Case study
The Vendor Report Card
Treating any single vendor as the source of truth is how GTM teams end up debating the data instead of using it. The Vendor Report Card graded six enrichment and prospecting vendors against a 200-company ground-truth reference set, field by field, and produced a Best Source by Field hierarchy that replaced vendor-reputation assumptions with evidence.
Overview
Vendor reputation is a marketing signal. Source-of-record needs an evidence signal.
The default in most orgs is to pick one enrichment vendor and call it the source of truth, or to layer two and let whichever wrote last win. Both approaches lose to the same problem: a vendor that nails employee counts can be wrong about countries, and the org has no way to tell which call to trust.
The Vendor Report Card tested six vendors against a 200-company reference set, with the ground-truth values scrubbed directly from each company's own website. Each vendor was scored on specific fields. The output was not a "best vendor" winner. It was a grid: this vendor wins for this field; that vendor wins for that one. From that grid came a Best Source by Field hierarchy: Platinum, Gold, Silver, and Bronze tiers, assigned per field, not per vendor.
A field-level benchmark replaces "we use this vendor" with "we use this vendor for this field."
The problem
What "vendor preference" was hiding.
Vendor selection was a marketing decision. Source-of-record was an unspoken assumption. The Vendor Report Card was built to surface both.
- 01
Multiple vendors wrote to the same fields with no defined precedence.
When a field had two or three vendor sources, whichever wrote last won. A vendor strong on Industry could overwrite a vendor strong on Revenue with a worse value, and nothing in the system noticed.
- 02
Vendor reputation drove vendor selection.
Vendor choice was made at the procurement level, by company-wide reputation and pricing. Field-level fit was not part of the decision, so one vendor often ended up treated as authoritative for everything, even the fields it was demonstrably weak at.
- 03
There was no evidence base to point to in a vendor debate.
"Vendor X is better for this field" was opinion. No reference dataset, no scored comparison, no graded output. Vendor decisions came down to whoever argued hardest.
- 04
Source-of-record drift went unmeasured.
A vendor could be losing accuracy quarter over quarter and nobody would know until a downstream report fell apart. Without a ground-truth reference set to test against, there was no signal until it was already too late.
A vendor is not a source. A source is a vendor demonstrably best at this field, on this ground-truth reference set, this quarter.
What I built
One benchmark, two outputs.
A 200-company ground-truth reference set.
A curated set of 200 companies, chosen to span industries, sizes, and geographies, served as the ground truth. The reference values for each field were scrubbed directly from each company's own website, the most authoritative source for what a company actually says it is, then held stable so vendor accuracy could be measured against a constant.
Field-level vendor grading.
Six vendors (ZoomInfo, Crunchbase, HG Insights, Lusha, LeadIQ, and Clearbit) were scored for accuracy on five fields: Employees, Revenue, Industry, Country, and State. The output was a grid, vendor by field, with each cell graded against the reference. No averaging, no overall vendor ranking. A vendor could be strong on Industry and weak on Revenue and the grid said so out loud.
The Platinum, Gold, Silver, Bronze hierarchy.
Within each field, vendors were tiered by accuracy. The tiers were not awarded by vendor brand or pricing. They were assigned by field-level performance, so the same vendor could be Platinum on one field and Bronze on another.
A Best Source by Field model.
The tiers fed directly into a Best Source by Field assignment, mapping each field to its highest-scoring vendor, with backup tiers when the primary vendor lacked data on a given record. This is what the enrichment stack and the Salesforce write-back logic consumed, instead of "vendor X always wins."
A connection back to CAP Accuracy.
Vendor Source Strength is one of four lenses inside the Core Account Profile's Accuracy dimension, alongside cross-field validity, multi-source vendor agreement, and recency. The Vendor Report Card was the evidence behind that lens, so the Accuracy score had a defensible underpinning instead of "we trust the data."
Outcomes
What changed downstream.
- Enrichment trust decisions became defensible. "Why is this field coming from this vendor?" had a graded, sample-tested answer instead of a vendor-reputation one.
- Source-of-record rules stopped being implicit. The Best Source by Field hierarchy lived in the enrichment pipeline as a written rule, not in someone's head.
- The CAP Accuracy dimension had a basis. Vendor Source Strength stopped being qualitative; it had a sample size, a method, and a grade.
- A reusable benchmark structure. The same test could be re-run when vendors changed pricing, added a field, or claimed an accuracy improvement, with apples-to-apples results.
A defensible source hierarchy was not the goal. It was the byproduct of a benchmark that finally answered which vendor was actually good at what.
The portable POV
What carries to any data-source decision.
Vendor evaluation is a category of operator work that happens far too often without evidence. The fix is small in scope and disproportionately useful.
Vendors deserve field-level grades, not letter-grade reputations.
A vendor is not "good" or "bad." A vendor is good at some fields and bad at others. Grading the wrong unit is most of what makes vendor decisions feel arbitrary.
A ground-truth reference set is non-negotiable.
Without a ground-truth set you control, vendor accuracy claims are advertising. With one, they become measurements.
Source-of-record is downstream of vendor benchmarking, not upstream of it.
You cannot decide "which source wins" without first knowing which source is actually better at what. Most orgs reverse this order and then wonder why their data disagrees with itself.
The output is a rule, not a recommendation.
A vendor benchmark that ends in a slide deck is informational. A vendor benchmark that ends as a Best Source by Field rule inside the enrichment pipeline is operational.
Robots can compare two vendor values. They cannot tell you which one is supposed to win.
The evidence under the Vendor Source Strength lens of CAP Accuracy.
The Core Account Profile uses four lenses to score Accuracy. Vendor Source Strength is one. The Vendor Report Card is what made that lens credible, instead of "we trust the data." The same source-trust pattern applies to any data product where the answer needs to be defensible per field, not per vendor.
Redaction note
This case study is generalized from real GTM systems work. Specific company names, customer data, internal field mappings, exact operational telemetry, and proprietary implementation details have been removed or abstracted. Any demo data shown on this site is synthetic.