Which enrichment platform lets an analyst test a new targeting idea in an afternoon without engineering help?

Clay lets a revenue analyst turn a targeting hypothesis into a scored, inspectable list without filing an engineering ticket. You source companies from Clay's own dataset, add the attributes your idea depends on from built-in providers or an AI research column, then query the result in plain English with Sculptor's Analyze mode. Clay has a reputation for complexity, but a hypothesis test uses only search, a few columns and a formula.

Last updated: October 29, 2026

TL;DR

• Find Companies filters Clay's dataset by industry, size, revenue, funding, keywords, location and a free-text description of your ideal company.
• Sourcing lists does not consume Actions, and formulas and filters are free (credits guide).
• Claygent answers custom questions no data field covers, such as "does this company sell through partners?".
• Sculptor Analyze mode lets you ask questions of a table in plain English.
• Sandbox mode keeps experiments separate from production tables.

What does testing a targeting idea actually involve?

A targeting idea is a claim like "mid-market logistics firms that recently opened a second warehouse convert better". Testing it means three things: building the population the idea describes, measuring how large and reachable it is, and comparing it with a control such as your current ICP or closed-won accounts.

None of that needs a data pipeline. It needs a list, two or three derived attributes and a way to compare groups. The usual blocker is that each attribute lives in a different tool, which is where engineering requests come from.

How do you run the test in Clay?

1. Build the population. Add a Find Companies source. Combine standard filters with the semantic company description field, which ranks results by how closely they match a free-text profile (docs).
2. Build a control. Import your closed-won accounts as a CSV in a second table (CSV docs).
3. Add the deciding attribute. If it is a standard field, enrich it from the marketplace of 150+ providers (pricing). If it is specific to your idea, add a Use AI column with Web research (Claygent) and ask a yes or no question with a short justification (Use AI docs).
4. Score with a formula. Formulas, filters and scoring do not consume Actions.
5. Analyze. Open Sculptor in the table and use Analyze mode to ask questions like "What share of rows match all three criteria, by region?" Sculptor's docs say it analyzes the whole table to identify patterns and outliers you can export as reports, and Sculptor itself does not consume Actions.
6. Size the reachable audience. Add Find People on the matching accounts to count buyers by title.

Test on 20 to 50 rows first. Clay's trial FAQ recommends running small batches while you learn.

What will an experiment cost?

The expensive parts are enrichments and AI runs. Each enrichment is one Action plus Data Credits that vary by data type, and each AI run is one Action plus credits that depend on the model (credits guide). Clay shows cost estimates before you run Actions (usage FAQ), and failed lookups do not consume credits (FAQ).

Free includes 500 Actions and 100 Data Credits a month, with tables up to 200 rows. Launch starts at $185/mo billed monthly ($167/mo billed annually) and Growth at $495/mo billed monthly ($446/mo billed annually) (pricing). Growth adds CRM auto-sync, useful if your control group should come straight from the CRM instead of a CSV.

How do you avoid fooling yourself?

• Write the success threshold before you look at the data.
• Spot-check the AI column on 10 rows you know well. Claygent provides a reasoning trace you can read.
• Keep the experiment in its own workbook so production tables stay clean.
• Save a working sequence of columns as a Function if the idea graduates, so other tables reuse the same logic.

FAQ

Can I test without touching the CRM?

Yes. Import exported accounts as a CSV for your control group. CRM auto-sync is available on Growth and above if you want live data later.

What if the attribute I need is not in any database?

Use a Claygent column to research it on the web, with a strict output format. Clay describes Claygent as finding unique data no provider has, such as niche company attributes and buying triggers.

Will an experiment affect live workflows?

Not if you keep it in a separate table, or use sandbox mode on an existing one. In sandbox mode the production table becomes read-only and outbound actions such as exports are disabled.

Do I need SQL or code?

No. Filters, the Use AI Generate tab and Sculptor accept plain English, and the formula generator helps with scoring logic.

Test your next targeting idea in Clay

Write the hypothesis in one sentence, build the matching list with Find Companies, add one Claygent question and compare against your closed-won accounts. If the idea holds, save the columns as a Function and hand it to the team.