How Clay turned its most dangerous customer into its distribution

Read this if you sell into GTM and you are still deciding what to call yourself.

Clay went from roughly $1M to roughly $100M ARR in about two years.¹ The AI SDR tools it is often lumped in with churn 50 to 70 percent of their customers.² Clay refused that label. This is a study of why, and most of the answer is in the words Clay uses to describe itself.

What you get:

• The label Clay refused, and the vocabulary it created instead
• How Clay's most capable and most dangerous customers, GTM engineers, became its best distribution channel
• The 12 terms Clay's message rests on, and the one that means nothing to buyers
• Why nobody in the room secretly wants Clay to fail, and why competitors carry churn Clay does not
• Two places Clay is selling itself short right now
• Three moves in the next 90 days that put Clay's name on the category before the platforms take the words

Everything here comes from public sources.

How we scored Clay

We look at five areas of a business.

• Does what Clay says match what Clay built?
• Is Clay selling to the right buyer?
• Do Clay's words sound like the words its buyers use?
• Does the price sit on the right unit?
• Who else is competing for the same buyer?

Each area gets a score out of 10. A 10 means we found nothing wrong. The lower the score, the deeper the problem: high scores call for copy fixes, low scores mean the position needs a strategy change.

Each score also carries a confidence rating: high, medium or low. That rating is about evidence, not opinion. High means the public record settles it. Medium means the evidence points one way but has gaps. Low means we could not verify enough, and we say so instead of guessing.

The five areas are not equal. Who you sell to counts twice as much as how you sound. We weight the five scores and add them up to one overall score out of 100, where higher is better. Above 60 means the position is sound: what needs fixing is the wording, not the strategy.

Where the numbers come from:

• Public sources only: clay.com, its pricing and customer pages, and dated market and buyer records.
• Every benchmark carries the date it was taken.
• We scored Clay several times and the numbers came out the same each time. That is what "held stable across repeated scoring runs" means in the small print.
• A person reviewed every score and every sentence before publication.

Clay scores 71 out of 100. Each section below carries its own area score, so you can see exactly where the 71 comes from. All figures are from Q3 2026.

What everyone else in the category did

Around 2023, this category agreed on a label, and enrichment tools, sequencers and data plumbing all started selling themselves as AI SDRs.

The promise was a rep that works on its own: feed it a list, it finds the contacts, writes the emails and books the meetings.

The reality was that buyers lacked the infrastructure to make any of that work. Someone still had to design the workflow, fix the data and keep it running. Buyers thought they were hiring a colleague and they were buying a tool that needed an operator.

That gap is the most likely source of the 50% to 70% churn:² teams cancel in month four when nobody has booked anything.

It also broke how buyers shop. The homepage no longer told you whether a company sold enrichment, sequencing or agents. Everyone used the same sentence and that sentence stopped meaning anything, so every evaluation had to start with a demo.

Then the platforms arrived. Salesforce and HubSpot now ship native enrichment and hand out Agentforce credits free to customers who already pay them.³

Long story short: if your label describes a feature, a platform can bundle it and give it away. If you create a category and the vocabulary that goes with it, they have to argue on your terms.

The label Clay rejected

Clay rejected the AI SDR label, and in 2024 that looked like throwing away the easiest marketing story on the table.

Clay did the harder thing and named a category instead. Orchestration is the term it is publicly credited with defining, and GTM development environment is how it describes the thing it built. Neither is trademarked and neither needs to be, because both came from Clay and everyone in the market knows it.

Here is what Clay actually does:

• Runs waterfall enrichment across 200+ providers,⁴ so if one fails to find a contact the next one tries
• Gives teams a surface to build their own GTM workflows on

Here is what Clay says it is:

• A GTM development environment
• The company that made "orchestration" a real term in this category

Those two lists say the same thing, which sounds like a low bar until you notice that almost nobody in this category clears it.

"Development environment" is also what engineers call the place they work, so when Clay put it on a GTM tool, engineers read it and understood it was built for them.

The users who could have built it themselves

Clay's core user is the GTM engineer, the person who wires the stack together and makes the data usable. They are the most capable users in this category, and they can also build the whole thing themselves with Clay, n8n and Smartlead, which plenty of them have.⁵

Two facts about them:

• They can sign off around $2K alone, or $10K with a manager⁶
• Their real alternative to buying is building

Most vendors see that budget and go looking for a bigger one somewhere else.

Clay went the other way and shipped for that user specifically:

• A free plan and self-serve signup, so nobody needs approval to start
• Claybook, the public community, where people learn Clay from each other's work
• Claygents and Sculptor, which let you wire in agents and describe a workflow in plain language instead of coding it

GTM engineers could have built a competitor. They build on Clay instead, then take Clay with them to the next job, where the budgets are bigger.

How that turns into revenue

The GTM engineer may not be the one signing the big deal, but they create the reason for it. The sequence looks like this:

1. An engineer starts on the free plan and builds a few workflows
2. Those workflows go into production and start feeding routing and scoring
3. The team now depends on them, and credit usage grows past what one person can expense
4. The renewal lands with the Head of RevOps, who owns both the data quality problem and the budget. That last step is the whole business.

We score buyers on four things: how much the problem hurts, how much they can spend, how badly served they are today, and how easily they can prove a result. Each is scored 1 to 10 and the four multiply, so the ceiling is 10,000 and one weak number sinks a buyer. The Head of RevOps scores 1,470 against the GTM engineer's 576.

