Seven Types of Customer Data Inside a CDP

There are seven types of customer data inside a CDP. Not seventy. Seven.

To get full value from your platform, you need all seven. And you need the data in good shape.

Once you see this, the CDP starts making sense. And when something breaks, you know where to look.

Profile data

Who the customer is. Name, email, company, segment. Comes from your backend, databases, signup forms, CRM, support tools, loyalty programs, logins, product subscriptions. This is the identity layer. Everything else attaches to it. Get this wrong and everything downstream gets dirty, loses traction, loses value.

You will never get this to 100%. That is the nature of profile data. But the difference between 95% and 99.9% is where most of the value sits. All other data connects to the profile, so it has to be paired correctly. No duplicates. No missing IDs. Even when your customer uses multiple devices, sometimes anonymously.

And it gets harder in layered services. Take telco. One ID is for billing. Another is the user. Another is household or service point. There are different services, at different locations, some paired or grouped, some not. Profile data sounds simple until you have to get it right.

Behavioral data

What the customer does. Page views, app events, searches, clicks. This is real-time signal. It tells you someone looked at pricing three times this week but never clicked buy. Or someone is searching how to cancel the service. Or checking if the sport channel is included in their current package. Or checking if the room has a garden view. Or browsing for more brown high heels in your catalog.

Most teams collect this data. Few actually use it well.

This is as close as you get to reading your customer's intent in real time. I am not saying it is easy. Seeing three page views does not tell you if your customer is fed up or in love. But it tells you something. Do not just collect it. Look at it. Listen. Get closer in small steps. Do not say awkward things. Adjust as you learn. That is what personalization actually is.

Transactional data

What the customer bought. Or wider. What the customer committed to. Orders, returns, subscriptions, payments, renewals, cancellations.

But look beyond e-commerce. What did the customer book? What did they reserve? What subscription did they renew? What service did they cancel? In insurance, when did the policy start and when did it lapse? In healthcare, when was the visit and what procedure was performed?

Every transaction is a moment with a business outcome. Someone bought, booked, paid, renewed, or walked away. That is what makes it transactional. If it is ongoing usage or a pattern over time, that is behavioral. If it is a response to something you offered, that is engagement.

This is where revenue lives. But on its own, it mostly tells you what already happened.

Engagement data

What you sent them and how they responded. Email opens, SMS clicks, push notifications delivered. This closes the loop between your campaigns and their behavior. Without it, you are guessing which messages work.

But engagement is not only about channels. It is also about what you offered. Was the upgrade even seen? Did the discount lead to anything? Was the retention offer accepted, ignored, or turned down?

Channel data tells you if you reached them. Engagement data tells you if what you put in front of them worked.

Consent data

What you are allowed to do. Opt-ins, opt-outs, legal basis. This is not a nice-to-have. Without this layer, the other six are a liability.

Straightforward in theory. But tightly connected to profile data in practice. You need to be sure that every action you take is lawful. That the consent tied to a profile is valid and current. And that there is no duplicate profile of the same customer sitting somewhere with a different, outdated consent status.

One thing worth separating: a preference center often collects consent and preference data at the same time. They look similar. They are not. "Unsubscribe from marketing emails" is consent. It is a legal boundary. "I'd rather hear from you by email" is preference. You should respect it, but ignoring it is bad manners, not a legal problem.

This is where identity problems become legal problems.

Preference data

What the customer says they want. Quiz answers, stated interests, communication choices. Not the same as profile. Profile says who they are. Preference says what they asked for. Explicit, high signal, and one of the most reliable inputs for personalization.

When was the last time you asked your customers what they want? What they value? What they expect? Not guessing from signals. Not inferring from clicks. Actually asking.

This is zero-party data. The foundation of every good relationship. Ask the other person how to make their life easier. Then act on it.

Derived data

What you calculated. Segments, scores, predictions, CLV, churn probability. Built from everything above. This is where most of the business value gets created. But it is only as good as the six layers feeding it.

Create what you can use. Stay curious. Optimize. Test. Start there and build from what you learn. The results will follow.

Those seven types cover the customer. Everything else in a CDP describes the world around the customer. Product catalogs, store locations, currency, shipping zones. That is metadata. It gives context to events and transactions, but it is not customer data. Identity resolution is not a data type either. It is the logic that makes profile data work. Everything else is supporting context.

This tells you what customer data is inside your CDP. But it does not tell you who collected it, or what rights you have. That is a different question.

Next post.

