The Roles You Need for a CDP Program to Work

A customer data platform goes live with a project team of eight. Six months later it’s run by one CRM manager who also owns the newsletter, the loyalty program, and the Black Friday calendar. Nobody decided that. The project ended, the vendor’s onboarding hours ran out, and the platform became whoever was left in the room.

That is the staffing pattern behind most of the month-four failures we get called into. The platform works. The roles that keep it working were never named, so they were never filled.

This piece names them. Six roles, what each one does every week, roughly how much time it takes, and which seats one person can hold at once without the program stalling.

Why a CDP needs more roles than an email tool

An email tool has one input (a list) and one output (a send). A CDP has many of each. Events arrive from the website, the app, the shop backend, the call center, the loyalty system. Profiles merge and split. Consent changes. Segments feed email, SMS, push, paid audiences, and the website itself. A change at any input changes every output.

That means the work isn’t campaign work. It’s plumbing, governance, and quality control, with campaigns on top. Gartner found in February 2025 that 63% of organizations lack, or can’t confirm they have, AI-ready data. The reason is rarely the platform. It’s that nobody holds the seats below.

The six roles

Fig 1 · Six seats, one platform · The six roles a CDP needs and the time each takes

The names are ours. Your org chart will call them something else. Each responsibility needs a named person and time in the calendar.

1. Platform owner

The person who is accountable for the platform being legible. Not the most senior marketer. The one who knows why every event exists, which segments are live, what changed last week, and what will break if a field is renamed.

Weekly work: checks new tracking requests against the tracking plan and approves or rejects new events and attributes. Keeps the naming convention, owns the change log, and runs the monthly data-quality check.

Time: in our experience, 25 to 40% of one senior person’s week. Below that, the role exists on paper and the platform drifts.

If nobody holds this role: three consultants, three naming styles, twelve events that mean “purchase.” We wrote about the fix in naming conventions that survive the third consultant. The cause is always the same empty seat.

2. Data engineer, or the person who plays one

Someone who can read the source systems and make data arrive in the shape the platform expects. Builds and maintains the imports, the API integrations, the webhook endpoints, the warehouse export. Fixes the feed when the shop changes a field name.

Weekly work: monitors imports and event volumes, investigates gaps, ships the small changes the platform owner approved.

Time: half a day a week in steady state; several days a week during any integration. This is the seat marketing teams most often lack, and the one that most often sits with an agency or an internal IT team that has other priorities.

If nobody holds this role: the catalog feed stops on a Friday and nobody sees it. The back-in-stock scenario sends nothing for eleven days. The first person to notice is a customer.

3. Campaign and journey builder

The role most teams do have. Builds scenarios, templates, weblayers, and segments. Runs A/B tests. Reads the results.

Weekly work: the campaign calendar. Also the upkeep nobody budgets for: retiring dead scenarios, updating templates when the brand changes, and checking that each segment still means what its name says.

Time: full-time in most mid-market teams. The mistake is letting this person also be roles 1, 2, and 4. They’ll do the campaigns, because campaigns have deadlines, and the rest slips.

4. QA and release

Someone other than the builder checks the work before it goes live. Rendering across clients and devices, personalization fallbacks, links, consent conditions, the audience count against expectation, frequency caps.

Weekly work: a checklist per release. For Bloomreach Marketing (Engagement) we run a platform-specific list because the platform has behaviors that standard Jinja and JavaScript don’t.

Time: an hour or two per campaign. Small until it’s skipped, at which point a Jinja null in a subject line reaches the whole list.

This seat can be shared, on a rotation, between two builders who check each other. It can’t be the builder checking their own work.

5. Data protection and consent

Owns the consent model: which purposes exist, how each one is captured, how expiry works, how every channel checks its own purpose before sending. Answers the data subject request. Talks to the DPO in the DPO’s language.

Weekly work: little, if the model was built right. Monthly: the consent report, the DSR log, the check that no scenario sends on a purpose it doesn’t hold.

Time: a few hours a month in steady state, days during a build. Often a legal or compliance person part-time, but the person must understand where consent physically lives on the profile, not only what the policy says. GDPR fines have passed €6 billion cumulative in EU records. The ones marketing teams cause come from the gap between the policy and the consent data on the profile.

6. Analyst and measurement owner

Decides what a campaign has to prove and whether it did. Sets up holdout groups, reads incrementality, keeps the dashboards honest, and tells the campaign builder when an open rate is a diagnostic rather than a result.

Weekly work: the performance review; the monthly work is the harder one, reconciling platform numbers with the warehouse and finance.

Time: a day a week for a program with more than a handful of live scenarios. Without this seat, the program reports activity and nobody can say what it earned.

Who can share a seat

Six roles doesn’t mean six people. It means six responsibilities with a name on each. Combinations that work:

• Platform owner + analyst. Both need the whole picture and both are about legibility. One senior person, if campaigns are someone else’s job.

• Campaign builder + QA, on rotation. Two builders checking each other. Never one builder checking themselves.

• Data engineer as a shared internal service or a retainer. The role is bursty. You need a named person who answers within a day, not a ticket queue with a three-week SLA.

• Consent as a part-time seat for legal or compliance, with the platform owner as the technical counterpart.

Combinations that fail:

• Campaign builder + platform owner. Deadlines win. Governance loses. Every time.

• Platform owner in IT with no marketing hours. The platform stays clean and nobody uses it.

• Everything with the vendor’s customer success manager. Their job is adoption of features, not the health of your data. Good CSMs will tell you this themselves.

What this looks like in a team of three

Fig 2 · Six seats, three people · How a team of three and a half covers all six roles

Most of the teams we work with are one to three people on the CRM side. Here is a version that holds:

Person A, senior: platform owner and analyst. Two days a week on the platform, the rest on strategy.

Person B: campaign builder. Full-time. QA-checked by Person A for anything above a set audience size.

Person C, part-time or external: data engineer. Named, reachable, with a monthly review slot.

Consent: legal, part-time, with Person A as the technical counterpart.

That’s three and a half people covering six seats, and it works because each seat has a name and a slot in the calendar. The version that doesn’t work is the same three people with “the CDP” listed as a shared responsibility. Shared means nobody’s.

The question to ask before the next project

Before the next integration, the next channel, the next AI feature: who holds each of the six seats today, by name, and how many hours a week? If two of the six have no name, the next project adds load to a structure that is already short. Fill the seats first. The platform will wait; the drift won’t.

Read next

• Naming conventions that survive the third consultant. The platform owner’s first job, written down.

• The first conversation with engineers before a CDP migration. What the data-engineering seat needs from marketing in week one.

• Identity resolution: deterministic, probabilistic, and where each breaks. Six decisions the platform owner writes down.