Stable large
Active, non-subsidiary companies with at least 1,000 employees, founded by 2015. This is the well-documented baseline every provider should resolve reliably.
282 companies · company enrichment
Company Enrichment Benchmark
8 APIs—People Data Labs, Apollo, Parallel, Predict Leads, Explorium, CompanyEnrich, ZoomInfo, and Exa—receive the same company domain and enrich it by API calls from their official docs. The scored fields are HQ country/city, founded year, industry, LinkedIn URL, and headcount. Headcount passes when an exact count is contained by the reference range, or the reference exact count is contained by the provider range.
This live cohort contains 282 reachable companies across stable-large, long-tail, subsidiary, and verified-rebrand slices. The headline metric is enrichment success rate: semantically correct returned fields divided by all available human-verified Ground Truth fields for that company. Field accuracy, company match rate, field coverage, and speed are shown beside it.
[00] verdict
Which company enrichment API is best?
There isn't one best enrichment API. The right choice depends on what happens next: filling CRM records, avoiding bad overwrites, matching a prospect list, returning complete company profiles, or powering a latency-sensitive product flow. Choose the job below; each recommendation is based on the metric that matters for it.
Which provider wins depends on the use case — CRM writeback, prospecting and matching, complete company profiles, or real-time product flows. The workflows below are ranked separately for each.
| Workflow | Recommended API | Measured evidence |
|---|---|---|
| Best API for keeping CRM records current | People Data Labs | 89.0% enrichment success rate |
| Best API for safe CRM writeback | Apollo | 92.6% field accuracy |
| Best API for the most complete company data | Parallel | 94.7% field coverage |
| Best API for matching a company list | Predict Leads | 100.0% company match rate |
| Best API for real-time product or agent enrichment | People Data Labs | 274 ms median latency |
[01] provider comparison
Compare provider trade-offs across data, coverage, speed, and cost.
Each column measures a different trade-off; providers are shown alphabetically, not as an overall ranking. Use the workflow recommendations above to select the metric that matters for your job. GPT-5.6-terra compares semantic reference/provider field pairs, while GPT-5.6-sol applies the dedicated numeric employee-band policy; missing reference data is excluded from the denominator. Estimated cost uses each provider's entry paid public self-serve tier; pricing details are in the methodology.
benchmarks/firmographic/company-firmographic-enrichment-web-research-v2-293live
| Provider | Official docs | Resolution | Enrichment success rate | Field accuracy | Field coverage | Median latency | Est. cost | Endpoint & configuration |
|---|---|---|---|---|---|---|---|---|
| Apollo | Apollo official docs | 96.8%273/282 domains | 88.6% | 92.6% | 93.0% | 654 ms | $10.88$0.0196 / credit | POST /api/v1/mixed_companies/search → GET /api/v1/organizations/{id}page=1; per_page=1 |
| CompanyEnrich | CompanyEnrich official docs | 98.6%278/282 domains | 69.7% | 72.5% | 93.7% | 309 ms | $2.76$0.0098 / credit | GET /companies/enrichdomain={company domain} |
| Exa | Exa official docs | 100.0%282/282 domains | 60.4% | 76.2% | 77.1% | 8,143 ms | $4.23$0.015 / deep-reasoning request | POST /searchtype=deep-reasoning; category=company; numResults=10; highlights + structured output |
| Explorium | Explorium official docs | 95.0%268/282 domains | 70.5% | 87.6% | 76.0% | 616 ms | $22.00$0.04 / credit | POST /v1/businesses/match → POST /v1/businessesmode=full; page=1; page_size=10; size=10 |
| Parallel | Parallel official docs | 99.7%281/282 domains | 86.0% | 89.4% | 94.7% | 21,317 ms | $14.10$0.050 / Responses request (medium reasoning) | POST /v1/responsesmodel=parallel; reasoning.effort=medium; JSON schema output |
| People Data Labs | People Data Labs official docs | 98.2%277/282 domains | 89.0% | 91.2% | 94.3% | 274 ms | $27.70$0.10 / match | GET /v5/company/enrichwebsite={company domain} |
| Predict Leads | Predict Leads official docs | 100.0%282/282 domains | 80.6% | 86.5% | 87.9% | 556 ms | $40.00$0.04 / API credit · 100 free; $40 monthly minimum | GET /api/v3/extended_companies/{domain}1 company domain per request; follows API redirect target when returned |
| ZoomInfo | ZoomInfo official docs | 94.7%267/282 domains | 63.9% | 70.4% | 88.7% | 1,132 ms | $93.45$0.35 / new company record | gtm companies enrich--domain {company domain} --fields name website domainList city state country primaryIndustry industries employeeCount employeeRange foundedYear socialMediaUrls |
[01.a] by attribute
No provider is best at everything.
