
Tools and providers
Part of Selecting conversion suppliers by one store job, one dossier and independent risk gates
Comparing conversion procurement routes on one evidence grid with no universal winner
Compare conversion procurement routes on one England store job, identical evidence fields, blank commercial assumptions and no universal winner.
Compare conversion procurement routes only after fixing one job. The case here is an England merchant that needs to reconcile checkout, payment and refund states before deciding whether a page problem exists.
The routes named below come with public list prices or labelled typical ranges. The buyer's dossier keeps candidate names and prices blank until exact-edition evidence or a controlled test arrives; the grid still allows a like-for-like comparison.
Research date: 6 September 2026. Method: compare the current manual route, a buyer-configured single-purpose service, a managed supplier service and buyer-built software on identical evidence fields. Exclusions: the buyer's dossier excludes brands, prices, invented capabilities, staffing assumptions and predicted results. Public brand and price facts below come from published sources. Winner: none; the buyer must set weights and thresholds.
What to take away
- No route wins on the evidence grid; the buyer must set weights and thresholds.
- Every route needs the same store states, privacy, security, consumer and payment fields recorded. Shopify exports, Microsoft Clarity and Hotjar face that test. So do Hallam, Croud, Segment and BigQuery.
- Missing evidence should be marked unknown, not scored lower as an invented penalty.
- An export that cannot recreate the decision record counts as incomplete evidence.
- A privacy, consumer, accessibility or security failure cannot be offset by convenience.
Compare the named routes: Shopify, Clarity, Hotjar, Hallam and Segment
The current manual route uses authorised exports or existing records and a documented reconciliation. In England that often means Shopify admin CSV exports for orders and refunds plus Google Analytics 4 exploratory reports.
Conversion procurement routes compared
Manual
- Examples
- Shopify, GA4
- Cost
- £25/mo + free
- Control
- Direct buyer
- Coverage
- Buyer-measured
- Delay
- Buyer-measured
- Labour
- Buyer-measured
Single-purpose
- Examples
- Clarity, Hotjar
- Cost
- Free to £32/mo
- Control
- Defined events
- Coverage
- Specified report
- Delay
- Alert timing
- Labour
- Low
Managed
- Examples
- Hallam, Croud
- Cost
- £2k–£10k/mo
- Control
- Agency operates
- Coverage
- Retainer scope
- Delay
- Agency SLA
- Labour
- Agency staff
Shopify's Basic plan is listed at about £25 a month in the UK and includes those exports. GA4 standard properties are free, with a limit of 10 million events a month per property.
The route preserves direct buyer control, but coverage, delay, labour and error remain buyer-measured fields.
A single-purpose service would ingest defined events and return a specified report or alert. Microsoft Clarity is free and records session replays and heatmaps with no stated traffic cap. Hotjar has a free tier; paid plans typically start around £32 a month for 100 daily sessions.
Contentsquare sits at the enterprise end with quote-based pricing. The dossier must still identify the legal entity, edition, data path and definitions. Support and exit need the same detail.
A general product category proves none of those points.
A managed service adds people who investigate or operate the process. UK agencies such as Hallam, based in Nottingham, and Croud, with offices in Shrewsbury and London, sell conversion research and testing retainers.
Published rates are rare; UK managed conversion retainers typically run from about £2,000 to £10,000 a month depending on scope.
Record their access, instructions, deliverables, evidence and correction authority. Subcontractors need the same treatment.
Do not infer expertise from the word managed.
A buyer-built route may provide configuration control, but the merchant owns development, monitoring, documentation, support and recovery. Segment's free tier covers 1,000 monthly tracked users; its Team plan starts at $120 a month. BigQuery gives 1 TB of queries and 10 GB of storage free each month.
dbt Cloud is free for one developer. That control is not automatically safer or cheaper.
Use one evidence grid
One evidence grid fields
- Store statescheckout started to order confirmed
- Refund requested, issued, dispute recorded
- Source, timezone, identity, duplicates
- Late records, missingness, version
- Privacydevice, personal data, purpose, controller
- Processor, recipients, retention, transfer, deletion
- Securityaccess, secrets, logging, incident, backup
The government's online-selling guidance keeps customer information, payment obligation and order correction within the operating case.
