How should you compare customer data platforms?

The short answer

Start with one customer decision, then test whether each platform can create and maintain the customer truth required to improve it. Compare the complete operating model, not only the software demonstration.

A strong shortlist should make identity, audience rules, permissions, destinations, refresh schedules, implementation work, ownership and total cost visible. If a vendor cannot explain how those pieces work together for your first use case, a longer feature list will not solve the problem.

1. Define the decision before comparing technology

Write the first use case in business language. For example: stop paying to acquire current customers, change media treatment when a lead qualifies, re-engage people at an agreed lapse point, or create an acquisition seed from customers with observable value.

Name the source field that proves eligibility, the destination where the audience will be used, how current it must be and the commercial measure that could change. This becomes the common test for every vendor.

2. Ask whether the platform maintains a customer record

A customer data platform should create and maintain a persistent unified customer record as the same person appears across systems and changes state. A campaign list assembled only when requested is useful activation work, but it is not the same customer layer.

Ask how records are merged and separated, which identifiers are used, how conflicting sources are prioritised, what history is retained and how corrections flow through. Fractyl uses visible deterministic identity and source-precedence rules so the logic can be inspected rather than treated as a black box.

3. Test the actual source data, not a generic connector catalogue

• Which CRM, warehouse, analytics, operational and structured-file sources are required for the first use case?
• Are those connections live, configurable or dependent on custom implementation?
• What fields, identifiers and timestamps are available and reliably maintained?
• How are consent, opt-out, invalid-record and lifecycle states represented?
• What data-quality problems must be fixed before the first audience is safe to use?

4. Confirm the destinations and refresh cadence you will use

A destination is useful only when the organisation can operate and measure the audience there. Confirm the specific advertising accounts, clean-room environments or owned channels required, then separate live integrations from beta and roadmap statements.

Fractyl currently activates governed audiences to Google Ads, Meta, TikTok, DV360 and supported third-party data clean rooms. Synchronisation can run on different schedules, including up to hourly, according to the customer decision and destination requirement. Email and SMS are planned rather than current live destinations.

5. Separate identity resolution, hashing and platform matching

Identity resolution determines whether source records belong to the same person. Hashing transforms selected identifiers into consistent non-plain-text values for governed matching or activation. A media platform then attempts to match those values to its own users.

These are different steps. Hashing does not clean data, resolve duplicates, create permission or guarantee destination match rate. Ask every vendor to show where each step occurs and who owns the rules.

6. Make governance operational

• A named purpose and accountable owner for each audience.
• Documented inclusion, exclusion, recency and suppression rules.
• Approved source fields, destinations and refresh schedules.
• Access controls and a change trail for audience definitions.
• A process for delayed data, failed synchronisation and incorrect matches.
• Clear treatment of consent, deletion, correction and opt-out signals.
• Client legal, privacy and security review for the proposed implementation.

7. Compare implementation promises with implementation work

Easy implementation should mean a smaller, better-defined first release, not that customer data stops being complex. Ask what the vendor needs from marketing, CRM, data, IT, privacy, security and the incumbent agency before an audience can go live.

A credible plan should cover data discovery, identity design, audience logic, destination configuration, validation, ownership and measurement. It should also show what is explicitly outside the first scope.

8. Choose the operating model as carefully as the platform

Some organisations have dedicated data engineering and marketing-operations teams. Others need a managed service, agency partner or shared model to keep data, audiences and destinations current.

Fractyl can be managed by the Fractyl team, licensed to an agency or client team, or operated through a shared model. It is agency-neutral, so the incumbent media agency can retain media strategy and execution while using the customer signal Fractyl provides.

9. Compare total cost, not only licence price

• Initial data discovery, identity design and implementation.
• Platform licence or managed-service fee.
• Cloud storage, processing and monitoring.
• Source, destination and clean-room integration work.
• Privacy, security, procurement and legal review.
• Internal or partner time for operation, quality assurance and measurement.
• Custom development required for future use cases or destinations.

10. Ask how value will be proved

The first use case should have an observable value path. For customer suppression, that might include source audience size, destination match rate, eligible acquisition exposure removed and mistaken exclusions. For lead-stage or value signals, it may include qualified conversion and downstream customer value.

Ask for the baseline, comparison or holdout design before launch. Treat audience delivery and match rate as operational evidence, not the final commercial result.

11. Score current capability separately from product direction

Create three columns in the evaluation: live now, beta, and roadmap. Apply the buying decision to current capability unless a contractual release commitment says otherwise.

For Fractyl, paid-media activation and supported third-party clean-room delivery are live. Media mix modelling is in beta. AI-assisted audience segments and attribution feedback are roadmap direction. These future capabilities should not be scored as if they are already available.

12. Use a weighted shortlist, not a feature-count contest

• First-use-case fit and measurable value: 25%.
• Identity, data quality and persistent customer record: 20%.
• Required sources, destinations and refresh cadence: 15%.
• Governance, privacy and security fit: 15%.
• Implementation and ongoing operating burden: 15%.
• Total cost and commercial flexibility: 10%.

A practical Fractyl fit check

Fractyl is strongest where an organisation has meaningful first-party customer or lead data, customer states that should change media treatment, ongoing paid-media investment and a team or partner able to own the audience decision.

It is less likely to be the right first choice when the primary requirement is broad real-time web personalisation, hundreds of prebuilt connectors or a global customer-experience suite. The working session should test the real data, audience and commercial case before a proposal is made.