
Strategy
Planning a customer data platform decision around the data journey, not the product class
Plan a customer data platform decision in England through a defined data journey, comparable routes, independent gates and reversible evidence tests.
Planning a customer data platform decision around the data journey, not the product class
Begin with a correction that must reach its intended destinations, not a platform label. For a team in England, the decision is which controlled route can carry an authorised correction and suppression instruction from a service record to two approved destinations while retaining provenance. Options include a process, approved tool, buyer-built route, hosted service or managed integration.
What would count as a useful first test?
A useful first test joins a defined buyer task to a reversible change, so the route is judged against evidence rather than appearance. A ninety-day window can organise planning if the buyer chooses it, but it is not a universal timetable; use the planning envelope as an optional frame.
Use fictional records or records specifically authorised for the exercise, with production activation disabled throughout. Set expected outcomes before running it; include duplicates, absent identifiers, conflicting preferences, uncertain joins, corrections, suppression and deletion. A versioned workflow is a useful operating reference, but the trial must evaluate the buyer's own route.
Acceptance should show that an independent reviewer can reproduce a join, provenance remains visible in an export, corrections reach every downstream copy, and uncertainty is not concealed. Check that administration and review tasks are accessible, and that recovery works after a failed change. The exercise should also allow inspection and stopping without sending a live audience or instruction.
Which parts of the work need a CDP at all?
First establish the functions and hand-offs the buyer actually needs; a CDP label alone does not show that an edition supplies them. Make required, excluded and combined functions explicit before assessing routes, using the selection checklist to record those boundaries.
A CRM may hold relationship records, while a warehouse or lake can store and transform broader data. Identity resolution suggests associations among identifiers. Consent and preference management records choices and their scope, tag management governs deployment of collection technologies, analytics describes events, and activation conveys a defined audience or instruction to another service.
A clean room is a bounded collaboration arrangement, while managed integration adds external labour and operating responsibility. Describe the decision consequence rather than a technology aspiration: better correction evidence can be observed, but improved personalisation remains vague until the affected population, action, baseline and acceptable proof are specified.
Do not set revenue, conversion or efficiency objectives without buyer-owned measurements and a credible design. A demonstration may illustrate some functions, but it cannot establish that the precise edition covers the buyer's task or that the remaining hand-offs suit the buyer's operating needs.
What should I be able to trace across the journey?
A reviewable record follows information from collection through change, destination, correction, suppression and deletion. Capture the owner, purpose, content, controls and evidence at each transition, tied to the actual task, rather than relying on a broad product description.
At every source, note its owner, collection authority, the person or organisation represented, identifiers, date range, quality concerns and retention. Record which fields are genuinely needed for the stated purpose, so data minimisation is considered alongside the source details rather than after a route has been chosen.
Document every join rule with a version and say whether it is deterministic or uncertain. Explain how a mistaken association can be repaired, preserve provenance for profiles and inferences, and do not turn a likelihood into a fact merely because it has been saved as a profile attribute.
At each destination, record the purpose, receiving legal entity, transferred fields, transmission method, permitted action, correction path and deletion proof. A suppression instruction should survive a later import; it should not be undone when a new identifier causes a separate profile to be formed.
The ICO's data protection principles are the UK reference for purpose limitation, accuracy and storage limitation. Inaccurate personal information must be corrected or erased without delay. These principles do not decide the buyer's lawful basis, retention period or identity approach.
Does technical capability give me permission?
No. The ability to ingest an identifier, infer a connection or send an audience does not establish collection authority, transparency or lawful use. Evaluate permission for each source and intended destination separately from technical capability, and retain evidence for both assessments.
Before creating profiles, inventory cookies, pixels, scripts, software development kits, link decoration and comparable technologies. PECR assessment depends on purpose and operation, not the chosen label. Calling a method first-party or server-side does not settle whether it is covered; PECR cookies guidance can inform the review.
Classify later communications separately: service notices, genuine research, advertising audiences and promotional electronic messages are different activities. The ICO updated its direct marketing guidance on 28 April 2026 to reflect the Data (Use and Access) Act commencement schedule. A preference flag is an input to review, not proof that every destination and communication is permitted; see the UK GDPR and PECR rules.
Who determines the purposes and essential means?
Work out the roles from actual decisions: who determines processing purposes, who chooses essential means, who follows instructions and who receives outputs for its own purpose. A role name in a proposal cannot settle the question; assess how the arrangement operates in practice.
Under the UK GDPR, a controller determines the purposes and means of processing, while a processor acts on behalf of the controller. The ICO explanation of roles sets out these definitions. An activation destination may have responsibilities beyond those shown in the buyer's first diagram.
Where the arrangement is controller-to-processor, qualified legal review should examine processor instructions, subprocessors, assistance, security, audit, return and deletion against the exact service. The ICO contracts and liabilities guidance is under review after the Data (Use and Access) Act, so check it again at contracting.