Privacy fields cover device access, personal data, purpose and controller. Processor, recipients, retention and transfer need separate rows. Deletion closes the set.
The ICO's storage and access technology guidance applies to relevant cookies and similar operations. A managed route does not shift the merchant's analysis by default.
Security fields include privileged access, secret handling, logging and incident evidence. Backup, restore and supplier dependency follow.
Use the NCSC's secure online service guidance as an authoritative topic map, then require evidence for the real connection.
Consumer rows record the seller identity, product description, total-price source and optional choices. Confirmation, cancellation, returns and support follow.
Product-safety rows identify warnings, traceability, correction and withdrawal authority. None of these fields receives a score from the event report. They are pass, hold or not applicable decisions made by the responsible reviewer.
Accessibility uses the same customer tasks, including error recovery and refunds. W3C WAI's evaluation overview explains why tools cannot alone determine accessibility. Every route therefore keeps a knowledgeable evaluation and user-relevant evidence field.
The buyer's dossier leaves commercial rows blank for currency, VAT, unit, period and term. Usage and support threshold stay blank too. Change notice, liability and insurance stay blank too.
Public list prices fill part of that gap. Shopify Basic is about £25 a month; Clarity is free; Hotjar paid plans typically start near £32 a month; Segment Team starts at $120 a month. BigQuery gives 1 TB of queries free each month.
Whole-life cost also includes buyer labour, integration, assurance, incidents and exit. An accountant and VAT adviser define money treatment.
Payment rows use the same definitions across routes. Authorisation request, authorised amount, capture, settlement and refund form one set. Dispute is separate.
Record the provider entity and reconciliation source. Stripe charges 1.5% plus 20p for UK cards, and refunds do not return that fee.
Where the managed service sees only a subset, or the manual route receives a delayed file, show that missingness rather than normalising it away.
Compare proof, not promises
Use synthetic or specifically authorised cases with known states. Require each route to reconcile the same inputs, expose missing records and preserve an audit trail.
Proof test sequence
- Use synthetic or authorised cases with known states
- Reconcile same inputs across all routes
- Expose missing records and preserve audit trail
- Rehearse pause, correction, export, restoration, deletion
- Mark route unknown when evidence is absent
- Run exit exercise against documented manifest
Require Shopify exports, Clarity replays, Hotjar recordings, agency reports and Segment or BigQuery outputs to pass that test.
Rehearse pause, correction, export, restoration and deletion. Mark a route unknown when its evidence is absent rather than awarding a lower invented score.
Run the exit exercise against a documented manifest of records, definitions and configuration. The buyer should confirm what can be restored without the supplier, what remains dependent and when residual copies are due to be removed.
An export file that cannot recreate the decision record is incomplete evidence.
The result may favour the current method, admit a route to further testing or hold all options. No route can win by offsetting a privacy, consumer, accessibility or security failure with convenience.
Before you act
Before you act
- Fix one job before comparing routes
- Record identical evidence fields for every route
- Define source, timezone, identity, duplicates, missingness
- Require evidence, not product category
- Rehearse pause, correction, export, restoration, deletion
- Run exit exercise against documented manifest
Common questions
Why does the article say no route can be declared the winner?
Because the buyer must set weights and thresholds, and no exact-edition evidence or controlled test has been supplied. Shopify exports, Clarity and Hotjar can be compared on the grid. So can Hallam, Croud, Segment and BigQuery. None wins by default.
What must each route reconcile on the same evidence grid?
Checkout started, payment authorised, payment captured and order confirmed. Also refund requested, refund issued and dispute recorded. Each route must expose missing records and preserve an audit trail, using synthetic or specifically authorised cases with known states.
What does the exit exercise require the buyer to confirm?
The buyer should confirm what can be restored without the supplier, what remains dependent and when residual copies are due to be removed. It runs against a documented manifest of records, definitions and configuration. An export file that cannot recreate the decision record is incomplete evidence.



