PMF Confidence Scorecard

Confidence in What, for Which Decision?

A client-reporting product has found a promising pocket. Five small agencies sent a client-ready report, four returned the next week, three accepted the standard paid pilot, and one referred another agency. The founder can tell a credible PMF story from those facts.

The same accounts tell a less comfortable story. Three needed help mapping their data before the first report. The evidence covers one weekly return for most accounts. Recruiting came mainly through the founder. Larger agencies will pay more, but only after custom roles and import work.

The question is not whether the product has PMF as a permanent condition. The question is whether the current evidence is strong enough for a particular commitment: focus the next cohort on small agencies, build a self-serve import path, increase acquisition, harden the system, or keep searching.

Use this scorecard when several evidence tools point in the same general direction and a consequential decision is due. It does not turn unlike signals into a numerical grade. It limits the claim, follows the evidence chain, exposes contradiction, and sets the next boundary at which confidence must be earned again.

Write the Claim Before Collecting Support

“We have PMF” is too broad to review. Confidence belongs to a claim with a segment, recurring job, delivery model, and time boundary. It also expires when the product, price, promise, acquisition source, or customer changes enough that the old evidence no longer describes the offer.

Begin here:

PMF CONFIDENCE BOUNDARY
Decision due:
Claim being judged:
Target segment, buyer, and operator:
Recurring job and observable value event:
Natural cycle of that job:
Product, offer, price, and acquisition paths represented:
Account, workspace, site, or user as the evidence unit:
Observation cutoff and eligible cohorts:
Founder-assisted work recorded separately: yes / no
Changes during the evidence window:
Commitment this review may authorize:
Commitment it may not authorize:

A useful claim for the reporting product is narrow:

Small agencies with weekly client meetings repeatedly send client-ready reports, will pay the standard pilot price, and can be served within a bounded setup path. The evidence may justify recruiting one more comparable cohort; it does not yet justify broad paid acquisition or a move upmarket.

That sentence can be contradicted. It also prevents evidence from freelancers, multi-office agencies, one-off custom imports, and immature cohorts from being pooled into a more flattering answer.

Follow the Chain Without Adding It Up

Read PMF evidence in sequence. Can the founder reach and qualify a recognizable customer? Does that customer reach first value? Does value recur when the job returns? Does payment follow the product’s value rather than bespoke persuasion? Does customer behavior show pull? Can one founder deliver the promise at a sustainable load?

Each link answers a different question. Three payments do not repay an activation debt. An enthusiastic survey response does not repair absent return behavior. Strong retention produced through weekly founder rescue can prove urgent pain while leaving the productization claim weak.

Build the review from account histories, product events, billing records, support and sales traces, customer language, and a log of founder work. For each link, write:

EVIDENCE LINK
Question this link answers:
Observed accounts and exact denominator:
Behavior and natural-cycle window:
Founder assistance, discounts, or exceptions:
Customer words attached to the behavior:
Evidence source and definition changes:
Strongest supporting observation:
Strongest counterexample or missing observation:
What this link supports:
What it cannot support yet:

Keep unknown links unknown. An account whose next cycle has not arrived is not retained and is not churned. A buyer who accepted a pilot has supplied payment evidence; until the operator reaches value, that account has supplied no activation or retention evidence. Missing data should narrow the commitment, not be converted into a neutral score.

Do not total the links. The first unsupported link controls how far the decision may travel. Later evidence can still reveal a service opportunity, pricing power, workflow urgency, or a promising exception, but it cannot lend its strength backward to a claim the earlier evidence does not support.

Give the Confidence Read a Decision Boundary

Use ordinary language before choosing one of four reads. The label is a summary, not a calculation.

Insufficient means the claim cannot yet be judged. The segment or value event may be vague, the eligible sample may be unknown, or the evidence may come entirely from intention. The next action is to repair the boundary or observe the missing behavior.

Directional means a coherent pattern is visible but still rests on a small sample, one customer cycle, founder selection, material assistance, or another untested condition. It can justify a bounded experiment or another comparable cohort. It cannot justify an irreversible scale commitment.

Decision-worthy means the chain has repeated under comparable conditions, the accounts behind the measures are inspectable, payment and customer language agree with behavior, and counterevidence has not broken the claim. It can justify concentration, productization of a repeated bottleneck, or a larger controlled test within the stated segment and operating boundary.

Durable within the boundary means the chain has survived multiple natural cycles or cohorts, meaningful product and acquisition variation, loss and counterexamples, and delivery without hidden founder rescue. It can support a more durable investment. It is not permission to ignore reliability, trust, economics, channel capacity, or the new failure modes created by scale.

Confidence can differ by link. “Decision-worthy retention, directional payment, insufficient self-serve delivery” is more useful than “medium PMF.” The overall commitment must respect the weakest link relevant to that decision. A manual import weakness may block self-serve acquisition while still allowing another founder-onboarded cohort. Weak renewal evidence may block annual planning while allowing an activation repair.

Read the Client-Reporting Evidence

At the current cutoff, nine qualified small agencies attempted the standard offer. Seven connected a data source. Five sent a client-ready report; three of those five needed founder help with mapping. All five have reached another weekly reporting cycle, and four sent another report. Three accepted the standard pilot after using the product. One retained agency referred a peer. Founder load averages 35 minutes per active account and is concentrated in initial setup, but the next cohort has not tested the proposed import repair.

The evidence is decision-worthy for focusing another bounded cohort on small agencies. Activation, natural return, payment, and referral point toward the same recurring job. The result is only directional for self-serve delivery: most successful activations still include founder mapping, the sample is small, and recruiting has not traveled far beyond the founder’s reach.

That distinction changes the decision. The founder should repair and test the common import boundary with the next four qualified small agencies. The expected result is a first client-ready report without data cleanup, followed by another report at the next weekly cycle. Recruiting more broadly would increase the very work the product has not learned to carry. Building custom roles for larger agencies would test a different product and delivery model.

Complete the Review

Use one review for one claim. Attach or link the underlying dashboard, cohort, revenue, qualitative, churn, and segment records rather than copying their contents into a polished summary.

PMF CONFIDENCE REVIEW — [CLAIM] — [DATE]

DECISION BOUNDARY
Decision due:
Authorized commitment if supported:
Evidence expiry or next review date:

CHAIN READ
Reach and qualification:
Activation and time to first value:
Return at the natural cycle:
Payment, renewal, or expansion after value:
Behavior-backed customer pull:
Founder load and repeatable delivery:

EVIDENCE QUALITY
Eligible accounts and cohorts:
Sources reconciled to account histories:
Product, price, promise, or channel changes:
Assistance, discounts, and custom work:
Missing or immature observations:

CONTRADICTION
Strongest supporting account or cohort:
Strongest counterexample:
Plausible alternative explanation:
What evidence would show that explanation is right:

CONFIDENCE READ
Insufficient / directional / decision-worthy / durable within the boundary:
Confidence by weak or contested link:
What the evidence supports:
What it does not support:

DECISION
Continue / concentrate / repair / narrow / hold / stop:
Action and accounts exposed:
Expected behavior by the next natural cycle:
Evidence that would reverse the decision:
Work, hiring, acquisition, or hardening explicitly deferred:

If supporting and contradictory evidence describe different segments, jobs, offers, or delivery modes, split the claim rather than averaging them. If the same accounts appear in several metrics, do not treat each appearance as independent confirmation. If definitions changed during the window, preserve both definitions and begin a new comparable period.

The review is complete when a skeptical future reader can trace the confidence read back to accounts and see why the chosen commitment is no larger than the evidence. Confidence should make the next bet more precise. Whether the product is ready for scale remains a separate decision.