A customer data platform pays only when the profile acts, not when it reports.
A customer data platform (CDP) collects data from every source into one profile per customer. For an online store, the profile is worth what it does: the products it picks on the site, the messages it sends and when, and the audiences it hands to your ads.
- Data and analytics CDPs unify and report. Actionable CDPs also act on the profile.
- Most store traffic is anonymous, so the profile must start at the first visit
- A CDP is not a CRM and not a DMP
A CDP gathers what you know about each customer, from the site, the app, orders, email, the CRM and the shops, and joins it into one profile. That is the easy half. The hard half is using the profile, and most CDPs were not built for it.
Early CDPs were built for data teams: collect, clean, model, report. To act on the result, marketers still had to export segments into an email tool, an ad account and a recommendation widget, each with its own copy of the truth. Actionable CDPs, sometimes called campaign CDPs, close that gap by acting on the profile in the same place it is kept.
For a store, three things decide whether a CDP pays: it profiles anonymous visitors, not only customers who gave an email; it predicts what each customer will do, not only what they did; and it acts on every channel from that one profile, with a holdout group that shows the lift in your own orders.
What a customer data platform is
A CDP is software that collects customer data from separate sources and joins it into one profile per person. The sources are the usual ones for a store: site and app events, orders, the email and SMS tools, the CRM, the help desk, analytics, the data warehouse and, for stores with shops, the till.
Joining them is identity resolution: working out that the phone that browsed on Monday, the laptop that bought on Wednesday and the email address that opened Friday's newsletter are one person. Once they are, the profile carries everything in one place: identifiers, what the person viewed and bought, which sizes, colours and prices they consider, which messages they opened, and what they told you in a survey.
A CDP also gives you one place to manage consent and data protection. Under GDPR that matters: a deletion request or a withdrawn consent has to reach every copy of the data, and the fewer copies there are, the easier that is.
Two kinds of CDP
All CDPs build a single customer view. They differ in what happens next.
Data and analytics CDPs
Data CDPs control and integrate the data: schemas, cleaning, governance, pipelines into the warehouse. Analytics CDPs sit on top and visualise it, so teams can run analyses and find insights. Both are built for technical roles. Even a marketer with the skills to use them has to move the unified data downstream to the tools that send email or buy ads, and each move adds a delay and a copy that drifts.
Actionable CDPs
Actionable CDPs also unify the data, but they carry the tools that use it: segments, predictions, on-site personalization, messages on several channels and journeys between them. The profile that knows a shopper looked at running shoes this morning is the same profile that decides which shoes the homepage shows tonight and what the email says tomorrow. Nothing is exported, so nothing goes stale.
For an online store, the second kind is the one that pays. A unified profile that only feeds a dashboard tells you what happened. A profile that acts changes what happens next.
CDP, CRM and DMP compared
Three tools that all hold customer data, built for different jobs.
| CDP | CRM | DMP | |
|---|---|---|---|
| Main job | One profile per customer from every source | Track conversations with sales and support | Target paid ads at anonymous audiences |
| Data it holds | Behaviour, orders, events, surveys, consent, contact details | Emails, tickets, calls, deals, contracts | Cookies, device IDs, IP addresses, mostly third-party |
| Who it knows | Known customers and, in the better ones, anonymous visitors | Leads and customers who spoke to your team | Anonymised segments, not people |
| Used for | Personalization, messages, segments, ad audiences | Sales pipeline and service history | Media buying through ad networks |
Three things a store should do with a CDP
1. Treat anonymous visitors as people
Most CDPs profile known customers: the ones who gave an email or a phone number. On a typical store that is a small share of the traffic (see why one agent). The rest browse, leave and are never seen again. A useful CDP starts a profile on the first visit, records the pages, products and attributes the visitor looks at, and uses it straight away: products on the site picked for that person, and a push notification to bring them back if they opted in (anonymous visitors).
When the visitor later gives an email or a phone number, nothing is lost. The anonymous history joins the known profile, and the messages pick up where the site left off.
2. Segment by value and predict what comes next
RFM segmentation scores each customer on recency (how recently they bought), frequency (how often) and monetary value (how much in total). Together the scores show who your best customers are and who is slipping away. RFM describes the past, so a good CDP adds predictions on top: how likely each customer is to buy again, to leave, to respond to a discount, and what they are likely to be worth over time (predicted customer lifetime value).
Take an outdoor gear store. RFM finds a group of high-value customers who bought often in recent months. A churn prediction flags the ones among them who have gone quiet. Those get a retention message with a loyalty reward, not a blanket discount. A second group, likely to respond to ads, gets new products that match what they browsed. The same profile drives both decisions.
3. Join online and offline
Stores with shops hold half the picture at the till. Feeding in-store purchases into the same profile means the site stops recommending what the customer bought in the shop last week, and recommends what goes with it instead. The data flows the other way too: what someone saved online can shape what the shop staff or the shop's messages show them.
What to check before you choose one
Who gets a profile
Ask whether anonymous visitors get a profile from the first visit, and whether that history merges when they identify themselves. If only known customers count, the CDP ignores most of your traffic.
Whether it predicts or only describes
Segments by demographics, device or last purchase are a start. Predictions of churn, purchase likelihood, discount affinity and lifetime value tell you what to do next. Ask what the predictions are and how they are tested.
Whether it acts without exports
Count the channels it acts on directly: the site's search, recommendations and banners, email, push, SMS, messaging apps, and ad audiences. Each export to another tool is a copy that drifts and a delay you pay for in relevance.
How it chooses the channel and the moment
A journey builder that sends every message on a fixed schedule leaves money behind. Look for per-person send times, a choice of channel based on where each customer responds, and A/B tests across channels, flows and timing.
How you will know it worked
The honest answer to the return on a CDP is that it depends on what it does with the data. Ask for a holdout group: a share of shoppers who do not get the personalization, compared with those who do, in your own orders. That measures incrementality, not opens or clicks.
One profile that sells, on every channel
Every visitor gets a profile, and a better page
The agent profiles every visitor from the first click, including those who never log in, and uses it to rank the search, pick the products and answer the chat.
How it sells →RetentionMessages and audiences from the same memory
Email, push, SMS, Viber and WhatsApp, and the audiences sent to your ads, all from one profile per shopper, toward the goal you choose.
How retention works →Questions about customer data platforms
What is a customer data platform?
Software that collects customer data from separate sources, such as the site, the app, orders, email tools and the CRM, and joins it into one profile per person. Actionable CDPs then use that profile for personalization, messages and ad audiences.
What is the difference between a CDP and a CRM?
A CRM tracks conversations between your team and leads or customers: emails, tickets, calls and deals. A CDP holds every customer and visitor signal regardless of source, including behaviour, orders, events and surveys, so it sees the whole journey, not only the moments someone spoke to sales or support.
What is the difference between a CDP and a DMP?
A DMP stores mostly third-party, anonymised data such as cookies and device IDs to target paid ads. A CDP builds profiles of real customers from first-party data and uses them across the site, messages and ads.
Does a CDP work for anonymous visitors?
Only if it starts a profile on the first visit. Many CDPs profile only customers who gave an email or phone number, which leaves out most of a store's traffic. The agent profiles every visitor and merges the history when they identify themselves.
What return can a store expect from a CDP?
It depends on what the CDP does with the data, not on the data itself. The reliable way to know is a holdout group: compare shoppers who get the personalization with shoppers who do not, in your own orders.
See which of last week's visitors you missed
One call, with your store and ad accounts open. Then 30 days free. A holdout group decides: if the agent doesn't add orders in 30 days, you don't pay.
See it on your store
You'll talk to one of us.