Appendix: how the seven types show up across industries

E-commerce

Profile: customer account, name, email, address, loyalty tier

Behavioral: product views, searches, filters applied, cart adds, wishlist activity, price comparison page visits

Transactional: order placed, return processed, refund issued, subscription renewed, gift card redeemed

Engagement: opened abandoned cart email, clicked discount campaign, ignored restock notification, responded to review request

Consent: marketing opt-in, cookie preference, data sharing choice

Preference: "favorite size is 42," "I want offers about running shoes," "gift wrapping by default"

Derived: CLV, purchase frequency, predicted next order date, preferred category affinity

Retail

Profile: loyalty card, POS identity, household, in-store account

Behavioral: store visits, dwell time, in-app browsing, barcode scans, fitting room requests

Transactional: till receipt, click-and-collect pickup, in-store return, loyalty points earned, gift card purchased

Engagement: responded to loyalty promotion, used in-store coupon from email, ignored push notification about flash sale

Consent: marketing opt-in, location tracking permission, loyalty program terms

Preference: "preferred store is downtown," "interested in organic products," "no paper receipts"

Derived: basket analysis, cross-store shopping pattern, churn risk, household spend estimate

Banking

Profile: KYC data, verified identity, address, account type, segment

Behavioral: app logins, balance checks, feature usage, branch visits, mortgage calculator runs

Transactional: transfer completed, loan drawdown, card payment, account opened, account closed, standing order set up

Engagement: responded to credit card offer, opened rate change notification, clicked financial health nudge, ignored savings prompt

Consent: marketing consent, data sharing consent, regulatory consent

Preference: "preferred branch is city center," "statements by email," "no calls about insurance"

Derived: credit risk score, product propensity, lifetime value, fraud probability

Telecoms

Profile: billing entity, user ID, household, service point

Behavioral: data consumption, call frequency, roaming usage, app usage patterns, how often they check their balance

Transactional: contract signed, plan upgraded, device purchased, bill paid, service cancelled, add-on activated

Engagement: responded to upsell offer, clicked retention campaign, accepted or declined a plan change you proposed

Consent: marketing consent, network analytics consent, location data consent

Preference: "I want paperless billing," "do not call me with offers"

Derived: average monthly spend, churn risk score, household structure

Hospitality

Profile: guest record, name, contact, loyalty tier, travel document

Behavioral: browsing room types, checking availability for specific dates, looking at restaurant menus, searching "late checkout"

Transactional: room booked, stay completed, spa treatment purchased, minibar charged, cancellation processed, loyalty points redeemed

Engagement: opened pre-arrival email, responded to upgrade offer, left a review after a prompt

Consent: marketing consent, data sharing across properties in a chain

Preference: "high floor," "garden view," "extra pillows," "early check-in when possible"

Derived: lifetime stays, average spend per visit, preferred property tier, rebooking likelihood

Fintech

Profile: onboarding data, verified identity, risk tier, account type, linked accounts

Behavioral: app usage frequency, feature adoption, portfolio checks, support chat initiated, comparison tool usage

Transactional: deposit made, withdrawal processed, investment executed, fee charged, subscription upgraded, payout completed

Engagement: responded to product recommendation, opened fee alert, clicked referral campaign, ignored educational content

Consent: regulatory consent, investment suitability, data sharing with partners

Preference: "weekly portfolio summary," "notify me only above €500," "risk appetite: moderate"

Derived: spending pattern, investment risk score, product fit prediction, upgrade likelihood

Travel / Airlines

Profile: passenger record, frequent flyer ID, travel document, tier status

Behavioral: route searches, fare comparisons, seat map views, app check-in frequency, destination browsing

Transactional: flight booked, upgrade purchased, ancillary added, cancellation processed, miles redeemed, lounge pass bought

Engagement: responded to fare alert, accepted rebooking offer, clicked disruption notification, ignored loyalty promotion

Consent: marketing consent, partner data sharing, location tracking

Preference: "aisle seat," "vegetarian meal," "morning departures," "home airport is Vienna"

Derived: route loyalty, ancillary spend propensity, upgrade likelihood, disruption sensitivity

Insurance

Profile: policyholder identity, coverage type, risk profile, household structure

Behavioral: quote comparisons, policy document views, claims portal usage, app logins, coverage calculator runs

Transactional: policy purchased, policy renewed, policy lapsed, claim submitted, claim settled, payout completed

Engagement: responded to renewal reminder, opened cross-sell offer, clicked claims status update, ignored prevention tips

Consent: marketing consent, health data sharing, third-party assessments

Preference: "contact by email," "monthly payment," "no paper documents"

Derived: claims propensity, risk score, retention probability, cross-sell fit

The pattern

If there is a moment, an exchange, and a business record, it is transactional. If it is ongoing activity or a pattern, that is behavioral. If it is a response to something you initiated, that is engagement. If the customer stated it, that is preference.