Open benchmarks
The results are already public.
Free, dated studies of B2B data providers, published with their method and raw data.
All benchmarksIndependent benchmarks for B2B data decisions.
We help companies make better B2B data decisions with free public benchmarks or custom benchmark projects on their own ICP.
Read the benchmarks- Benchmarks published
- 2
- Vendors tested
- 9
- Search queries scored
- 7,000
- People matched
- 2,406
All scoring data is open under CC BY 4.0.
Your next best prospect may be missing from your current data provider. Benchology independently evaluates B2B data sources to help you uncover missed prospects, verify contact accuracy, and build a more effective data stack.
Our benchmarks
Two published studies, with their methods, results and supporting data.
Each benchmark is a snapshot: it shows how providers performed when we measured them. Providers change their data and products, so results may no longer hold, and we do not guarantee that they still do. Read each study's dates, limitations and any re-verification schedule.
Benchology Benchmark 02
Apollo vs Clay: how different are the people they find?
- Apollo only
- Both (overlap)
- Clay only
Bar length = people found by either vendor
ICPstress axis
Overlappeople found by either vendor · total 2,406
Valid row rateshare of returned rows that are valid
- RevOps · San Francisco201–1000 employeesUS tech84.1%Apollo92.6%Clay
- Sales · Munich201–1000 employeesNon-English75.0%Apollo88.2%Clay
- Ops · Ohiomanufacturing, 201–500Industrial79.9%Apollo83.0%Clay
- Marketing · Singapore201–1000 employeesAPAC87.7%Apollo79.8%Clay
- Founders · AustinSaaS, 11–20 employeesSmall / SaaS84.7%Apollo72.2%Clay
† Set 5's two encodings are not equivalent. Overlap figures are floors and unique-contribution figures are ceilings: the identity matcher recovers about 91% of true cross-vendor pairs, so the error runs one way.
See the full comparison1,000 search tasks
Other benchmark
Web Search Tools
Retrieval accuracy (URL-exact match) across the same 1,000 search tasks. Whiskers and the range under each name are the 95% interval.
Higher accuracy is better
Accuracy
0%20%40%60%80%100%
Search provider
Under this study's reporting rule, providers with overlapping intervals are not ranked apart.
● Same tasks● Same inputs● Observed resultsBar color = provider brand
Benchology Benchmark 01 · Reliability
What each vendor commits to
Scored disclosures, not measurements: what each vendor publicly commits to, 0–4 per axis. A low score means thin disclosure, not poor practice. Verified 2026-09-03, valid until 2026-12-02.
All 30 scores as a table
| Vendor | Price | Independence | Docs | Compliance | Procurement |
|---|---|---|---|---|---|
| Exa | 2 | 4 | 2 | 3 | 3 |
| Tavily | 4 | 2 | 1 | 4 | 4 |
| Linkup | 3 | 3 | 2 | 3 | 3 |
| You.com | 4 | 1 | 1 | 2 | 2 |
| Brave Search API | 1 | 4 | 1 | 1 | 2 |
| Keenable | insufficient history | 3 | 0 | 0 | 1 |
The five dimensions
Each scored 0 at the centre to 4 at the outer ring.
- Price
- How much a buyer's cost basis has moved, and how existing customers were treated when it did.
- Independence
- Whether the vendor can be cut off or repriced by an upstream it does not control.
- Docs
- Whether the vendor has made the operational commitments a production integration depends on.
- Compliance
- Whether the vendor clears the first filter of a standard vendor security review.
- Procurement
- Whether an organisation with a procurement function can actually buy this, and how much friction it will meet.
A dashed axis means insufficient history, which is not the same as a score of zero — Keenable had been public two weeks at the verification date. Serper was scored and removed after scoring at the floor on all five dimensions; it stays in the measurements above and in the published data file.
| Provider | Accuracy | 95% interval | Mean latency | Price per 1,000 |
|---|---|---|---|---|
| Exa | 91.1% | 89.2–92.7% | 1.71s | $7.00 |
| Brave | 90.0% | 88.0–91.7% | 532ms | $5.00 |
| You.com | 88.9% | 86.8–90.7% | 672ms | $5.00 |
| Linkup | 84.8% | 82.4–86.9% | 2.21s | $5.00 |
| Serper | 84.4% | 82.0–86.5% | 1.93s | $1.00 |
| Tavily | 83.4% | 81.0–85.6% | 2.23s | $7.50 |
| Keenable | 75.9% | 73.2–78.4% | 404ms | $4.00 |
Custom benchmarks
Run the same test on your own ICP.
Define the audience and providers you want to compare. We agree the scope, test the data and document the findings and limitations.
How a custom benchmark worksIndependence
Public benchmarks are free to read. Buyers pay for custom benchmark projects; we do not accept vendor funding, sponsorships or referral commissions. Vendors cannot pay for inclusion, removal or changes to our results.
Our business modelWho we areEvery number on this page comes from a run we did ourselves, a vendor price we verified and dated, or a disclosure we scored from published documents.
Custom benchmarks
Run the same test on your own ICP.
Bring your ICP and the data providers you use or are considering. We scope a custom evaluation around your decision and show you the evidence: who each provider finds, who your current source misses, and how accurate the records are.