How to Build Your First Clay Table, Step by Step

Clay is a spreadsheet-shaped workspace for go-to-market data. A row is a company or a person, a column is a field, and the interesting part is that a column can go and fetch its own value from a data provider or an AI prompt rather than sitting there holding whatever you typed.

That is genuinely powerful, and it is also why the first table trips people up. A spreadsheet you fill in by hand costs nothing to get wrong. A spreadsheet that runs 200 lookups the moment you paste a list has a bill attached.

So this walkthrough goes in the order that protects you. Source first, then a small test, then the enrichment columns, then the full run. It is the order Clay's own credit conservation guidance points at, and it is the order every experienced Clay user seems to arrive at eventually.

What is a source in Clay, and why does it come first?

A source is how rows get into your table, and Clay's documentation calls sources the foundation of how data flows in. Nothing else in the table works until you have chosen one, because every enrichment column reads from the fields your source produced.

Clay's docs name several source types. There are the built-in list builders, Find Companies and Find People. There are CRM connections for HubSpot and Salesforce. There are webhooks for receiving records from external systems in real time. And there is plain CSV import.

• Start with CSV if you already have a list you trust, from an event, a partner, or an export from another tool.
• Start with Find Companies if you know the shape of the account but not the names, and you want to work top down from firmographics.
• Start with Find People if you know the job titles you want and you want Clay to go find who holds them.
• Start with a CRM connection if the job is to clean or extend records you already own. Clay's docs note CRM sources need columns configured to match what the source expects.

How do you import a CSV into Clay?

Clay documents two routes, one for a brand new table and one for a table that already exists. Both end at the same place: a column mapping screen and a choice about whether to run the rows immediately.

1. For a new table, create from CSV — From your workspace home, choose Create new and search for CSV. Upload the file with Browse Files or drag and drop it in.
2. For an existing table, use Tools then Import — Open the table, click Tools, and select Import. Upload the same way. Clay then asks whether to add to the current table, which appends rows, or replace the current table, which overwrites them. Read that choice twice.
3. Map your columns — Match each column in your CSV to a field in the Clay table. This is where a messy export gets tidied, so do not rush it. A domain column mapped to a name field will quietly poison every enrichment downstream.
4. Choose whether to run — Clay offers Save and run rows in this CSV, or Save and don't run. On your first table, choose not to run. You want to look at the imported data before anything starts fetching against it.

How do you build a list with Find People?

Find People is Clay's built-in prospect search. According to Clay's docs you start it by clicking Add at the bottom of a workbook and searching for Find People, then you narrow the search with filters until the result count looks like a list you would actually work.

The filter set is wide. Clay's documentation describes company attributes such as size, industry, and description keywords, job title filters with different matching modes including must contain exact and must match exactly, experience duration and keywords, geography with both include and exclude, profile characteristics such as names and follower counts, certifications, languages, education, and the ability to point the search at specific companies held in another Clay table or a custom list.

There are documented ceilings worth knowing before you plan a campaign around it. Clay states you can set a maximum of up to 50,000 records per search, up to 100 people returned per company, and up to 500 results per cell if you reduce the data you request. Exclusions cover up to 300,000 people across three exclusion sets, and those exclusions work from LinkedIn URLs.

Spend your time here rather than downstream. Filtering is free and enrichment is not, so every row you keep out of the table is money you never spend. Clay's own credit guidance says the same thing in plainer words: narrow the dataset to the records you actually care about before you enrich anything.

How do you add an enrichment column?

In a Clay table you click Add enrichment and search for the one you want, then pick the integration and the account it should run under. Clay's docs describe enrichments as columns that transform your data by pulling extra information from outside sources, with examples like verifying an email, gathering company details, or finding a social profile.

Two settings on that column matter more than the enrichment you picked.

• Only run if. Conditional logic that controls when the enrichment fires. Clay's docs note you can click Use AI to have it generate the conditional formula for you. This is how you stop paying to re-enrich rows that already have the answer.
• Auto-update. When it is on, Clay's docs say the enrichment re-runs whenever its input values change. Useful on a live table you maintain, expensive on a table you are still building, because every fix you make to an input can trigger another run.

Test on ten rows before you run three thousand

Clay recommends running an enrichment on roughly ten rows first, checking the output, and only then letting it loose on the column. That single habit is the difference between a cheap mistake and an expensive one, because a mis-mapped input does not fail loudly, it just returns nothing across the whole table while the meter runs.

Look at three things in your sample. Did the enrichment return anything at all? Is what it returned actually correct for those specific people, checked by eye against a real profile? And is the format usable, or will you need another column to clean it up?

The order that saves the most money

Do your company-level work before your person-level work. If twenty contacts share one employer, researching that employer once at the company level costs one lookup, and researching it twenty times because you did it on a people table costs twenty. This is the most repeated piece of practical advice in the community tutorials, and it holds up.

So the shape of a well-built first table is: build the account list, enrich the accounts, filter down to accounts that survived the filter, and only then find people at the survivors. You end up enriching a fraction of the contacts and you throw away almost nothing you paid for.

Gotchas the docs do not lead with

A few things are easy to learn the hard way. None of them are flaws in Clay, they are just consequences of a tool that is powerful enough to do a lot of work without asking.

• Replace current table is destructive. On CSV import into an existing table, the replace option overwrites rows. Append is the safe default.
• Empty is not the same as wrong. An enrichment that returns nothing looks identical to one you configured incorrectly. Always check a sample against a source you can see with your own eyes.
• Your table gets wide fast. Waterfall enrichments in particular add provider columns. Many teams adopt a colour convention, one colour for source data, one for enrichment output, one for AI logic, one for export-ready fields, simply to stay oriented.
• Credits have two flavours. Clay separates Actions, which measure platform work, from Data Credits, which pay for the data itself. A single enrichment typically uses one Action plus a variable number of Data Credits.

None of this makes Clay a bad tool. It is a very good one, and the people who get the most out of it treat it like what it is, a data engineering surface for go-to-market work. The question worth asking before you invest a week in it is simply whether you want to own that surface.

Table you architect, or employee you brief

Clay hands you the building blocks. You decide the columns, the order, the conditions, the providers, and the thresholds, and when the campaign changes you go back and rewire it. For teams with a dedicated operator who enjoys that work, the ceiling is very high.

Sistava starts from the other end. You hire an AI Employee for a role, you describe the outcome you want in a sentence or two, and it works out the steps. When the campaign changes you tell it, the same way you would tell a colleague. Neither approach is universally right. If you want maximum control over every lookup, build the table. If you want the result without owning the machinery, brief the employee.

Sources

• Clay documentation, Sources
• Clay documentation, How to import your CSV into Clay
• Clay documentation, Find People in Clay
• Clay documentation, Enrichments
• Clay documentation, Actions and Data Credits
• Clay documentation, Ways to save Clay credits
• OneAway, How to Use Clay: The Complete Beginner Tutorial for B2B Sales Teams
• Outbound Masters, Clay Tutorial: Setup to Situation Targeting