learn
Do You Need a Customer Data Platform (CDP)? A 12-Question Checklist
By Art Kukharenko · Published · Updated
Browse learn resourcesYou probably need a customer data platform (CDP) if your customer data sits in three or more systems, you cannot reliably recognise the same person across them, you need that unified view in minutes for marketing, service or compliance, and you have people to run it. If one CRM or engagement platform already does the job, you may not. Score yourself with the 12 questions below.
Who this is for: heads of marketing, CRM, data and IT deciding whether to buy, build or wait. You will get a scored checklist, the alternatives, common wrong reasons to buy, readiness prerequisites and next steps. More explainers: the Learn hub.
What a customer data platform does, and does not do
The CDP Institute defines a customer data platform as software that creates and maintains a persistent, unified customer record that other systems can access 1. It groups vendors into data, analytics, campaign and delivery CDPs, each adding functions to the one before 1. It is not a CRM (relationship records), a data warehouse (analytical storage), a DMP (anonymous advertising audiences) or master data management (golden records); see CDP vs CRM vs DMP vs MDM vs data warehouse. New to the category? Start with what a customer data platform is.
The 12-question checklist
Score each question 0 (no), 1 (partly) or 2 (yes). Questions 1 to 8 measure need (maximum 16); questions 9 to 12 measure readiness (maximum 8).
Scroll horizontally to compare all columns ↔
| # | Question | Why it matters |
|---|---|---|
| 1 | Is customer data spread across three or more systems (web, app, CRM, billing, support, branch or call centre) with no shared record? | Unifying records is the core job of the category 1 |
| 2 | Does the same person appear under different identifiers (email, device, account, loyalty ID) that you cannot reliably join? | Identity resolution is the hardest, most valuable part (explained) |
| 3 | Do marketing or service teams wait days for audience or data pulls? | Self-service audiences are the usual first win |
| 4 | Must you act within minutes or seconds of behaviour (cart, failed login, balance event)? | Batch exports cannot; real-time raises the architecture bar |
| 5 | Do you send the same audience to many tools (ads, email, SMS, push, support)? | Point-to-point integrations multiply |
| 6 | Do consent and preferences differ across systems, so you cannot prove what was sent under which consent? | One consent record and audit trail reduce risk |
| 7 | Do law, contract or policy limit where customer data may live or who may process it? | This shapes deployment more than features do |
| 8 | Is segmentation limited by what your current tool can see (for example, no billing or product-usage data)? | The gap is data access, not features |
| 9 | Do you have stable identity keys (login or customer ID, verified email or phone)? | Without keys, unification is guesswork |
| 10 | Do you have named owners for data engineering, marketing operations and privacy? | Platforms fail from missing owners, not missing features |
| 11 | Can you name two or three use cases with measurable outcomes in the next 12 months? | Use cases justify cost and scope the proof of concept |
| 12 | Have you tried to meet those use cases with your current stack, and can you show where it falls short? | Proves the need rather than assuming it |
What your score means
These bands are our rule of thumb, not an industry standard.
Scroll horizontally to compare all columns ↔
| Need score | Readiness 0 to 3 | Readiness 4 to 6 | Readiness 7 to 8 |
|---|---|---|---|
| 0 to 5 | Fix data basics; do not buy | Do not buy; revisit yearly | Optimise the current stack |
| 6 to 10 | Fix prerequisites; try alternatives | Try alternatives first; re-score in six months | Try alternatives; run a small proof of concept |
| 11 to 16 | Fix prerequisites first, then shortlist | Shortlist and run a proof of concept | Shortlist, proof of concept and RFP |
The alternatives
Scroll horizontally to compare all columns ↔
| Option | Good for | Limits | Evidence |
|---|---|---|---|
| Warehouse plus reverse ETL (composable) | Teams with a warehouse and data engineers | Engineering effort; real-time access is weak 2 | Hightouch says data activation platforms are powered by reverse ETL 6; Fivetran sells Activations 7 |
| CRM extension | Organisations centred on one CRM | Credit or profile pricing; scope follows the CRM | Salesforce Data 360: Profiles $240 per 1,000 per year; Flex Credits $500 per 100,000 4 |
| Engagement platform with built-in profiles | One platform covers most channels | Profile model centred on messaging; other systems' data must be sent in | Braze profiles can be anonymous or identified and merge by documented rules 8; Bloomreach stores consent in the profile 9 |
| Open-source or self-hosted CDP | Control and customisation | You operate it | Apache Unomi is Apache 2.0 10; see open-source options compared |
| Do nothing, fix the basics | Low need or low readiness | Problems persist | Identity keys, consent capture and a few well-owned SQL views |
Packaged platforms are faster to deploy and easier for business users, while composable approaches reuse your warehouse but need more engineering 2; the full trade-off is in composable vs packaged vs private.
