An unsolicited concept study — not a live service. Every person shown here is invented.
Receipts

Every claim comes with the receipt attached.

Start Scope

What I would build, in what order

Three weeks, and each one ends with something you can measure. The order is deliberate: the cheapest test comes first, so you can stop after week one and still be ahead.

  1. 01

    Week 1 — the search page

    Server-rendered search and results, visible labels, a real 404, the identity strip free in every tier, and the price line on the button.

    It is the page your money walks through, and it is the one measurable in a week. If this does not move your numbers, stop there — you will have paid for one week and learned something.

  2. 02

    Week 2 — the report

    Provenance on every line, the outcome before the charge, the headline counting documents, and the sourced/fun split.

    This is the part that changes what the product IS. It is also the part your support inbox notices first.

  3. 03

    Week 3 — money and the second party

    Renewal and cancellation at the point of sale, one refund rule from one source, automatic refunds on empty and wrong, and the view for the person in the file.

    Chargebacks and removal requests are a cost line. This week is the one that pays for itself without needing a single extra sale.

The five questions your lawyer will ask

Better here than in week three. Where I do not know, it says so.

Does removing the score lose us conversions?
Unknown, and that is why it is a switch rather than a rewrite. Both states render from one config file; run it as an A/B for two weeks and keep whichever wins. My expectation is neutral on conversion and better on refunds, but expectation is not evidence.
Does stating cancellation up front kill subscriptions?
Some, yes. It also removes the chargeback, the angry review and the support thread — and in this category those are not rare events. It is the one change where I would accept a small measured loss.
Are we more exposed by showing the searched person their file?
Less, not more. The exposure in this category comes from wrong data that nobody can correct. A correction channel converts a liability into a data quality signal — and you are the one who gets to see it first.
Is any of this required by law?
Not in the US for this use, no — that is the point of not being a consumer reporting agency. It is required the moment you take European traffic, and it is cheaper to build it now than to retrofit it under a deadline.
Who owns the code?
You do, on delivery, with the tests and the guards. Everything you see in this demo — the rules, the probes, the acceptance gate — comes with it.

And the honest part

This demo was built from the outside, in an afternoon, with invented data. I have not seen your data layer, your billing edge cases or your support queue, and every estimate above will move once I do. What is not an estimate is the code you are looking at: it runs, it is measured, and the guards that keep it honest ship with it.