How to Build a Lookalike Account List From Closed-Won
TL;DR: A useful lookalike list starts with a clean closed-won seed set, separates true fit from deal-history noise, and ranks candidates with attributes that explain why the original customers bought. Validate the list against holdout accounts before activating it, then keep eligibility and exclusions in the workflow.
What is a closed-won lookalike account list?
A closed-won lookalike list is a set of companies that resemble proven customers on characteristics relevant to the purchase. It should not simply copy industry, employee count, and geography from the CRM.
The strongest lists combine durable fit attributes with buying context. Durable attributes help define who could buy, while current signals help decide who deserves attention now.
| Input | What it contributes | Quality check |
|---|---|---|
| Closed-won seed accounts | Examples of customers that completed the buying process | Remove one-off deals and records with missing firmographics. |
| Firmographic profile | Industry, company size, geography, and business model | Normalize values before comparing companies. |
| Technographic and operating context | Relevant tools, workflows, hiring patterns, or market motion | Keep only attributes connected to the use case. |
| Commercial context | Deal size, cycle, product, and expansion potential | Separate attractive customers from merely easy-to-close customers. |
| Current signals | Website, product, hiring, relationship, or research activity | Use signals for timing, not as a substitute for fit. |
Build the seed set before generating candidates
- Define the outcome: decide whether the list should optimize for fast acquisition, strategic fit, expansion potential, or another commercial goal.
- Choose a coherent cohort: group customers by product, segment, region, or motion before looking for similarities.
- Remove noisy wins: exclude exceptional founder relationships, inherited contracts, and deals that do not represent a repeatable motion.
- Normalize the CRM: deduplicate accounts, standardize domains and industries, and repair missing company associations.
- Write the hypothesis: state why each selected attribute should predict fit. If the reason is unclear, do not use the attribute.
- Reserve a holdout: keep known good accounts out of the seed set so the method can be checked before activation.
Score fit separately from timing
Fit and timing answer different questions. A company may resemble your best customers but have no current reason to engage, while a high-intent account may be outside the segment you serve well.
| Layer | Question | Examples of evidence |
|---|---|---|
| Fit | Does this company resemble customers the team can serve successfully? | Segment, operating model, use case, geography, and relevant technology context. |
| Timing | Is there a current reason to prioritize outreach? | Website activity, product usage, hiring change, relationship signal, or verified business event. |
| Governance | Is the account eligible to contact? | Customer status, open opportunity, territory, suppression, and recent-contact rules. |
Validate with a holdout test
Before using the list, check whether the model rediscovers the held-out accounts and whether its highest-ranked new accounts make sense to sellers who know the segment. This is a diagnostic, not proof of future conversion.
Review false positives closely. They often reveal an attribute that is common but not causal, a taxonomy problem in the CRM, or a seed cohort that mixes different buying motions.
Activate the list with exclusions intact
Unify’s current product recap describes Lookalikes as a signal that identifies companies similar to selected customers. The Unify Knowledge Base also notes that a lookalike Play runs against its current seed set and becomes static, so expanding the seed list is part of keeping the workflow useful.
Activation should still pass through persona selection, enrichment, qualification, ownership, and suppression before anyone enters a sequence. The list proposes accounts; the governed workflow decides what happens next.
Build a lookalike workflow in Unify.
Frequently asked questions
How many closed-won accounts are needed for a lookalike list?
There is no universal minimum. Use a coherent cohort large enough to reveal repeatable attributes, and test the output against held-out accounts before activation.
Should intent signals be included in the lookalike model?
Use durable attributes to define fit and current signals to prioritize timing. Keeping the layers separate makes the logic easier to audit.
How often should the list be refreshed?
Refresh when the customer base, product, segment strategy, or seed cohort changes, and review static outputs before reactivation.
Sources
- This Year in Product, Unify, updated June 2026.
- Getting started with Plays, Unify Knowledge Base, updated June 2026.
- Unify - Signals, Unify, accessed September 2026.
