Choosing a customer data platform is one of the highest-stakes technology decisions a marketing or data team makes. The platform shapes your data architecture, sets what you can activate and fixes your integration options for years. Yet many evaluations still run on vendor demos, analyst quadrants and internal politics rather than a structured, weighted comparison.
This framework is the scoring method CdpHub uses in client engagements, published as an open resource. This edition grounds the first dimension, integration coverage, in the CdpHub directory: 6,699 integrations documented by 44 platforms, each with its direction, how it runs where the page says, and a link to the vendor's page. It also adds a worked example and the viability signals the directory picked up during 2026.
The seven dimensions
| Dimension | Default weight | What to check | Best evidence |
|---|---|---|---|
| 1. Integration coverage | 20% | Every tool in your stack, in the direction you need, and how each integration runs | The CdpHub directory, then the vendor's docs |
| 2. Identity resolution | 15% | Deterministic and probabilistic matching, cross-device stitching, how transparent and reversible merges are | Vendor docs, then a proof of concept on your own data |
| 3. Real-time capability | 15% | Event latency, real-time audience qualification, streaming destinations | Vendor docs, then a load test |
| 4. Data governance and privacy | 15% | Consent integration, data residency, schema enforcement, how much runs in the browser | Vendor docs, security review, the directory's run types |
| 5. Ease of implementation | 10% | Time to value, documentation quality, templates, migration tooling | References and a pilot |
| 6. Total cost of ownership | 15% | Licensing model, services, operating overhead, cost at 2x, 5x and 10x volume | Quotes and your own cost model |
| 7. Vendor viability | 10% | Financial stability, roadmap, ownership changes | Filings, news and the vendor's documentation |
Default Weight Distribution
The weights are a starting point. A retailer with fragmented customer records should push identity resolution to 25 percent or more. A media company targeting ads in real time should weight real-time capability higher. Adjust the weights before anyone scores a vendor, so the weights can't be tuned to favor one.
1. Integration coverage, measured
Integration breadth is the dimension vendors most like to quote and buyers can most easily check. A headline count says little: of the directory's 2,365 tools, 70 percent are documented by only one platform. What matters is whether a platform documents your tools, in your direction.
How to score it
- List your stack. Write down every tool that must send data to the CDP or receive data from it, and which way the data moves.
- Look each tool up for each finalist. In the CdpHub directory, check that the integration is documented, that it runs in the direction you need, and how it runs: a client-side tag, a server-side connection or through a warehouse.
- Score from 1 to 5. 5: every tool documented in the needed direction, with a server-side or warehouse path. 4: every tool documented, a few only through browser tags. 3: one gap you can bridge, for example through your warehouse or a webhook. 2: gaps in tools you can't replace. 1: several core gaps.
A worked example. Take a ten-tool retail stack. Shopify sends orders and customers in. Meta, Google Ads and TikTok receive audiences and conversions, Klaviyo and Attentive receive email and SMS audiences, and Google Analytics receives events. Snowflake, Zendesk and Salesforce can connect either way.
| Platform | Tools documented in the needed direction | What's missing |
|---|---|---|
| Amperity, BlueConic, Lytics, mParticle, RudderStack, Segment | 10 of 10 | Nothing, and every one has a server-side or warehouse path |
| Hightouch, Fivetran Activations | 9 of 10 | Shopify appears only as a destination in their catalogs, so plan how Shopify data reaches your warehouse |
| DinMo | 9 of 10 | No Google Analytics integration |
| ActionIQ | 9 of 10 | Google Analytics only as a source, and no run type documented for eight of the nine |
| Tealium | 8 of 10 | No Attentive or Zendesk integration in Tealium's documentation |
Six of the 44 platforms document all ten. That is a short list built from documentation in minutes, before any demo. It is not a ranking: a tool missing from one vendor's documentation may still connect through a webhook, the warehouse or the partner's own integration, so ask each finalist about its gaps and get the answer in writing.
What the directory adds beyond coverage
- Direction. A catalog heavy on inbound connectors is built to assemble the profile, and one heavy on outbound connectors is built to activate it. Salesforce Data 360 documents 156 inbound integrations, 11 outbound and 10 both ways. Tealium documents 49 inbound, 832 outbound and 5 both ways.
- How integrations run. The share of a catalog that depends on code in the browser bears on dimension 4. It ranges from 67 percent of Tealium's listings to none of Fivetran Activations' or Salesforce Data 360's.
- Warehouse depth. Warehouse share runs from 95 percent or more for the composable platforms (Hightouch, Fivetran Activations, DinMo) to 3 percent or less for Tealium, Segment, mParticle and Klaviyo, and none at all for four smaller platforms. If the warehouse is your system of record, it belongs in the score.
7. Vendor viability: watch the documentation
Ownership changes, renames and shutdowns show up in the documentation, often early. While CdpHub read the 44 catalogs during 2026, it recorded these:
- ActionIQ now documents under Uniphore, and BlueVenn under Upland.
- Blueshift's documentation says it is part of BlueConic.
- Census joined Fivetran and is now Fivetran Activations.
- Totango became Odie, and CrossEngage became Spotler ActivatePro.
- SAP renamed SAP Emarsys to SAP Engagement Cloud in February 2026.
- Delighted shut down on June 30, 2026, and Salesloft began phasing out Drift in March 2026.
None of these is a reason to rule a vendor out. Each is a reason to ask about the roadmap, contract terms and support commitments.
Signals worth checking for each finalist:
- Who owns the documentation domain?
- Do catalog pages redirect to a new owner?
- Are integration pages still being updated?
- Do partners still list the product?
The last test matters for your own stack too: a tool that is being retired can take an integration with it.
Scoring methodology
Score each sub-criterion from 1 to 5: 1 does not meet requirements, 3 meets them, and 5 exceeds them with differentiated value. Average the sub-criteria into a dimension score, divide it by 5 and multiply by the dimension's weight. The seven results add up to a total out of 100. For example, a 4 on integration coverage at 20 percent contributes 16 points.
Have at least three people score each platform independently, then average the scores and discuss the outliers. The goal is not one winning number. It is to see where platforms are strong, where they are weak, and where your own team disagrees about what matters.
How to use this framework
- Define your requirements first. Document your use cases, data sources, activation destinations and technical constraints before scoring any vendor.
- Set your weights. Adjust the seven weights to your priorities before the first demo.
- Shortlist with data. Use the CDP Market Landscape 2026 report and the directory to narrow the field to three to five finalists, and apply the framework only to them.
- Score independently, then discuss together. Independent scoring prevents groupthink, and the discussion surfaces the disagreements that matter.
- Check every claim against the documentation. Verify integration claims in the CdpHub directory, which links each listing to the vendor page it came from, rather than relying on sales materials.
How we count
Integration figures are every integration with a third-party tool that each of 44 platforms documents on a public page, as of October 2026. The platform's own SDKs, libraries, APIs and features are not counted. Counts change as catalogs change; the live directory is the reference.
