Tools and providers
Selecting conversion suppliers by one store job, one dossier and independent risk gates
Select ecommerce conversion suppliers by one defined store job, current evidence, controlled acceptance, independent risk gates and a tested exit.
Pick one store decision first, then look at suppliers. This guide names no specific products or suppliers and ranks nobody. It sets out a process that runs from a named job to signed acceptance criteria.
What to take away
- Name one store decision, its owner and its date before you open a supplier page.
- No specific suppliers or tools are named here, so treat this as a process guide rather than a comparison.
- Build a candidate dossier from official records before you accept a demonstration.
- Put every candidate on one meter, converting sessions and identified profiles to a single figure.
- Fix acceptance criteria and whole-life cost before the trial starts.
Start from one store decision
Write one sentence with one owner and one date. A merchant diagnosing unexplained checkout exits needs different evidence from one monitoring failed payments. Do not start with a category name such as analytics or personalisation.
Attach the journey. Name the legal seller, the product, the adult customer population, the England location rule and the channel. Identify the failing transition and the evidence you already hold.
The baseline may be an existing report, a manual reconciliation, moderated research or no intervention. Record its labour, delay, error risk and coverage. Do not invent a performance figure for it.
- Write the decision, owner and date.
- Name the journey and the failing transition.
- Describe the current baseline and its coverage.
- List the inputs the supplier must receive and return.
Assess each tool job separately. Journey analytics records events. Experimentation allocates units and keeps exposure evidence. A consent interface does not settle the lawful basis for later use.
Build the candidate dossier
Confirm the contracting entity and any trading name through the government's company information service. That record confirms registered facts. It is not an endorsement or a solvency opinion, and it says nothing about the product.
Then check the wider identity, role and exit position through conversion vendor due diligence before you book a call.
Keep a fixed checklist:
- Entity name, trading name and company number.
- Service, edition, UK availability and published change date.
- Inputs, outputs, event definitions and limits.
- Data locations, subprocessors, retention and deletion.
- Security owner, incident route, export and termination.
Example dossier entry, filled in
Field / Entry
- Decision
- Diagnose guest checkout exits
- Owner
- Head of ecommerce
- Entity
- Registered name and number, checked on the company register
- Edition
- Named plan, UK availability, change date
- Data
- Locations, subprocessors, retention, deletion
- Exit
- Export format and deletion on termination
Mark an absent point unknown once. Do not infer a feature from a plan family name.
Draw the data path before capability
Trace each identifier and record. Give the source, purpose and controller. Then give the processor, recipients, retention and deletion. The ICO's storage and access technology guidance explains current PECR treatment of cookies and similar methods. A server-side design or an analytics label does not settle whether consent or an exception applies.
Define missingness on purpose. A refused optional storage, a blocked script, a late refund and an unmatched payment are separate states. The supplier must not quietly drop them from a denominator.
Ask security questions that fit the connection
Map admin rights, production write access, scripts, interfaces and logs. The NCSC's supplier assurance questions cover governance, incidents, networks, data and independent assurance. Tailor them to the service instead of treating them as a pass sheet. A trust badge cannot show the safety of an unexamined integration.
Convert every quote to one meter
| Quoted unit | What it counts | Step to one meter |
|---|---|---|
| Per session | One browsing session | Multiply by your session-to-profile ratio |
| Per identified profile | One matched customer | Divide the fee by matched sessions |
| Per order | One completed order | Divide the fee by completed orders |
| Per experiment | One test run | Divide by planned tests and months |
Illustrative arithmetic only: if your log shows 40,000 sessions and a 25 per cent match rate, the profile-equivalent count is 10,000. Divide the per-profile fee by that count. Compare structures on identical meters in conversion pricing models, which measures seven against whole-life cost.
Whole-life cost covers buyer labour, configuration and assurance. Add integration, support, incident response and exit. A free trial is not a free operation.
Set acceptance criteria before the trial
Predeclare acceptance for the exact job. Cover eligible cases, expected state transitions and missingness treatment. Add definition version, security evidence and accessibility cases.
Example acceptance criterion: eligible cases are guest checkouts in the England store from 1 to 31 August. Refused storage is reported as its own state, not dropped. A named accessibility specialist signs off keyboard and screen-reader cases before release.
Put support boundaries, change notice and audit to the supplier as contract questions. Add export, deletion, termination and liability after legal review.
Common questions
Do I need named suppliers to choose one?
No. The dossier fields and the acceptance criteria do the work. A shortlist follows once candidates return complete entries.
Which rules shape the trial?
PECR governs storage, the CMA rules cover misleading conduct and price transparency, and the Equality Act applies to service provision in England and Wales. Security assurance follows the NCSC questions.
What if a supplier will not answer?
Treat the gap as evidence. Score the candidate on what it published and supplied, not on what a sales call promised.
Do I need a lawyer?
Yes for the data-processing terms and the contract. The buyer can prepare the dossier and the acceptance criteria first.
In this guide
- Ecommerce conversion software selection: the acceptance testRun a bounded acceptance test of ecommerce conversion software for an England store: predeclared cases, data checks, an exit rehearsal and a clear verdict.
- No conversion tool makes the shortlist yet, and the supplier evidence that would change thatSee why the current England conversion-tool shortlist is empty and what exact supplier evidence and controlled buyer tests would change that hold.
- Comparing conversion procurement routes on one evidence grid with no universal winnerCompare conversion procurement routes on one England store job, identical evidence fields, blank commercial assumptions and no universal winner.
- Vetting a conversion supplier's identity, data roles and exit before you buyInvestigate a conversion supplier's identity, exact service, data roles, security, customer effects, commercial terms and workable exit before purchase.
- Ecommerce conversion tool implementation: tags, consent and six release gatesHow to implement an ecommerce conversion tool: tool categories, tag and consent set-up, a segregated test, six independent release gates and a rollback plan before connecting a store tool.