International transfers depend on access and recipient facts. Check ICO guidance where a restricted transfer may arise. A region name on a sales page does not resolve support access, replication or subprocessor routes; trace the wider data protection duties across the same arrangement.
How can I compare routes without favouring one?
Assess each route against the same buyer case and evidence fields. This makes buyer effort, proof needed before a trial and exit conditions visible beside technical functions, while giving each option a fair basis for rejection.
- Current manual process:The buyer handles extracts, rules, review and deletion. Before a trial, require repeatable steps, an access record and an exception log. For exit, another operator should be able to restore the process and explain its decisions.
- Existing approved tools:The buyer configures an already governed set of systems. Before a trial, require the exact approved scope, current configuration and a gap analysis. For exit, required records must export without an unapproved dependency.
- Buyer-built route:The buyer owns code, infrastructure and continuity. Before a trial, require versioned lineage, identity logic, tests, access details and recovery evidence. For exit, a different team should be able to operate and recover the route.
- Licensed hosted service:The supplier provides a defined service; the buyer controls configuration and acceptance. Before a trial, require the legal entity, edition, roles, subprocessors, security, accessibility, contract and export details. For exit, a synthetic export should restore with usable provenance.
- Managed integration:External people design or run an agreed journey. Before a trial, require deliverables, methods, evidence rights, staffing continuity and handover details. For exit, the buyer should receive mappings, records and deletion evidence.
Put commercial units beside the technical assessment, not inside its score. The cost guide identifies fields to record: currency, VAT basis, billing period, term, profiles, events, sources, destinations, seats, service hours, implementation and buyer labour. Leave each amount blank until a current applicable record is verified.
What could make me reject an otherwise attractive route?
Give privacy and PECR authority, identity accuracy and rights, security, accessibility and exit their own pass-or-fail decisions. A low quote, polished interface or strong result elsewhere must not outweigh a failure in any one check for the route being assessed.
Assess security of processing against the actual service and the buyer's evidence needs. Include least privilege, administrator and service accounts, logs, support access, incident notification, recovery and change control; do not rely only on a general assurance statement.
Test accessibility requirements against the buyer's critical configured tasks and outputs. For SaaS routes, ask about audit, incident management, return, deletion and termination. Make these checks against the service and configuration being considered, not a generic description of a product class.
Where processing is likely to result in high risk, screen the design for a data protection impact assessment. A completed template cannot replace assessing the real sources, people and consequences. Keep this assessment distinct from the other checks rather than treating it as a substitute for them.
Stop if data authority is unclear, false joins cannot be explained, suppression is rebuilt, privileged access is unknown, an essential task is inaccessible or exit evidence fails. Keep buyer-set thresholds blank until the owner can justify them, so a test does not acquire arbitrary pass conditions.
Claims about accuracy, lift or commercial outcomes need evidence that matches the wording and population. The CAP Code requires marketers to hold documentary evidence for objective claims before publication. A synthetic functional test cannot support a claim about live customer results.
When should I review the decision again?
Record the chosen route, rejected options, supporting evidence, remaining unknowns, owners, costs still blank, conditions of use, review date and exit trigger. Revisit the decision when a material part of the journey or its operating arrangement changes.
Use the planning template for a blank decision record. Improving the current process rather than buying a platform may be the right outcome. If approvals, source authority or recovery proof are unavailable, pause instead of compressing the work to fit a chosen date.
Reopen the plan if the purpose, source, identifier, join, inference, destination, subprocessor, ownership or contract changes. Recheck each regulator and supplier record on publication day and before any live processing, so the evidence still applies to the route in use.
What should I do first?
Start with a short record of the buyer's task, the journey it touches and the evidence needed to decide whether the route works. Make the scope testable before comparing products or services by taking these first steps:
- Name the decision consequence, source, state change and destinations.
- Mark required, excluded and combined functions, including each hand-off.
- Record sources, transformations, destinations and who makes processing decisions.
- Inventory cookies, pixels, scripts and similar technologies before planning new profiles.
- Compare routes on one buyer case, with separate privacy, identity, security, accessibility and exit checks.
- Test fictional or authorised records with production activation disabled, then record stop conditions.
In this guide
- Customer data platform objections: a direct-marketing suppression runbookHandle direct-marketing objections in an England CDP with daily hand-offs, profile and destination checks, suppression tests and objection deadlines.
- Customer data platforms planning template: traceability fieldsUse a customer data platforms planning template to trace journey stages, role decisions and evidence links so a UK reviewer can follow each record.
- Five delivery routes for one customer data journey, compared without picking a winnerCompare five CDP delivery routes for one England buyer journey using common data, control, cost, accessibility and exit criteria without a winner.
- Eight customer data platforms strategy mistakes to avoidAvoid eight observable customer data platforms strategy mistakes, from the single-view slogan to exit, with regulator-backed corrections and evidence limits.
- Customer data platforms: five evidence gatesFor a customer data platforms ninety day plan, use five evidence gates—not a timetable—to decide whether an England buyer should adopt, extend or retain.



