Lookalike accounts are companies that resemble a set of example companies - usually your best customers - on the attributes that made those examples good: industry and sub-industry, size, technology stack, hiring pattern, business model, and the way they describe themselves. Instead of writing filters and hoping they capture what a good customer looks like, you hand over the customers themselves and let similarity do the describing. It is the most direct way to expand a list from evidence rather than assumption.
The definition
A filter-based list answers "which companies match these criteria". A lookalike list answers "which companies are most like these ten". The difference matters when the criteria are hard to write down - a niche vertical that standard industry codes misclassify, a business model that firmographics do not capture, a "feel" the best reps recognize and cannot express as a filter. The seed carries that information implicitly, and the similarity model extracts it. It is the same idea the ad platforms use for lookalike audiences, applied to companies rather than people, and it is one of the ways the prospecting agent in the company data catalog identifies accounts.
How similarity is computed
| Input | Captures | Caution |
|---|---|---|
| Firmographics | Size, industry, revenue band, location, ownership | Coarse alone; it is the baseline, never the whole answer |
| Technographics | Stack and stack changes | Detection confidence varies by method |
| Descriptions and keywords | What the company says it does, in its own words | Marketing language drifts; weight it with the structured fields |
| Hiring and funding pattern | Growth stage and where investment is going | Time-sensitive; decays on its own clock |
| Relationships | Partners, integrations, customers, suppliers | Sparse for smaller companies |
The weighting is where lookalike tools differ. A model that leans on descriptions finds companies that talk like your customers; one that leans on firmographics finds companies shaped like them; the useful models combine both and let the seed's own variance say which attributes matter. The honest check is to hold out a few real customers from the seed and see whether the model ranks them near the top.
Choosing the seed
The seed is the whole game. Use closed-won customers that stayed and expanded, weighted toward the ones with the best margin or the shortest cycle, and leave out the accidental wins, the churned accounts and the deals that only closed on a discount - each of them teaches the model the wrong shape. Apply the suppression list before you run, so existing customers, competitors and their subsidiaries do not come back as "new" lookalikes. Ten to fifty well-chosen seeds outperform two hundred mixed ones, and the seed should be revisited whenever the ICP is revalidated against what actually converted.
Where lookalikes fit in a list build
Lookalike discovery is a first step, never the list. The similar companies still need firmographic confirmation (similarity can surface a namesake or a subsidiary), a waterfall to fill the fields the ICP scores on, an ICP re-check on the enriched row, the people at each account, verification before any send, and dedupe against accounts already worked. Run as an agent mission, the discovery and the six steps behind it are one objective: "200 companies most like our best forty customers, enriched, verified, with the head of operations at each". Run as a standing mission, the lookalike set refreshes as new customers close and the seed improves.
Frequently asked questions
What are lookalike accounts?
Companies that resemble a seed set of example companies - usually your best customers - on the attributes that made the examples good: industry, size, stack, hiring and funding pattern, description and relationships. They are found by similarity to real examples rather than by writing filters.
How is a lookalike account different from an ICP filter?
A filter captures what you can write down; a lookalike captures what the seed implies, including attributes you never articulated. Both belong in a list build: the ICP scores and re-checks the rows that lookalike discovery surfaces.
How many seed companies do I need?
Ten to fifty well-chosen closed-won, retained, high-value customers outperform hundreds of mixed ones. Leave out accidental wins, churned accounts and discount-only deals, and apply the suppression list before running.
Are lookalike accounts the same as lookalike audiences?
Same idea, different object. Lookalike audiences expand a seed of people on an ad platform for targeting; lookalike accounts expand a seed of companies for prospecting. A lookalike account list can later become an ABM audience on LinkedIn.
Sources
- Gartner - the B2B buying journey research behind account-level targeting
- Ehrenberg-Bass Institute - category buyer research behind in-market share assumptions
Every playbook on this blog ships as a runnable mission.
Open a workspace and the playbook library is waiting - describe the outcome and the agents carry it end to end, on your plan's monthly credits.