How to Test a Data Vendor's Coverage Claims Yourself

Testing a data vendor's coverage claims yourself means replacing the marketing number with a sample you can personally verify. Pick a small set of companies or people whose real details you already know, query the vendor's data for them, and score what comes back. Three checks do most of the work: record coverage, field fill, and recency.

01What Does Testing Coverage Claims Mean?

Every data vendor publishes a headline number: profiles, companies, postings, updates per month. The number describes the vendor's own dataset, measured by the vendor's own method, which is exactly why it cannot settle anything. Testing coverage claims yourself means building a ground truth you control, a sample of entities whose real details you know independently, and scoring the vendor against it. The goal is not to audit a billion records. It is to convert one marketing number into four smaller ones you can act on: coverage, completeness, accuracy, and freshness.

02How to Run the Tests Yourself

1. 01Build ground truth from what you already know: Your current employer. Your previous one. Twenty businesses on the streets you walk. Your own role and title. You are the authority on this sample, no methodology document required.
2. 02Add a random sample: Convenience samples flatter vendors. Pull 50 to 100 companies from a public list you did not curate: a city business directory, an industry association's member list, a funding announcement page.
3. 03Run the lookups: Query the vendor for each entity by name, domain, or whatever key the API supports. Log what came back and what did not.
4. 04Score four dimensions: Coverage, what share of your sample exists at all. Completeness, on records that exist, what share of the fields you care about are filled. Accuracy, spot-check values against what you know. Freshness, compare dates to changes you know happened this quarter.
5. 05Run the known-list test: Take a canonical list, the Fortune 500, the top 100 companies in your target industry, and measure recall against it. Missing 8 of 100 names you know should be there is a finding, regardless of the total count on the website.
6. 06Repeat across two vendors and compare pairs: The same entities, queried on both datasets. Where they disagree, you have found either a bug or a methodology difference, and either one is worth more than any headline number.

03Self-Testing vs. Trusting the Claim

04Key Benefits of Testing Data Claims Yourself

Numbers stop being marketing. A billion anything is unfalsifiable in a demo. Fifty companies you picked are not. Once you have run the test, the conversation with every vendor starts from your spreadsheet, not their slide.

You find your gaps, not the average gap. A dataset can cover 95% of US tech companies and still miss the exact segment you sell to. A self-run test measures the coverage of your market, which is the only coverage that decides your ROI.

Freshness becomes measurable before it becomes expensive. Compare five records against changes you know happened this quarter. If the vendor is three months behind on entities you can check, the monthly update claim has already answered itself.

Vendor disagreements turn into signals. When two datasets differ on the same company, one of three things is true: one is stale, one is wrong, or they define the field differently. Each explanation changes your buying decision, and only a side-by-side test surfaces it.

You negotiate from evidence. "Your data missed 12 of the 50 companies in my test sample" is a sentence that changes pricing conversations in a way that "your competitor says they are bigger" never will.

05Bottom Line

A coverage claim is a promise, and promises are cheap in both directions. An afternoon with a spreadsheet of entities you personally verified will tell you more than any number on a pricing page. Run the test before the demo, bring the results to it, and let every vendor respond to the same rows.

MixRank asks for exactly this. Send us your test sample, or the fields where our data disagrees with what you know, and we will tell you why, including the rows where we are the one who is wrong. That offer is not confidence. It is what hourly-refreshed data with preserved source records makes possible.