Lookalike finder

Ranks the hundred companies most like your best customers, and shows the buying committee at the top ten.

For: Account executives
You provide: Two or three companies you already sell to
You get: 100 companies ranked by similarity
The buying committee at the top ten
The top accounts imported to your workspace

Get the prompt

A lookalike finder for an Adobe enterprise seller.

The ranked list narrows to those twenty-three.

Built as a demo for Adobe, against a real Trayo workspace

The prompt

Paste it into any coding agent with the Trayo MCP connected.

+ any MCP client

You have the Trayo MCP server connected. Carry out this recipe against the real API.

# Find companies like these

When to use: You have example companies and want a ranked list of similar companies.
Cost: Nothing. This search is not metered.

## Steps - findLookalikeCompanies: Send {"companies":[{"linkedinUrl":"https://www.linkedin.com/company/stripe"}],"limit":100}. Each of the 1–50 seeds can carry companyId for an exact company reference, or linkedinUrl, website and name for matching in that order. companyId accepts a number or numeric string; it takes precedence when supplied and is not a workspace account UUID. The limit is the total number of results wanted, up to 500. One response, no pagination.

## Then Read seeds to check matched identities and any not_found or not_indexed inputs. Reuse companyId from a matched seed or result in a later search. The companies come best match first, each with a preliminary high, medium or low similarity label. filtered counts the seed subsidiaries, repeat listings and companies outside the seeds' employee range that were left out; send matchSize false to allow any size. Results also rank higher when people at your workspace's accounts used to work there; send useWorkspaceAccounts false to rank on the seeds alone. Nothing is saved. To save selected companies, use batchCreateAccounts with name and a website url including https://, in batches of 200. Also send linkedinHandle from linkedinUrl when available; omit rows without a name or usable website.

Take the tour

A narrated walkthrough of the app.

Sound off

The building blocks

What this app asked Trayo for

Every name, number and email on the page came back from one of these. None of it was typed in.

Reads your workspace

What you sell, who you sell to, and the buyer personas already saved in Trayo.

Finds companies

Surfaces accounts that match your ICP, plus lookalikes of the ones already working.

Builds your account list

Imports, updates and organises accounts so the work survives the session.

How it was built

The data behind each screen, and what's faked for the demo.

Built in 15 min · 3 capabilities · 9 live API calls

Lookalike Finder is a screen for an Adobe enterprise seller who has a few good customers and wants the next hundred companies that look like them. The Targets view lists the ranked result with its similarity band, score, headcount, country and whether the company is already an account in the Adobe workspace; it filters by band, country and free text, pages at 25 rows, and opens a row for the match detail and the people found there. Buying committee shows those people per company against Adobe’s five saved personas, and How this was built records every operation behind the screens.

Everything shown came from the live Trayo API on 2026-10-08. The Adobe workspace had no accounts, so the seeds came from its saved ICP — global consumer brands with large in-house creative and campaign teams: Nike, L’Oréal and Sephora. One trayo_find_lookalikes call with those three seeds and the ICP size band as a hard filter (1,000–100,000 employees) returned 100 companies, 62 high similarity and 38 medium, and reported 3 seed subsidiaries and 5 repeat listings left out. A trayo_search_stakeholders_batch run over the top ten results, using the workspace’s own stakeholder definition, screened 174 candidates and returned 34 people with Trayo’s reason for each. The top 20 companies were then imported with trayo_import_accounts; all 20 were created and show as saved accounts in the UI.

Two earlier searches were thrown away and are written up on the method page: seeding by website with wpp.com and omnicomgroup.com resolved to unrelated companies and came back not_indexed, and a run with no size filter returned corner shops, because the seed employee range did not apply on its own. No emails or phone numbers were enriched, so the people carry no contact rows — those lookups are metered and were out of scope. Push to Salesforce, Share to Slack, Start sequence and Draft intro email are previews: they confirm with a toast that says nothing was sent. Export CSV is real and writes the filtered rows as Trayo returned them. The brand module took #FB0F01 as primary and reported its adjustments, which are printed on the method page: no background colour was supplied, so the page is a white canvas tinted #fff8f7; accent text drops to #c50700 to clear 4.5:1 on that page; and the top bar takes a deep step of the primary, #810300.