How do company lookalike APIs work?

There is no single method. Providers implement lookalikes differently: neural web search (Exa), a global company graph (Ocean.io), shared tech / news / jobs co-signals (PredictLeads), and agentic research (Parallel). Because the mechanism differs, the relevance of the results differs — which is exactly what an independent benchmark measures.

How do you evaluate a company lookalike API?

Give every provider the same seed companies and the same input, then score how many of the companies each one returns are actually relevant. Top-of-list quality (are the first ~10 good?) and long-list quality (are the first ~100 good?) are different axes — a provider can win one and lose the other. Total coverage rounds out the picture. Openbenchmarks runs exactly this comparison with an LLM judge on identical inputs.

benchmarks/lookalikes/what is a company lookalike API?

lookalike benchmark · concept

What is a company lookalike API?

Q: What is a company lookalike API?

A company lookalike API (also called a similar-companies API) takes a seed company and returns other companies that resemble it — by product, market, size, technology, or business model. Teams use it to expand an Ideal Customer Profile, build prospecting lists, or find competitors and comparables without manual research.

A company lookalike API — also called a similar-companies API— takes one seed company and returns other companies that resemble it. Teams use it to expand an Ideal Customer Profile, build prospecting lists, or find competitors and comparables programmatically, without manual research. The hard part isn't getting a list back — it's whether the companies on it are actually relevant. That's what this benchmark measures.

route by workflow

There is no single “best” company lookalike API — this independent benchmark points to a different winner depending on how you will use the results. Pick the workflow that matches yours:

Low volume, high accuracyPredictLeadstop 10 P@10 95.8%
Reps hand-work a handful of accounts, so the top of the list has to be right. Sharpest top-of-list precision in the benchmark.
High volume, broad reachParallellong list P@100 67.5%
A big list runs through automated sequences, so relevance has to hold deep into the results. Best long-list precision in the benchmark.
Market / TAM sizingParallelrelevant 3,242
You are counting the universe of similar companies, not actioning a shortlist. Returns the most relevant companies across the cohort.
Real-time in-product featurePredictLeadsavg latency 739 ms
A live “similar companies” lookup called on page-load, where response time is the constraint. Fastest average latency in the benchmark.

Full ranking, per-seed evidence, and average latency: the lookalike benchmark →

[01] how it works

How each provider implements lookalikes — and where each is strongest

There is no single method. Each provider builds "similar" differently, which is why their results — and their strengths — diverge. The four approaches on this benchmark:

Provider	How it finds lookalikes	Strongest at (measured)	Precision@100
Parallel	agentic web research with natural-language lookalike prompts	Best for long-list relevance (Precision@100)	67.5%
Extruct	domain-based similar-company ranking from a company index	Strongest at long-list relevance (Precision@100) (#2 of 5)	61.2%
Ocean.io	AI-driven lookalike search across a global company graph	Strongest at long-list relevance (Precision@100) (#3 of 5)	56.5%
Exa	neural company search using a seed name and description	Strongest at top-of-list precision (Precision@10) (#2 of 7)	52.4%
Discolike	domain lookalike discovery over BizData company profiles	Strongest at long-list relevance (Precision@100) (#5 of 5)	35.2%
PredictLeads	similar companies via shared tech / news / jobs co-signals	Best for top-of-list precision (Precision@10)	—
CUFinder	company lookalike finder using a name, domain, or LinkedIn URL	Strongest at top-of-list precision (Precision@10) (#7 of 7)	—

[02] by industry

Where each provider shines, by industry

We read the per-industry results as patterns, not precise rankings — the seed set is focused, and only B2B SaaS and fintechhave enough seeds to call confidently. Aggregating each vendor's per-seed relevance by the seed company's industry, the shape that emerges:

Provider	What the per-industry results suggest
Parallel	Leads B2B SaaS, Cybersecurity, Fintech, Dev tools, E-commerce, Healthtech, and Home services — its best-sampled win. Beyond that, strongest in Hospitality and Logistics.
Extruct	Leads Logistics — its best-sampled win and tops Energy. Beyond that, strongest in Home services and Hospitality.
Ocean.io	Leads Hospitality — its best-sampled win and tops Real estate. Beyond that, strongest in Home services and Fintech.
Exa	Leads Industrial — its best-sampled win. Beyond that, strongest in Home services and Logistics.
Discolike	Tops no industry on long-list relevance. Its relatively best sectors are Logistics and Hospitality, but it trails the category leaders everywhere — built for the first handful of results, not the full hundred.
PredictLeads	Tops no industry on long-list relevance, and posts the best top-10 precision on the whole board. Its relatively best sectors are Fintech and Energy, but it trails the category leaders everywhere — built for the first handful of results, not the full hundred.
CUFinder	Tops no industry on long-list relevance. Its relatively best sectors are Energy and Healthtech, but it trails the category leaders everywhere — built for the first handful of results, not the full hundred.

Relevance varies by sector — the per-industry breakdown above shows where each vendor leads, beyond the single headline number. See the full per-seed matrix →