Where the data came from, and whether the accuracy claim means anything

Each carries a different obligation for the buyer, and a provider that will not name the mix is asking you to accept all of them.

Contribution pricing is often the real price. The discount is paid in your own contact records, and that is a decision your privacy counsel should make rather than your procurement team.

Continuous and real time are not measurable claims. A monthly re verification share is, and a provider that tracks it can state it.

An accuracy rate on a vendor chosen sample tells you about the sample. Ask for the denominator before you ask for the percentage.

38 of 249 vendors disclose the model behind a score. A score whose inputs you cannot name cannot be diagnosed when it stops working.

This costs the buyer an afternoon and settles more than a procurement cycle of reference calls. Reluctance is itself an answer.

Where do B2B contact data providers source their data and is it licensed?

Most do not say. Of the 95 vendors in the index that sell or resell B2B contact data, 17 document their sourcing and licensing at a standard a buyer can check, 18 percent, and 16 carry a D, meaning the question is not addressed at all. The 4 graded at the top of the scale are Apollo.io, GTM Workspace by ZoomInfo, Instantly, Saleshandy. Where sourcing is disclosed, the common structure is a mix of public web collection, user contributed data from a community or extension, licensed third party files, and partner contribution networks where customers exchange their own contact records for access to everyone else's. Those four carry very different obligations for the buyer, which is why the index grades the disclosure rather than the sourcing method: a provider that names its inputs can be assessed, and one that does not is transferring that assessment to you.

How do you evaluate the accuracy of a B2B contact data provider?

Never on the headline percentage, because almost every provider publishes one and 67 of 95 publish it with no method attached. 28 of 95 state a measurement basis a buyer could reproduce, 29 percent, and 23 carry a D. Three things make a claim assessable. First, the denominator: accuracy measured against records the provider chose to sample is not accuracy across the file you will actually query. Second, the date and the method of verification, since a bounce tested address and a pattern generated one can both be called verified. Third, the segment, because coverage in North American mid market technology tells you nothing about European manufacturing. The practical test costs nothing: send the provider a list of 200 accounts you already know well, ask for their records against it, and measure the result yourself. A provider confident in its file will run that test. Treat reluctance as the answer.

How often do B2B contact data providers refresh their records?

A refresh cadence is a claim like any other, and the index grades whether the claim carries a method rather than whether the number is impressive. Continuous, real time and always on are the phrases used most often across the 95 vendors in this lane, and they are not verifiable statements. 28 of them publish an evidence standard behind their operational claims, so the great majority of published refresh claims have nothing behind them. The questions that turn the claim into something checkable are these: what triggers a record to be re verified, is it a fixed calendar cycle or an observed signal such as a job change; what share of the file is re verified in a given month rather than merely re scored; how long a departed contact stays live before removal; and whether the same cadence applies to email addresses, phone numbers and firmographics, which usually decay at very different speeds. Ask for the monthly re verification share as a number. Vendors that measure it will tell you, and vendors that do not will describe the process instead.

How do B2B intent data providers differ in how they collect intent signals?

They differ more than the category's marketing suggests, and the difference is mostly undisclosed. Of the 28 intent providers in the index, 6 document where their signals come from at a checkable standard, 21 percent, and 4 carry a D. The ones that do document it are GTM Workspace by ZoomInfo, Clevenio, Lead411, Neuron, Sendr, What's Trending. The collection methods behind the category are genuinely different products: bidstream data observed through advertising exchanges, publisher cooperatives where a network of trade sites shares reader behaviour, first party signals from your own site and product, review site activity, and inferred signals derived from hiring, funding or technology change. Bidstream and cooperative data describe an anonymous company; first party data describes a known person. The model that turns any of these into a score is disclosed by 4 of 28. Not one vendor in this lane documents all three of provenance, evidence and the model behind the score. Ask which of those five inputs is in the feed you are buying, in what proportion, and what happens to the score when one goes away.

Which revenue intelligence platforms have evidence their forecasts are accurate?

23 of the 33 revenue intelligence platforms in the index publish evidence behind their accuracy and outcome claims at a standard a buyer can check, 70 percent, which is the strongest showing of any data adjacent category on this page. 6 are graded at the top of the scale: Clari, DealHub, Gong, GTM Workspace by ZoomInfo, HubSpot Sales Hub, Scratchpad. That is a materially better record than contact data or intent, and the likely reason is that forecast accuracy is measured against an outcome that arrives on its own, the quarter closes whether or not the vendor wants it to, whereas contact accuracy is measured against a file only the vendor holds. Two cautions before treating a published figure as settled. Forecast accuracy is usually reported as a variance against the vendor's own submitted forecast rather than against the naive baseline of last quarter plus growth, and a model that beats a rep's gut but not the naive baseline has not earned its licence. And 10 of 33 disclose what the model is, so the number is often published without the method that produced it. 6 platforms document provenance, evidence and the model together: Adobe Marketo Engage, Demodesk, Gong, Revenue.io, SAP Sales Cloud, Scratchpad.

How does the GTM Tech Index grade data provenance and evidence?

On three axes, applied identically to all 249 vendors: Data Licensing and Provenance for where the data came from and on what licence, Operational and Outcome Evidence for whether a performance or accuracy claim carries a method, and AI Disclosure and Model Transparency for what produced a score. A grade records what the vendor has published, so a low grade means the public record is thin rather than that the practice is bad, and a vendor may well answer in a contract what it does not answer on its site. Across the index, 12 of 249 vendors document all three and 105 document none. AI Disclosure and Model Transparency is the thinnest of the three at 15 percent with 56 vendors at a D or below. Grades are never aggregated into a composite score, and every grade carries a source basis and a verification date on the vendor profile.