CDP · CRM · 8 MIN

The CDP Reckoning: Your Next CDP Is a Data Platform You Already Pay For

SEP 02, 2026

TC Tim Christensen

The customer data platform had a good run as a category. The pitch was simple: your customer data is scattered across dozens of systems, so buy a platform that ingests all of it, resolves it into one profile per person, and pushes audiences back out to every tool you run. The premise was real. The buying decision built on it is what needs a second look, because the two jobs that justify a CDP's premium, identity resolution and audience building, now run on infrastructure a lot of companies already own and already pay for.

This is not a forecast. It is visible in pricing pages and acquisition announcements from the last two years, and it points one direction: the CDP layer is moving from product to feature.

What a CDP actually costs

Verified from vendor pages as of this writing, because CDP pricing is where transparency goes to die:

Twilio Segment publishes Connections pricing: $0 for 1,000 visitors a month, Team from $120/month for 10,000 monthly visitors, then $10 per extra thousand. That is the pipeline product. The actual CDP, Connections plus Unify identity resolution plus Engage activation, is contact-sales only.
Tealium publishes exactly one number: $1,000/month for Data Cloud Activation, billed annually, which is reverse-ETL that runs against your warehouse. The tiers above it, the ones with real-time profiles and identity resolution, are quote-based.
mParticle, now part of Rokt, prices through usage "credits" tied to event volume, profile counts, and feature usage. Also quote-based.

Third-party analyses fill in the ranges the vendors won't print. CDP.com's pricing guide from August 2026 puts small and midsize deployments at $1,000 to $10,000 per month and enterprise deals at $200,000 to $500,000+ per year, with implementation fees stacked on top.

Then there is the billing unit itself. Segment bills on monthly tracked users, and unassociated anonymous IDs count toward that total. If you run a high-traffic consumer site, part of your bill pays for tracking people you cannot identify yet.

You are buying compute, not data

Strip the marketing off a CDP and it sells two jobs. Job one: identity resolution, matching the cookie to the logged-in account to the email address to the mobile device and merging them into one profile. Job two: audience building, rules over those profiles, pushed out to ad platforms, email tools, CRMs. The audience half overlaps with what a DMP did for cookie-based targeting, except CDPs attach the audiences to known customer identities instead of anonymous segments.

Neither job gives you the data. You send the vendor your events, they process them, they hand back segments. The thing you rent is compute and matching logic applied to first-party data that was already yours. If your warehouse already holds that data, the compute is the only part you still need to buy, and warehouse compute is a line item you already have.

The vendors are already retreating to your warehouse

The strongest evidence that the center of gravity moved is that the CDP vendors are re-architecting around their customers' warehouses instead of their own storage.

Segment shipped Linked Audiences in 2024, a mode that queries data directly in Snowflake, BigQuery, Redshift, and Databricks without copying it into Segment. Tealium sells a Snowflake Audience Discovery app and frames its whole $1,000/month tier as warehouse-native activation: reverse-ETL, context API, unlimited cloud data sources. Read those two product pages next to each other and you can hear the category negotiating its own surrender. The vendors that used to say "bring your data to us" now say "we will meet your data where it lives."

From the other direction, the data-movement companies are coming for the same budget. Fivetran bought Census, the reverse-ETL pioneer, in May 2025, and Census (now Fivetran Activations) joined Fivetran's consumption pricing in February 2026. We covered the broader version of this move last month: Fivetran showing a health system getting AI-ready data without rebuilding its platform. The CDP budget is the next line item in that same argument.

What consolidation looks like from the losing side

Follow the money and it gets less comfortable. Twilio bought Segment for $3.2 billion in November 2020. In Q4 2023 it recorded a $285.7 million impairment on Segment-related intangible assets, which is accounting language for "this asset will not earn what we paid." mParticle did not get a standalone exit at all: Rokt bought it for a reported $300 million in January 2025 and now markets it as Rokt mParticle, a data layer inside an ecommerce personalization company.

None of this means the products are bad. Segment is still a Leader in IDC's CDP MarketScape, and Tealium's governance story keeps winning in healthcare and financial services. It means the standalone CDP, priced as its own enterprise budget line, is getting squeezed from both ends: warehouse-native activation is cheaper for teams with a working data stack, and suite vendors are absorbing the CDP feature set for everyone else.

The math to run before you sign

The honest version of "do we need a CDP" comes down to two counts.

Count your data sources. If the list is your website analytics, your email platform, your CRM, and your payment processor, that is four systems, and a competent analyst can union them in a week. CDPs earn their fee when the source count gets high enough that maintaining hand-built pipelines becomes a full-time job with an on-call rotation.

Count your identity rules. If customers log in before they buy, and one stable key like email or customer ID follows them through every system, identity resolution is a join. SQL does joins. CDPs earn their fee when identity is genuinely messy: anonymous traffic you need to convert, journeys that span devices and channels, online and offline stores, or household-level matching instead of individual-level.

Here is the part most vendor demos skip: a small stack with clean keys may not need identity resolution at all. If every record already shares a common key, the product you are being sold solves a problem you do not have.

The standalone CDP still wins where both counts run high: consumer businesses with large anonymous traffic, real-time profile needs measured in milliseconds rather than minutes, and regulated industries where Tealium has built its entire franchise on governance. The question is whether your counts are high, not whether the sales deck assumes they are.

The verdict: count before you contract. The two jobs a CDP charges premium for, identity resolution and audience activation, increasingly run on warehouse infrastructure companies already pay for, and the vendors' own roadmaps concede the point. The standalone CDP still earns its budget where identity is genuinely hard and source counts are high. Everywhere else, it is a premium layer over infrastructure you own. Run the two counts. If either is high, negotiate hard, because the alternative is credible now. If both are low, the money belongs in your data stack, not in a renewal.

What to do this week

Inventory your sources and your keys. List every system that holds customer data and the identifier it uses. If you can draw the join path between any two systems in one line, your identity problem is smaller than a CDP quote assumes.

Price warehouse-native activation for your volumes. If you already run Snowflake, BigQuery, or Redshift, get a consumption estimate for pushing audiences out of the warehouse: Fivetran Activations, a standalone reverse-ETL tool, or Tealium's $1,000/month tier. Put that number next to the CDP quote on the same page.

Ask every CDP vendor the warehouse question. In each demo, ask what runs in their system versus what can run in yours. Segment's Linked Audiences and Tealium's Snowflake app exist because buyers started asking. If the answer is that everything must flow through the vendor's platform, you are renting compute at a markup.

Your next CDP may be a set of capabilities you already pay for: a warehouse, a sync layer, and a few dbt models that define what a customer is. The reckoning is just the market catching up to that.

Compare CDPs against warehouse-native options before renewal

Our directory breaks down customer data platforms and activation tools by pricing model, identity features, and what they actually need to run. Check the renewal quote against the lean option.

BROWSE THE TOOL DIRECTORY →

Sources: Twilio Segment Connections pricing (retrieved Sep 2, 2026) · Tealium pricing (retrieved Sep 2, 2026) · CDP.com: CDP Pricing, Models, Ranges, and Hidden Costs (Aug 31, 2026) · CDP.com: What Is Twilio Segment? (Aug 13, 2026) · Twilio Segment billing docs: MTUs and throughput · Twilio Q4 2023 earnings release (Feb 14, 2024) · Fivetran: agreement to acquire Census (May 1, 2025) · Fivetran: Census joins consumption-based pricing (Feb 2, 2026) · Rokt and mParticle merger announcement (Jan 16, 2025)