What is a Customer Data Platform?

What a CDP is

A Customer Data Platform is a piece of software that collects data about your customers from every system and touchpoint in your business, then stitches it together into a single, persistent profile for each person. That profile updates in real time as new data comes in, and it's accessible to other tools in your stack – your marketing platform, your customer service software, your analytics tools – via integrations or APIs.

The key word is unified. Without a CDP, the same customer might appear in three different systems under three different identifiers: a customer ID in your e-commerce platform, an email address in your ESP and a reference number in your CRM. These records don't talk to each other. You can't tell, at a glance, that a customer who clicked your last promotional email also bought from you six months ago and raised a support ticket last week. A CDP resolves that problem.

It's worth being clear about what a CDP is not. It's not a CRM replacement, and it's not an analytics dashboard. It's data infrastructure – the layer that sits beneath your tools and makes them smarter by feeding them accurate, complete information about real people.

How it differs from a CRM and an ESP

The terminology in this space is genuinely confusing, so it's worth defining each category properly before drawing comparisons.

A CRM (Customer Relationship Management system) is designed to manage your relationship with known contacts – leads, customers and accounts. It stores contact details, interaction history, deal stages and notes. Your sales and account management teams live in it. CRMs like Salesforce, HubSpot and Pipedrive are built around manual data entry and structured around the needs of salespeople. They're excellent at what they do, but they don't automatically pull in behavioural data from your website, your app or your e-commerce platform.

An ESP (Email Service Provider) is the tool you use to send marketing emails. Mailchimp, Klaviyo and Dotdigital are ESPs. They hold subscriber lists, manage unsubscribes and track opens and clicks. More sophisticated ESPs can trigger automated sequences based on behaviour, but they typically only see what happens inside their own environment. They don't know what a contact did on your website after they clicked through, or what they bought, or whether they called your support line afterwards.

A DMP (Data Management Platform) is worth distinguishing here because it sounds similar to a CDP but serves a different purpose. DMPs aggregate anonymised audience data – largely from third-party cookies – for advertising targeting. They deal in segments, not individuals, and the data doesn't persist for long. As third-party cookies become less available, DMPs are declining in relevance for most businesses.

A CDP differs from all of these in two important ways. First, it ingests data from multiple sources automatically – behavioural, transactional and demographic – rather than relying on manual input or staying inside one tool's ecosystem. Second, it performs identity resolution: the process of matching data points from different systems to the same real person. If someone browses your site anonymously, then completes a purchase, then emails customer support, identity resolution connects those three events to a single profile. That unified customer profile – a complete, up-to-date record of a person's interactions with your business across every channel – is what a CDP produces and maintains.

What a CDP actually does

The core functions of a CDP break down into four areas.

Data collection. A CDP connects to your existing systems – your website, your CRM, your e-commerce platform, your point-of-sale system, your support tool – and pulls data from all of them continuously. This happens through pre-built integrations, API connections or tracking scripts. The CDP doesn't replace those systems; it reads from them.

Identity resolution. Once data arrives from multiple sources, the CDP works out which records belong to the same person. It does this by matching shared identifiers: email addresses, phone numbers, device IDs, login tokens. The more signals it has, the more accurate the matching. This is the technical work that transforms a pile of disconnected data into coherent customer profiles.

Segmentation. With complete profiles in place, you can build audience segments that would be impossible with fragmented data. Not just "customers who bought in the last 90 days" but "customers who bought in the last 90 days, haven't opened an email in 60 days, and have a lifetime spend above £500." The granularity is only useful if the underlying data is accurate – which is why identity resolution comes first.

Real-time activation. A CDP pushes those segments and profiles out to the tools that act on them. Your ESP receives an updated list before your next campaign runs. Your advertising platform gets a suppression list of recent purchasers. Your customer service tool shows an agent the full history of a caller before they say a word. This is where a CDP creates direct commercial value: the right message to the right person at the right time, with the data to back it up.