Common wrong reasons to buy
Readiness prerequisites
- Data quality: you have profiled duplicates, nulls and stale records in your main sources.
- Consent capture: consent is recorded with purpose, time and source, and withdrawals reach every system.
- Identity keys: at least one persistent key exists in each major source.
- Ownership: someone owns schema, identity rules, quality checks and the budget.
- Use cases: each has a metric, a baseline and a decision date.
- Constraints: residency, contract and security requirements are written down before vendor demos.
Cost reality check
Public prices show how models differ. Twilio Segment's Free plan covers 1,000 monthly tracked users, and Team starts at $120 per month for 10,000; both cap reverse ETL records and syncs per day 3. Salesforce lists Profiles at $240 per 1,000 per year 4; at one million profiles that is $240,000 a year for that line alone at list price, before credits, other clouds or services (our arithmetic; contracts vary). Stable Kernel estimates a CDP programme at $300,000 to $2.5 million over three years (a secondary estimate) 5. Flat-licence models such as PrivyCDP remove per-profile exposure but add infrastructure and DevOps you fund. See CDP pricing explained and the 3-year TCO model.
What to do next
- Low scores: fix identity keys, consent capture and ownership; revisit in six months.
- Middle scores: test a warehouse plus reverse ETL, or the profile features of tools you own.
- High scores: write requirements with the RFP template and test with a proof of concept on synthetic data.
Frequently asked questions
Do I need a CDP?
You need one if customer data is fragmented across systems, identities cannot be joined, teams need unified audiences or triggers faster than batch exports allow, and you have owners to run it. If a single CRM or engagement platform covers your use cases, you may not. Use the checklist above and re-score as your stack changes.
When do you need a customer data platform?
Typically when point-to-point integrations multiply, consent and preferences diverge between systems, and marketing or service teams wait on engineering for data. Regulation can also force the question when you must prove what was sent under which consent, or control where data lives. Need without readiness means fixing prerequisites first.
CDP vs data warehouse: do I need both?
Often, in some form. A warehouse stores and analyses data; a customer data platform adds identity resolution, audiences and activation. Composable approaches build those functions on the warehouse, so you may need one platform plus tooling rather than two products 2. Decide by real-time needs, engineering capacity and where data must live.
What are the signs you need a CDP?
Common signs are customer records in three or more systems, duplicate people under different IDs, slow audience requests, inconsistent consent across tools, many activation destinations and a need to react within minutes. Two or three of these together, plus the readiness items, justify a shortlist.
What are CDP alternatives?
The main alternatives are a warehouse with reverse ETL, a CRM extension such as Salesforce Data 360, an engagement platform with built-in profiles such as Braze or Bloomreach, an open-source platform such as Apache Unomi, or fixing data basics first 48910. Each trades speed, control and engineering effort differently.
Can a small business use a CDP?
Yes, at small scale. Twilio Segment's Free plan covers 1,000 monthly tracked users and Team starts at $120 per month for 10,000 3. At small volumes, the engagement platform you already use often has enough profile features, so test that first. Buy when fragmentation, identity or consent pain is real.
How do I run a CDP readiness assessment?
Score the 12 questions, then check the readiness list: data quality, consent capture, identity keys, ownership, use cases and constraints. Profile a sample of records for duplicates and missing keys, interview the teams that wait on data, and write down residency and security limits before talking to vendors.
Sources
- CDP Institute, What is a CDP? Open source
- CDP Institute, Composable CDP Knowledge Hub Open source
- Twilio Segment pricing Open source
- Salesforce Data 360 pricing Open source
- Stable Kernel, CDP total cost of ownership (secondary) Open source
- Hightouch, Composable CDP Open source
- Fivetran Activations Open source
- Braze, User profile lifecycle Open source
- Bloomreach, Consent management Open source
- Apache Unomi Open source
- Google Privacy Sandbox, 17 October 2025 Open source
Put this evaluation into practice.
Review the total cost modelFurther reading
- Learn about CDP
- CDP Pricing Explained
- CDP Proof of Concept With Synthetic Data
- CDP RFP Template: 50 Requirements
- CDP Total Cost of Ownership: 3-Year
Published: · Updated:
Publication date recorded for the approved release. Editorial updates do not establish legal approval or a customer result.