Run one frozen fictional pack across four discovery routes

A route comparison is only fair if every route sees the same records. Build one pack of fictional identities, freeze it, and run it on the manual register, a marketplace, a paid database and a managed service. Then write down what each route actually did, not what its sales page says.

What to take away

What goes into the pack

The pack uses fictional identities only, so no real creator is exposed and no contact can be sent by accident. It holds two near-duplicate names, an ambiguous spelling, a creator who changes representative, a refused contact request and a tagged tracking link.

Each item has a job. The near-duplicates and the ambiguous spelling test how a route resolves identity. The representative change tests whether it notices a stale record. The refused contact request tests suppression. The tagged link shows what the route does with a record that carries campaign tags.

Keep the pack frozen. Change a name, a link or a consent state and the results stop being comparable.

Which routes can run the pack today

Only the buyer-owned manual register runs today with no purchase decision. It needs no subscription, account or contract, so it sets the baseline that other routes are measured against. Every other route waits on something the buyer does not yet hold.

Names identify route categories only, and vendors often combine functions, so test the contracted edition rather than the brand. Hold the marketplace rows until the relevant ads account exists. Hold managed services until a Companies House check confirms the trading entity.

GOV.UK lists company information, such as registered address and date of incorporation, among the details available for free. That settles identity only. It says nothing about how the route handles creator data.

Record six outputs on every route

Run the same pack on each route and record the following, in this order, before moving to the next route.

Setting the expected result first matters, because it stops a route's output from rewriting the standard it is judged against. A pass shows only that the route behaved as observed under these conditions. A contract promise is not observed behaviour.

Record the source and status of every field. The ICO's accuracy checklist asks whether you have processes to check the accuracy of the data you collect and whether you record its source.

Score the same six gates on every route

Every route is scored on the same gates, and a pass on one gate cannot offset a failure on another. A route with an unknown legal entity, an incomplete dossier or an unverified exit goes on hold, however good its search results look.

Disclosure sits with the brand as well as the creator. The ASA advertising codes set out how paid posts must be identified, so the pack's disclosure case should be judged against them and not against a platform label.

When an earlier comparison stops being usable

A changed source, a new edition, a different subprocessor, a revised identity method or an altered export format makes an earlier comparison unusable. Record who approved the change and what forces a re-run, so the next reviewer knows which results still stand.

Re-run the same frozen pack, not a new one. A fresh pack would reset the evidence and hide whether the route got better or worse. Keep the old results beside the new ones with their dates.

Cost and route choice sit outside this protocol. The head-to-head test of a manual register against one database scores both in buyer-defined units. For the second pass, the budget template holds a blank total that carries exit costs.

Common questions

Which route can be tested first?

The buyer-owned manual register. It runs today with no subscription, account or contract, so it sets the baseline for every other route.

What does a passing result prove?

Only that the route behaved as observed under these conditions. A contract promise is not observed behaviour, and a pass on one gate does not offset a failure on another.

What evidence should I keep before committing?

Keep the dossier exports, the evidence state per cell, the dates, the tester name and the live exit export.

What makes an earlier comparison unusable?

A changed source, a new edition, a different subprocessor, a revised identity method or an altered export format. Record who approved the change and what forces a re-run.