Signs you've outgrown your CRM or ESP

A CRM and an ESP are sufficient for a long time. The point at which they stop being sufficient tends to creep up gradually rather than announce itself clearly. These are the situations that typically indicate the threshold has been crossed.

Your marketing team builds audience segments by exporting spreadsheets from multiple systems and merging them manually before each campaign. This process takes hours, introduces errors and means your lists are already slightly out of date by the time you use them. If this is a regular part of how campaigns get built, the problem is structural rather than process-related.

You can't answer basic questions about customer behaviour across channels. You know your email open rates. You know your e-commerce conversion rate. But you can't tell which customers engage heavily with your emails but never convert, or which channel your highest-value customers came from originally. The data exists; it's just not connected.

You're running personalisation that isn't actually personalised. Your ESP lets you insert a first name and send different emails to different lists, but the segmentation logic relies on simple tags that were set months ago and haven't been updated since. Customers receive messages that don't reflect their actual recent behaviour – sometimes with jarring results, like a promotional offer sent to someone who complained last week.

Your customer view differs depending on which team you ask. Sales describes a customer one way; marketing has different data; support has no visibility of either. Decisions about how to treat that customer – whether to discount, escalate or retain – get made on incomplete information. A single source of truth would change that.

You're spending money on advertising to people who are already customers. Without a reliable, real-time suppression list fed from accurate customer data, paid campaigns regularly reach existing buyers – wasting budget and occasionally irritating the people you most want to keep.

What to look for when evaluating a CDP

The CDP market includes dozens of vendors at very different price points and capability levels. Evaluating them on marketing materials alone is difficult. These are the criteria that actually matter.

Integration coverage. A CDP is only as useful as the systems it can connect to. Before evaluating any platform, map out every system in your current stack that holds customer data. Then confirm – with specifics, not assurances – that the CDP has working integrations with each one. "We can integrate with anything via API" is a common answer that often means "your development team will need to build it."

Identity resolution quality. Ask vendors specifically how they handle identity resolution when data arrives without a clear shared identifier. What happens when someone browses on mobile without logging in, then converts on desktop? How does the system handle duplicate profiles? Poor identity resolution means your unified profiles aren't actually unified – which undermines everything built on top of them.

Real-time capability. Some CDPs process data in batches – hourly or daily. Others update profiles continuously. Which you need depends on your use cases. If you're sending triggered emails based on abandonment behaviour, batch processing may be fast enough. If you're personalising a website experience in the same session that a customer updates their preferences, you need genuine real-time throughput.

Cost model. CDP pricing typically scales with data volume – the number of profiles stored or the number of events processed per month. Understand what that model looks like at your current scale and at the scale you expect in two years. Costs that look reasonable at 50,000 customers can become significant at 500,000.

Organisational ownership. This is less about the technology and more about your business. A CDP needs someone accountable for it – configuring integrations, maintaining data quality, building and updating segments. If no one in your organisation owns that function today, the platform won't run itself.

Common pitfalls

Garbage in, garbage out. A CDP aggregates and unifies your existing data. If that data is inconsistent, duplicated or poorly structured – which it usually is, to varying degrees – those problems don't disappear inside the CDP. They get consolidated. Businesses that skip a data audit before implementing a CDP often find they've spent significant budget building a very efficient system for delivering inaccurate information to their tools. A data warehousing or data quality project before or alongside CDP implementation is usually time well spent.

Integration complexity is routinely underestimated. The vendor demo will show a clean, pre-configured stack. Your actual stack will have custom fields, legacy data structures, systems that weren't built with open APIs and internal tools nobody documented. Connecting everything takes longer and costs more than the initial estimate in almost every implementation. Build contingency into the timeline and budget, and get integration scope confirmed in writing before signing.

Organisational readiness matters as much as the technology. A CDP creates a single view of the customer. But that view is only valuable if marketing, sales and customer service actually use it – and align their processes around it. Businesses that implement a CDP without addressing how teams will work differently often find adoption stalls within a few months. The technology is the easier part of the problem.