Skip to content
← All posts

Tools

Clay for beginners: build your first enrichment table (step by step)

A beginner's Clay tutorial: core concepts, building a company-to-contact table, waterfall enrichment, AI research, conditional runs, cost control and exporting.

Victor Mendonça
Victor Mendonça

Founder, OretonGTM · ·6 min read

Clay has become one of the core tools in outbound and GTM engineering. It’s also intimidating at first: a blank table, hundreds of integrations and a pricing model built on credits.

The good news: you don’t need to learn all of Clay to get value from it. This tutorial walks through the core concepts and then builds one practical table step by step: from a list of companies to verified, qualified contacts ready for outreach.

What Clay is

Think of Clay as a spreadsheet where each column can do work. A column can:

  • Pull data from a provider (emails, phones, company data)
  • Run an AI prompt on the row’s data
  • Scrape a website
  • Calculate a formula
  • Call any API
  • Send data to another tool

Each row is usually a company or a person. You chain columns so data flows from left to right: company → qualification → people → emails → verification → CRM.

This is why Clay is often described as a data orchestration tool: it doesn’t just store data, it coordinates many sources and actions in one place.

Core concepts

Concept What it is
Workspace Your team’s account, with tables, integrations and credits
Table A list of rows (companies or people) with columns
Source Where rows come from: import, search, webhook, integration
Enrichment A column that pulls data from a provider
Waterfall An enrichment that tries several providers in order
AI column A column that runs a prompt for each row
Formula A column that calculates from other columns
Run conditions Rules that decide whether a column runs for a row
Credits What you spend when enrichments and actions run

Before you build: plan the table

Write down, on paper:

  1. Input: where do companies come from?
  2. Qualification: what makes a company a fit?
  3. People: which titles do you want, and how many per company?
  4. Contact data: emails only, or phones too?
  5. Output: where do results go?

This takes ten minutes and saves a lot of credits. Base it on your ICP.

My take: The most expensive Clay mistake is building while running. Turn off auto-run, build the whole flow on five rows, check every column, then scale.

Step 1: Get companies into a table

Options:

  • Import a CSV of companies (name + domain at minimum)
  • Use Clay’s company search to find companies by filters
  • Pull from your CRM (e.g., target accounts in HubSpot)
  • Receive via webhook from another tool

For a first table, start with 20 to 50 companies. Make sure each has a domain; it’s the most reliable key for enrichment.

Step 2: Enrich company data

Add enrichment columns for the company data you need to qualify:

  • Headcount and industry
  • Location
  • Description and LinkedIn URL
  • Funding data, if relevant
  • Technologies, if relevant

Run them on your sample and check the results. If a provider returns poor data for your market, try another.

See firmographic data and technographic targeting.

Step 3: Qualify with AI

Databases can’t filter on everything. Use an AI column to answer specific questions from the company’s website.

  1. Add a column that scrapes the homepage (or uses an existing description).
  2. Add an AI column with a prompt like:
Based ONLY on the website text below, answer:
1. sells_b2b: yes / no / unknown
2. has_sales_team: yes / no / unknown
3. reason: one sentence with evidence from the text

Return JSON.

Website text:
{{homepage_text}}
  1. Add a formula column “qualified” that’s true when the answers match your criteria.

Check 20 results manually. Adjust the prompt where the model gets confused. See AI lead qualification and Clay AI columns.

Step 4: Find people at qualified companies

Now find contacts, only for qualified companies:

  1. Use a find people action that searches each company for specific titles.
  2. Set run conditions so it only runs when “qualified” is true.
  3. Define titles carefully, including variations (Head of Sales, VP Sales, Director of Sales…).
  4. Limit the number of people per company (two to four is typical; see contacts per account).

People usually go into a new table linked to the company table, one row per person.

Step 5: Waterfall for emails

In the people table, add a work email waterfall:

  1. Choose several email providers.
  2. Order them (e.g., cost-effective providers first).
  3. Include verification so only valid emails are accepted.

The waterfall tries each provider until it finds a verified email. See waterfall enrichment and email verification.

For phones, add a separate waterfall with a run condition (e.g., only for Tier A accounts), since phones cost more. See find direct dials.

