How the tool makes money and what that does to the score

A rating tool has to make money somehow, and the "somehow" is not neutral background information. It is the closest thing to a design document a tool has, because whatever behaviour keeps the lights on is the behaviour the whole product ends up quietly optimised for. Three models cover most of the category, and each rewards a different action from the user.

Subscription: wants you to stay

A subscription tool's revenue depends on retention, not on any single visit. The behaviour that serves that model is a number the subscriber does not feel cheated by - not necessarily flattering, but stable and worth checking back on, because a subscriber who feels burned by one result cancels before the next billing cycle.

This tends to show up as consistency rather than generosity: a subscription tool has less reason to inflate any single score and more reason to make its scale feel dependable over months of use, since the product it is selling is the ongoing relationship, not the one number. It also has a reason to keep adding things to check back for - history, trends, new axes - because a stagnant feature set is the fastest way to make a recurring charge feel unjustified.

Credits: wants you rescoring

A credit-based tool sells attempts, and its revenue scales directly with how many times a user submits. The behaviour that serves that model is a result that makes another attempt feel worthwhile - a score close enough to a threshold, or a tips section implying a retake with a small change might do better.

This is the model with the clearest incentive toward the reveal animation and the "tips to improve" box doing real product work rather than being purely decorative. A credit tool that resolves your uncertainty in one attempt has sold you exactly one credit's worth of business; a credit tool that leaves you wondering has sold you a second attempt, and the design incentive points accordingly.

Ads: wants you sharing

An ad-supported or share-driven tool makes money from traffic, and traffic comes from results people want to post. The behaviour that serves this model is a shareable outcome - a badge, a striking share card, a number flattering enough to post without embarrassment.

This is the model with the strongest pull toward generosity at the top of the scale, because a tool whose median user gets an unshareable result has a much smaller multiplier on each visit than one whose median user gets something worth a screenshot. What a share card keeps and what it drops is itself shaped by this incentive - it is built to travel, not to inform.

Reading direction, not magnitude

None of this is measurable from outside with any precision - there is no way to know, from a single result, how much a specific business model moved a specific number. What is available is direction: a tool funded by ads has more reason to skew generous than one funded by subscriptions, and a credit tool has more reason to leave you wanting another attempt than either. Treat these as leanings a business model creates, not as proof that any particular tool has acted on them.

This is not the same as the free-versus-paid question

Cost model shapes the whole product's incentive; it is a different question from whether a specific paid tier scores the same submission more kindly than the free one, which is a narrower, directly testable claim about one tool's two tiers. It is also separate from what a subscription buys in features - more axes, saved history, a longer write-up - which is a real value independent of any incentive on the number itself.

Where this fits in evaluating a tool

Cost model is part of the disclosure check in the broader evaluation method: a tool that states plainly how it makes money has told you which of these three pulls to watch for, and that statement belongs alongside the rest of what a transparent tool volunteers. A tool that will not say how it is funded has not hidden anything about the model doing the scoring - the vision model underneath is largely the same across the category regardless of business model - but it has hidden the one piece of context that explains which direction its own number is more likely to lean.

None of these three pulls apply to a human reviewer in the same way - a person being paid to look at a submission has a payment relationship, not a scale-wide incentive baked into a product, and the two are worth keeping separate when weighing what either kind of result is optimised toward.

A tool that names its model, states which of these three it runs on, and reports components rather than a single blended figure is a tool that has left less room for the incentive to hide in. Rate Cock is a subscription-funded example that reports six axes rather than a total specifically for that reason - a breakdown is harder to quietly skew than one number, because each axis has to move on its own and the movement is visible. A physical measurement sidesteps this incentive question almost entirely, since a recorded length does not have a scale a business model can lean on the way a modelled score does.