that an override requires sign-off above the sponsor, with a written rationale, changes how the room behaves. The alternative is worse: an override by exhaustion, where the decision gets remade every week until the answer changes.

Continuous reassessment applies here too. The go/no-go decision is not a single event, and conditions can change after the start, requiring a fresh call per this glossary definition of go/no-go decisions. Pick at least one reassessment point inside the pilot (for example, at 30 days) and attach a second, simpler stop rule to it, such as "if fewer than half of the committed buyers have started using the pilot, stop and reclassify."

One more instrument worth borrowing: stakeholder engagement measurement works best when targets are set as specific countable items, such as numbers of attendees, response rates and resolution times, rather than as a general sense of goodwill per this guide to measuring stakeholder engagement. Your commitment tally is that same idea applied to a decision.

If the tally comes back in the hold band and the extension work needs a working artifact to show committed buyers something concrete, that is the moment a scoped build helps, and a specialist can pick up the description and build it with you at Quanta.