Step 6: Add personalization inputs (optional)

If you plan to personalize:

  • Add a column that pulls recent job posts or news for the company.
  • Add an AI column that extracts relevant facts.
  • Add an AI column that writes an opening line with strict rules, returning “NONE” when nothing relevant exists.

See AI personalization in cold email and personalization prompts.

Step 7: Dedupe and send

Before sending anywhere:

  • Dedupe against your CRM (lookup by email or domain) to skip customers, open deals and existing contacts (dedupe lists).
  • Filter to rows with verified emails.

Then send:

  • To your sequencer (Instantly, Smartlead and others) with the right campaign and custom fields.
  • To your CRM (HubSpot, Salesforce) to create or update contacts and companies with source, segment and signal fields.

Step 8: Run at scale

Once every column works on your sample:

  1. Import or source the full list.
  2. Run columns step by step, checking results after each.
  3. Watch credit usage.
  4. Spot-check outputs, especially AI columns.

Controlling credits

Credits go fast if you’re not careful:

  • Turn off auto-run while building.
  • Test on small samples first.
  • Qualify companies before finding people. Biggest savings.
  • Use run conditions for every expensive step.
  • Dedupe before enriching.
  • Avoid re-running columns on rows that already have results.
  • Use your own API keys where it makes sense.

More in Clay credits and cost control.

A complete example

Goal: heads of sales at B2B SaaS companies with 50–200 employees in the US, with verified emails, sent to an outbound sequence.

Column Type Runs when
Company name, domain Source (search) —
Headcount, industry, location Enrichment Always
Homepage text Scrape Headcount 50–200
B2B + sales team check AI Homepage text exists
Qualified Formula —
Find people (sales leaders) Action Qualified = true
Work email Waterfall + verification Person found
Recent job posts Enrichment Email valid
Opening line AI Job posts exist
CRM lookup Integration Email valid
Send to sequencer Integration Email valid and not in CRM

Beyond the first table

Once you’re comfortable, Clay can do much more:

Common beginner mistakes

  1. Running everything on the full list immediately.
  2. Finding people before qualifying companies.
  3. No run conditions on expensive steps.
  4. Vague AI prompts that return essays instead of data.
  5. Skipping verification.
  6. Not deduping against the CRM.
  7. Building one giant table instead of linked company and people tables.

Where to start

Take 20 companies you know are a good fit, build the table above step by step, and check every column. By the time you’ve sent those contacts to a sequence, you’ll understand Clay’s core workflow. If you’re deciding between Clay and an all-in-one database, see Clay vs Apollo.

Frequently asked questions

What is Clay used for?

Clay is used to build, enrich and qualify lead lists. It connects to many data providers, runs AI research on each row, and sends results to CRMs, sequencers and other tools, which makes it a central tool for outbound and GTM engineering.

Is Clay hard to learn?

The basics are approachable if you're comfortable with spreadsheets. The learning curve comes with waterfalls, AI prompts, conditional logic, formulas and integrations. Starting with a small, simple table helps a lot.

How much does Clay cost?

Clay has a free tier and paid plans, and charges for data and actions through credits. Costs depend on how many rows you enrich and which providers you use. Check Clay's current pricing page for details.

Do I need to bring my own data provider accounts to Clay?

Not necessarily. Clay gives access to many providers through its own credits. For some providers you can connect your own API keys, which can change the cost structure.

Can Clay send leads to my CRM or sequencer?

Yes. Clay integrates with CRMs like HubSpot and Salesforce and with many sequencers, and can also send data anywhere through HTTP API calls and webhooks.

How do I avoid wasting Clay credits?

Test on small samples, qualify companies before finding people, use conditional runs so expensive enrichments only run on qualifying rows, dedupe before enriching and turn off auto-run while building.

What's the difference between Clay and Apollo?

Apollo is mainly an all-in-one database with built-in sequencing. Clay is an orchestration layer that combines many data sources, AI and integrations in flexible workflows. Many teams use Apollo as one of the data sources inside Clay.

Ready to stop doing outbound by hand?

Book a strategy call. We’ll map your ICP, review what’s working (and what isn’t) and show you exactly what we’d automate first.

Book a strategy call

Keep reading