The comparison above blends every attribute into one number, which hides where a provider actually wins. Each cell below is the share of companies where that provider returned the correct value, counted only over companies that carry a verified reference for that attribute — so the denominators differ by column. Bold marks the leader. A dash means the provider returned nothing we could score.
benchmarks/firmographic/by-attributelive
| Provider | Official docs | HQ location | Industry | Employee band | Founded year | LinkedIn URL |
|---|---|---|---|---|---|---|
| Apollo | Apollo official docs | 87.3%240/275 | 95.4%269/282 | 78.2%219/280 | 91.8%191/208 | 90.8%256/282 |
| CompanyEnrich | CompanyEnrich official docs | 50.9%140/275 | 74.8%211/282 | 47.5%133/280 | 90.4%188/208 | 91.1%257/282 |
| Exa | Exa official docs | 82.9%228/275 | 96.1%271/282 | 33.2%93/280 | 97.6%203/208 | 5.3%15/282 |
| Explorium | Explorium official docs | 73.5%202/275 | 84.8%239/282 | 88.9%249/280 | 0.0%0/208 | 86.5%244/282 |
| Parallel | Parallel official docs | 90.9%250/275 | 92.9%262/282 | 80.7%226/280 | 76.4%159/208 | 86.2%243/282 |
| People Data Labs | People Data Labs official docs | 90.2%248/275 | 94.0%265/282 | 86.4%242/280 | 93.8%195/208 | 84.0%237/282 |
| Predict Leads | Predict Leads official docs | 68.4%188/275 | 88.3%249/282 | 78.2%219/280 | 89.9%187/208 | 84.8%239/282 |
| ZoomInfo | ZoomInfo official docs | 73.1%201/275 | 74.5%210/282 | 47.9%134/280 | 61.4%129/210 | 62.4%176/282 |
[02] methodology+
[02.a] cohort design
Four slices, from baseline to identity stress
The cohort covers four company types: established large companies, smaller companies across languages and countries, subsidiaries, and recent rebrands. Each group tests a different source of enrichment error: limited public data, parent-company confusion, or outdated identity records.
Long tail
Active, non-subsidiary companies with 10–200 employees, sampled across U.S. English, non-U.S. English, and non-English strata. This tests sparse profiles and international coverage.
Subsidiaries
Companies marked as subsidiaries in the discovery index. This tests whether a provider returns the requested entity instead of its parent; the source flag selected candidates but was not used as reference data.
Verified rebrands
Recent one-company name or domain changes verified from Business Wire plus current-site, redirect, company, or regulatory evidence. This tests stale names, domains, and LinkedIn identities.
[02.b] sampling and freeze
From candidate sources to one fixed 282-company input list
- Draw stable-large, long-tail, and subsidiary candidates from a read-only company index using fixed-seed random sampling, country and language diversity caps, and global company/domain deduplication.
- Review 100 recent rows from Business Wire's maintained Company Name Change feed. Accept only unambiguous, one-company transitions with a verified current identity; reject mergers, divisions, partial histories, and nonprofits.
- Human labellers review the candidate frame, remove invalid or unreachable companies, and freeze the remaining 282 inputs before scoring. Candidate-source attributes help select companies but never become expected field values. Every provider receives the same final 282 company domains.
[02.c] identity and reference data
Resolve the entity before judging its fields
- Human labellers refreshed the Ground Truth for the entire current dataset, reviewing each company and every scored field rather than treating LinkedIn as the sole source of truth.
- Review uses the company's official website, filings, company and registry sources, news, and reputable reference sources such as Wikipedia where appropriate. Labellers also verify redirects, alternate domains, and canonical or alternate LinkedIn company/school URLs.
