Checklist: How to choose a customer data platform (CDP)

Choosing a CDP for retail is harder than it should be. Every vendor claims to do everything. Feature lists are long and largely identical. And the stakes are high enough that a wrong decision costs you 12 to 24 months of switching cost on top of the platform fee.

Here is what retail marketers should actually look for in a CDP, and the questions to ask before you sign anything.

What should retailers look for in a CDP?

The capabilities that matter for retail are different from the capabilities that matter for B2B SaaS or financial services. If a vendor's reference customers are mostly outside retail, their platform is probably built for a different set of problems.

A retail-native data model

Retail customer data comes from multiple sources that most CDP vendors treat as an afterthought: point of sale transactions, loyalty programme events, in-store interactions, ecommerce orders, email engagement, and paid media behaviour. A CDP built for retail should ingest all of these natively, without custom engineering, and map them to a customer profile model that understands what a retail purchase record looks like.

Ask: which retail data sources does the platform connect to out of the box? What does the standard retail data schema look like?

Identity resolution that works offline-to-online

A customer who shops in-store with a loyalty card, then browses online from a different device, then clicks a paid ad on mobile is one person. Most CDPs handle online identity resolution reasonably well. Fewer handle the offline-to-online connection without significant custom work. For omnichannel retailers, this is the capability that determines whether the platform delivers a single customer view or another siloed dataset.

Your customer identity resolution layer should connect deterministic identifiers (email, loyalty number, phone) with probabilistic signals (device, behaviour) and update unified profiles continuously, not at batch intervals.

Real-time segment activation

A segment that was accurate yesterday is not useful for today's campaign. If your CDP processes segments in nightly batches, a customer who crossed a high-value threshold this morning will not receive the relevant communication until tomorrow. For triggered campaigns, including churn prevention, win-back, and post-purchase, real-time or near-real-time segment updates are a functional requirement, not a nice-to-have.

Check: how frequently does the platform update segment membership? What is the latency between a customer event and a segment change?

Activation reach across your channels

The CDP is only valuable if its segments reach every channel you operate. Look for native or near-native connectors to your ESP (Klaviyo, Braze, Attentive), paid media platforms (Meta, Google), your in-store or clienteling tool, and your website personalisation layer. Each additional integration point that requires custom development is a maintenance burden and a timeline risk.

AI and first-party data readiness

This one is newer but increasingly important. As third-party cookies continue to deprecate and AI-driven search changes how buyers find vendors, CDPs that can generate structured, enriched first-party signals are more valuable than those that simply store data. Predictive models for CLV, churn risk, and next-best-product, built from your own first-party data, are the signals that improve both your marketing performance and your visibility in AI-powered search results.

Ask: what predictive models does the platform run natively? Can it export structured audience data for use in downstream AI applications?

Speed to value and support model

A CDP that takes 12 months to implement is a CDP that costs you 12 months of opportunity. Ask specifically about implementation timelines, what is required from your team, and what happens when something goes wrong post-launch. The difference between a vendor who gives you software and a vendor who acts as a partner is most visible when the implementation hits its first complication.

What questions should you ask CDP vendors in a demo?

Most vendor demos are designed to show you the platform at its best. These questions are designed to reveal where it struggles.

1. Can you show us how a specific retail customer event, such as an in-store POS transaction, flows through to a unified profile in real time? (Not a pre-built demo scenario — ask them to show the live data flow.)
2. How does the platform handle identity resolution when there is no email address captured at POS? (This is the most common offline data gap in retail. Generic answers here are a red flag.)
3. What is your standard implementation timeline for a mid-market omnichannel retailer with three to five data sources? (Get a commitment, not a range.)
4. Which of your retail reference customers could we speak with? (Ask for two or three names, and follow through. Peers are your best source of honest implementation experience.)
5. How do segment updates work? What is the latency between a customer event and the customer appearing in the relevant segment? (Anything longer than a few minutes is worth pushing on for triggered campaign use cases.)
6. What predictive models does the platform run natively, without professional services? (CLV, churn risk, and propensity models should be table stakes. If these require an add-on or custom build, factor that into the cost.)
7. How does the platform handle customers who appear in your POS system but have never transacted online? (The answer reveals how well the platform was actually built for omnichannel retail.)
8. What does your support model look like post-implementation? (Is there a dedicated customer success team? What are the SLAs? Who owns the relationship day to day?)
9. Can your platform export structured audience data in a format optimised for use in paid media AI targeting or LLM retrieval? (This is forward-looking but increasingly relevant as AI-powered ad platforms and search tools rely more heavily on structured first-party signals.)
10. What are the most common reasons implementations fail on your platform, and what do you do to prevent them? (A vendor who cannot answer this honestly is not a vendor who will help you when things get hard.)

What separates retail CDPs from generic CDPs?

The distinction is in the data model, the out-of-box connectors, the default segment logic, and the reference customer base.

A generic CDP is designed for a broad range of industries. Its data model is flexible, which means it is also undifferentiated. You can usually make it work for retail, but you will spend significant time configuring what a retail-native CDP handles by default: POS transaction schema, loyalty event types, in-store behaviour capture, product catalogue integration.

A retail CDP is built around the retail customer relationship as a starting point. The default data model understands what a transaction record looks like. The default segment library includes RFM-based segments, lifecycle stages (active, at-risk, lapsed, new to brand), and category affinity groupings. The connectors are pre-built for the platforms retail teams actually use.

For retail marketers built for retail, the practical difference is months of implementation time and a lower ongoing engineering cost.

The most common CDP selection mistakes

Evaluating features instead of outcomes. A long feature checklist is not a buying framework. Anchor your evaluation on two or three specific use cases you need to solve in the first 90 days, and assess each vendor against those use cases specifically.

Ignoring the data capture gap upstream. A CDP is only as good as the data flowing into it. If your POS does not capture email addresses at checkout, or your loyalty programme is not connected to your ecommerce platform, the CDP cannot unify what it cannot see. Audit your data capture before you select a platform, and choose a vendor who will help you close those gaps.

Underestimating total cost of ownership. Platform fee plus implementation plus professional services plus ongoing support can look very different from the headline number. Get a total cost breakdown for year one and year two before comparing vendors.

Not involving the marketing team in vendor selection. IT-led CDP selections often optimise for data architecture and security. Marketing-led selections often optimise for ease of use and channel activation. The best evaluations involve both. The CDP has to serve marketing's day-to-day use cases, not just pass a technical audit.

FAQs

What should retailers look for in a CDP?

Retailers should prioritise: a retail-native data model that handles POS and loyalty data natively; identity resolution that links offline and online customer records; real-time or near-real-time segment updates; activation reach across ESP, paid media, and in-store channels; native predictive models for CLV and churn risk; and a support model that extends beyond implementation.

What questions should I ask CDP vendors in a demo?

The most revealing questions focus on the hard parts: how the platform handles offline-to-online identity resolution, what the real implementation timeline looks like for your specific data sources, what predictive models are native versus add-on, what the post-implementation support model includes, and what the most common reasons for implementation failure are on their platform.

How do I evaluate a CDP for retail?

Start with two or three specific use cases you need to solve in the first 90 days. Use those as the anchor for every vendor evaluation. Assess real-time segment capabilities, retail data source connectors, identity resolution approach, and total cost of ownership across year one and year two. Run a proof of concept with your actual data before making a final decision.