Find lookalike companies that resemble your best customers

What are lookalike companies?

Lookalike companies resemble the ones already buying from you: what they sell, who they sell to, roughly how big they are and how they make money. A lookalike company finder starts from examples instead of filters. You show it a few customers you are glad you won (FoxKnock calls them seeds), and it looks for more companies of the same kind.

"Similar" in FoxKnock means similar in what companies say about themselves, not just in industry labels. It reads each seed's website, turns the text into an embedding (a numeric fingerprint of what the company does) and compares candidates with that fingerprint.

With a meaning-based embedding model, which FoxKnock uses when one is set up (a key on the platform or one you add in Settings), two companies with different industry tags can still land close together because they describe the same product for the same buyer. Without one, FoxKnock falls back to a built-in embedder that follows shared words and phrases, so matches are rougher.

For example, a 3-person agency whose best clients are direct-to-consumer skincare brands could paste five of those brands. The run would look for other brands that sell and describe themselves in a similar way, and tell the agency what each one shares with its clients and where it differs. This is an illustration, not a result.

How to find companies similar to your customers

Open Lookalikes mode and paste the domains of 1 to 10 customers you are glad you won. Add lost customers as negative seeds if you want poor fits pushed down, and pick a segment if you want its criteria in the score. Every run follows the same path, and each part is listed in the run details.

How the lookalike company finder ranks results

Candidates come from three places:

• A similar-company search, run once for each seed.
• A web search built from what your seeds have in common.
• Your CRM: up to 30 companies that sit closest to your seeds as a group, among those FoxKnock has already turned into embeddings.

The combined list is cut at 150 before scoring, in that order, so a very large result count can crowd out the later sources. Your seeds, lost customers and own domain are left out, and so is anything in your CRM marked customer, competitor or disqualified, and anything you rejected recently.

Each candidate is compared with every seed and keeps its best match, and the row names that nearest seed. That is useful when your customers fall into a couple of types. It also means one unusual seed can pull in unusual results, which is why the tips below suggest one customer type per run.

The top 12 matches also get a short explanation: what the company shares with your seeds and how it differs. FoxKnock sends the text it gathered about each company to Claude, by Anthropic, which is told to write the note from that text only. If AI is unavailable, the notes fall back to keyword overlap and the run details say so.

How to get better results from a similar companies search

• Use one customer type per run. If you serve agencies and in-house teams, run them separately. A mixed seed list gives a mixed result.
• Choose seeds for quality, not size. Pick customers who stayed, got value and were easy to win. Three to five is a good start, and the form allows up to ten.
• Add lost customers only when there is a pattern. Two churned accounts that were all too small to afford you make good negative seeds. A random loss does not.
• Select a segment when you know hard criteria. Country, employee range and excluded domains come from the segment. Leave it off if you want pure similarity.
• Read the "differs" note before accepting. It is the quickest way to spot a company that looks right on paper and is wrong for you.
• Run small, then adjust. Searching is free, so start with the default 25 results, accept a few, refine the seeds and run again.
• Check that your seed sites load. When a seed's site cannot be read and your CRM has nothing on it, FoxKnock falls back to its domain name and your segment description, and similarity gets rough.
• Know the run limits. Discovery runs of every mode are capped at 12 an hour and 40 a day per workspace, and at 8 an hour and 30 a day per person. A daily AI budget also applies.

Not sure who your best customers are? The ideal customer profile guide walks through it, and the guide to finding similar companies covers the wider tactics.

Lookalike search, filters or Describe mode?

The usual way to build a target list is to filter a company database by industry, headcount and country. That works when you can state the criteria in advance. It works less well when your best customers share something filters cannot express, such as a business model, a type of buyer or the way they describe their product.

Use lookalikes when you can point at good customers but cannot name the pattern. Use Describe mode when you can say who you want in a sentence, for example "Series A fintechs in DACH hiring a first SDR". An AI planner turns the sentence into web searches and company-data filters, flags companies you already have, and scores each result against your segment (or against your description when you pick none) with evidence for every criterion.

Both modes end in the same staged results list and the same pay-on-accept rule, so you can run each on the same segment and compare what they find.

What a lookalike company search costs

• Searching: free. Results are staged, and nothing enters your CRM until you accept it.
• Accepting a company: 15 credits at default prices, which can change, so check the pricing page for current numbers. One credit is $0.01 at list price, so accepting 20 companies costs 300 credits at default prices.
• Companies already in your CRM: free to accept. A record is charged once ever.
• Follow-up work: accepting can also queue standard research or contact finding for up to 50 companies at a time. Each is charged only when it succeeds, and the price is shown before you click.

Staged results you never accept, including rejected ones, are deleted after 30 days, and a rejection is remembered only until then. On the Free plan, the 500 one-time credits cover about 33 accepted companies at default prices. Paid plans start at $49 a month with 5,000 credits. Work emails are a separate step once you have contacts, and email finding has its own credits.

Try it on your own market: sign up free, paste a few of your best customers and use the free credits to accept the companies worth keeping.