- Provider responses are normalized to the same five scored fields and judged against this human-reviewed Ground Truth with dedicated structured models: GPT-5.6-terra for semantic fields and GPT-5.6-sol for employee band. Missing Ground Truth is excluded from the denominator; every verdict retains its concise rationale for review.
[02.d] headline metric
Enrichment success rate
correct fields ÷ fields with available human-verified Ground Truth- Missing or incorrect provider values lower the score. Fields without reference data do not count.
[02.e] cost calculation
Estimated USD cost
- Recorded usage is multiplied by the public unit rate on each provider's entry paid self-serve tier as of August 3, 2026.
- Fiber
$0.020 / credit; Ocean.io$0.064 / credit; People Data Labs$0.10 / match; Predict Leads$0.04 / API credit · 100 free; $40 monthly minimum; Apollo$0.0196 / credit; CompanyEnrich$0.0098 / credit; Explorium$0.04 / credit; Exa$0.015 / deep-reasoning request; Parallel$0.050 / Responses request (medium reasoning); ZoomInfo$0.35 / new company record. - Free allowances, taxes, minimum commitments, and volume discounts are excluded. Apollo and Explorium use recorded request/credit-event proxies because their responses did not expose an exact charge.
[03] changelog+
2026-08-03 current snapshot
- Human labellers refreshed the entire dataset and every scored Ground Truth field, then removed 11 invalid company/domain inputs, leaving 282 companies.
- Added ZoomInfo GTM CLI/API enrichment results, bringing the comparison to eight providers on the same frozen inputs.
- Rejudged employee band for every provider with the dedicated arithmetic policy in
gpt-5.6-solat medium reasoning effort; it handles reviewed range containment, exact-count tolerance, and boundary equivalents more reliably. - HQ location, industry, founded year, and LinkedIn URL remain field-judged with
gpt-5.6-terraat medium reasoning effort. Each verdict stores a concise rationale. - Expanded equivalence rules for city/metro and UK country labels, near-match exact headcounts and range boundaries, matching LinkedIn school/company slugs, and audited alternate LinkedIn URLs.
- Separated labeller review state from snapshot-owned data and preserve manually edited Ground Truth during future refreshes.
[04] per-company evidence
Every company × every provider.
Each cell shows enrichment success rate, with field accuracy on returned evaluable fields below it. Hover a cell for coverage, status, and latency.
coverage matrix2256 runs
| Company | People Data Labs | Apollo | Parallel | Predict Leads | Explorium | CompanyEnrich | ZoomInfo | Exa |
|---|---|---|---|---|---|---|---|---|
| Envirotech Vehicles, Inc.evtvusa.com | 100.0%accuracy 100.0% | 100.0%accuracy 100.0% | 100.0%accuracy 100.0% | 40.0%accuracy 50.0% | 60.0%accuracy 75.0% | 80.0%accuracy 80.0% | 80.0%accuracy 80.0% | 60.0%accuracy 75.0% |
| Del Monte Corporationfreshdelmonte.com | 75.0%accuracy 75.0% | 75.0%accuracy 75.0% | 75.0%accuracy 75.0% | 75.0%accuracy 75.0% | 50.0%accuracy 50.0% | 100.0%accuracy 100.0% | 75.0%accuracy 75.0% | 50.0%accuracy 66.7% |
| Currentcurrent.co | not foundnot_found | 66.7%accuracy 66.7% | 100.0%accuracy 100.0% | 0.0%accuracy -- | not foundnot_found | 66.7%accuracy 66.7% | 100.0%accuracy 100.0% | 33.3%accuracy 50.0% |
| Airspeedgoairspeed.com | not foundnot_found | 75.0%accuracy 100.0% | 100.0%accuracy 100.0% | 0.0%accuracy 0.0% | not foundnot_found | 50.0%accuracy 50.0% | 75.0%accuracy 75.0% | 50.0%accuracy 66.7% |
| Kinislakinisla.com | 60.0%accuracy 60.0% | 80.0%accuracy 80.0% | 100.0%accuracy 100.0% | 0.0%accuracy 0.0% | not foundnot_found | 60.0%accuracy 75.0% | 60.0%accuracy 60.0% | 60.0%accuracy 75.0% |
[05] faq
Company enrichment benchmark — common questions
Which company enrichment API is best?
There is no universal best because company enrichment supports different jobs. Choose by the next action in your workflow: CRM writeback needs reliable, complete fields; protecting an existing record needs fewer wrong values; prospecting and long-tail research need strong company matching; hierarchy work needs subsidiary performance; rebrands need domain-migration handling; segmentation needs accurate industry and headcount; product or agent flows need low latency; and batch backfills need low cost. The workflow table on this page names the provider with the strongest measured evidence for each job, rather than rolling unlike needs into one overall rank.
Which company enrichment provider is most accurate?
Two different questions: on overall correctness (correct fields ÷ all available reference fields), People Data Labs leads at 89.0%. On accuracy of what is returned (right when present, fewer wrong values written to your CRM), Apollo leads at 92.6%. GPT-5.6-terra judges HQ, industry, founded year, and LinkedIn URL; GPT-5.6-sol at medium reasoning applies the dedicated numeric employee-band policy, both against human-reviewed Ground Truth.
What is the cheapest company enrichment API?
At each provider's entry paid self-serve rate, CompanyEnrich was the cheapest for this cohort ($2.76 total). Estimates use public unit rates as of July 14, 2026; free allowances and volume discounts are excluded, and Apollo and Explorium use recorded usage proxies. Cheapest is not best — check enrichment success rate and field accuracy beside cost.
Are ZoomInfo and Clay in this benchmark?
ZoomInfo is measured in this release through its GTM CLI/API company-enrichment flow. Clay is a workflow tool that runs waterfall enrichment across multiple providers inside its tables rather than exposing a single isolatable enrichment API — a Clay export was used only as a temporary reference during development and was fully replaced by human-reviewed Ground Truth in the published benchmark.
What is enrichment success rate?
The headline metric: correct returned fields divided by all fields with available human-verified Ground Truth for that company. It includes both completeness and correctness: a provider loses points for wrong values and missing values, but never for fields where no Ground Truth exists.
How is the ground truth verified?
Human labellers verified every scored Ground Truth field across official company sites, filings, registries, news, reputable reference sources, redirects, and canonical or alternate LinkedIn URLs. Invalid or unreachable candidates were removed; the frozen cohort contains 282 reachable companies before provider scoring. The resulting judgments and audit artifacts are public and reproducible.
Does this benchmark cover contact or email enrichment?
No — it measures company-level firmographic enrichment software and APIs only: five scored fields (HQ location, industry, employee band, founded year, and LinkedIn URL) from a company domain. Company name and domain are retained for identity audit, not score denominators. Audited domain and LinkedIn redirects are treated as the same identity. Headcount accepts equivalent ranges and reviewed exact-count tolerances. Contact/person enrichment, email finding, and phone data are not measured here; treat provider accuracy on those as vendor claims until measured.
Which company data enrichment provider has the best data?
Two measured answers on identical inputs: People Data Labs has the highest enrichment success rate (89.0%), and Apollo the highest field accuracy (92.6%). Note the scope: this measures company-level (firmographic) data enrichment — contact/person data is a different product, not measured here.
Is this benchmark independent?
Yes. No provider pays for inclusion, presentation, or removal, and there are no equity or referral relationships with the benchmarked vendors. Inputs, normalized provider outputs, and every field-level judgment are public in the open repository, so any number can be reproduced or disputed.
[06] go deeper
Choosing a company enrichment provider
Best company enrichment API — compare measured data → · The most accurate enrichment API (both definitions) → · What is a company enrichment API? → · Clearbit alternatives, measured → · Best enrichment providers & vendors — the landscape → · Best enrichment tools — a buyer's guide → · Best company data enrichment tools — independent comparison → · Cheapest enrichment API — cost beside accuracy → · Best company lookup API — match rate, measured → · ZoomInfo alternatives → · Clay alternatives → · Company enrichment for US companies → · Company enrichment for international companies → · Company industry classification API accuracy → · Employee headcount data API accuracy → · Company headquarters location API accuracy → · Company LinkedIn URL finder API accuracy → · Company founding year data API accuracy → · Apollo vs People Data Labs → · Apollo vs Explorium → · People Data Labs vs Explorium → · Exa vs Parallel →