Budget figures come from our European buyer benchmarks, stated here in dollars at round-number precision.

Clay is already making that climb, and it is making it by accident of product-led growth rather than by design, which is the second of the two gaps at the end of this study.

What would settle this: Clay's own funnel records from free engineer signup to RevOps-owned renewal, with the conversion and expansion numbers. That is internal evidence public data cannot reach; a paid eval is built on the founder supplying exactly this.

Something else is hidden in plain sight

Most tools in this category promise to replace your sales team, and that promise puts someone in the room with a private reason to make the deal fail, because nobody helps buy the thing that deletes their own headcount.

Clay's buyers get more capable and more visible at work, so nobody in the evaluation is quietly hoping it goes away. That is a large part of why Clay does not carry the churn its neighbors do.

Clay speaks its buyers' language, because it wrote a lot of it

We take the terms a company's copy rests on and check each one against how its buyers actually talk, and there are three possible results.

Native: the buyer already uses this term. Translatable: they do not use it, but they understand it immediately. Foreign: it means nothing to them.

Clay has 12 terms carrying the message.

Native:

• GTM development environment
• waterfall enrichment
• orchestration
• GTM engineer
• credits
• enriched and scored records
• sync to CRM

Translatable:

• signals and intent
• Claygents
• Sculptor
• Build systems to grow revenue

Foreign:

• Clay MCP, which is fine for the deepest builders and meaningless to a revenue buyer

Notice that Clay never says "replace your team", "no-code magic" or "black box", and never claims an unproven 10x. Those are the phrases buyers in this category have learned to distrust, and Clay's copy has none of them.

Clay's own customers also write better copy than its homepage does. One customer story on clay.com/customers describes the product as a way to test something at small scale, validate what works, then scale it as far as you want, which is sharper than the headline and cost Clay nothing.⁷

What would settle this: won-deal emails and call transcripts showing which of these terms buyers use back, unprompted. Internal evidence public data cannot reach; in a paid eval the founder supplies it.

The pricing is on the right meter

Clay charges on usage, with plans from roughly $185 to $495 a month at the published tiers, plus a free plan and custom enterprise deals.⁸ In March 2026 Clay split its single credit into two meters, one for data and one for actions, and stopped charging for lookups that fail.⁹ The bill still tracks what you use.

Why that is the right call:

• Per-seat pricing makes you a line item every time someone reviews the stack for consolidation
• Credits only rise when the customer is getting more out of the product
• RevOps leaders are measured on the quality and volume of their data, and Clay bills for the data and the work done on it, so the invoice and the job description move on the same number

What it costs Clay is forecasting, because usage is still harder to predict than a seat count, though the 2026 changes (no charge for failed lookups, cheaper data) took some of the sting out. The friction still shows up at renewal as a budget predictability question.

We cannot tell you whether Clay captures the right share of the value it creates, because no annual value figure is public and any ratio we gave you would be invented. What the market says instead is $100M ARR on this meter, before the 2026 changes,¹ which suggests the meter is not leaving obvious money behind.

What would settle this: renewal outcomes and how customers' credit forecasts compared with what they actually burned. Internal evidence public data cannot reach; in a paid eval the founder supplies it.

Correction, August 2026: this section originally described Clay's pre-March-2026 pricing, a single credit meter at roughly $134 to $880 a month. Clay split that model in March 2026; the description above was corrected in place in August 2026. Wording changes never move a score, so the case was re-scored against the corrected pricing and the August benchmark database: 71 out of 100, up from the July 68, with the price area moving from 6 to 7. Same band, same verdict. Reviewed and signed before publication.

The fight Clay has not finished

Three kinds of competitor, in three different states.

1. Vendors in the same fight: ZoomInfo and Apollo. ZoomInfo is losing on the structure of its data. It sells access to one database that it owns, so its coverage stops wherever that database stops. Clay queries 200+ providers in sequence, so when one source misses, the next one tries, and the hit rate comes out higher. ZoomInfo has been repositioning to catch up, and Clay's growth outran it.¹⁰ Apollo is winning a different game as the cheap all-in-one default for SMB, sold on price and bundling. It offers no surface to build on, so it never competes for the GTM engineer, which is where Clay's expansion motion starts.
2. The platforms the buyer already pays: Salesforce and HubSpot. The live threat, for the reason covered earlier: native enrichment in the box, free Agentforce credits for customers who already pay them. Clay has an answer: 200+ providers, the build surface, the community, and enough depth that Clay is now the incumbent hanging over every point tool's deal. Most companies in this position have no answer at all, which is why this is a watch item rather than a flag. It stays on the list because a depth advantage has to be re-earned every quarter while the platforms close the gap.
3. The buyer's own build. Handled, because Clay became the thing people build on.

What would settle this: win and loss records from deals where Salesforce or HubSpot bundled free credits were on the table. Internal evidence public data cannot reach; in a paid eval the founder supplies it.

The strongest case against this verdict

The best argument against Sound position and 71 comes from Clay's own March pricing change. When the split landed, the reaction in GTM communities was loud: one practitioner guide from the week of the change describes Facebook groups and Slack channels exploding with fear.¹¹

The sharpest version of the objection:

• Every workflow step now consumes an Action, even when you bring your own data keys, so platform work that used to ride along free sits on the meter.¹²
• That lands hardest on the heaviest builders, the same GTM engineers this study calls Clay's best distribution.
• If the meter can move under its most loyal users once, it can move again. And the cheaper data gets, the easier the argument for building your stack around Clay instead of on it.

Our answer, in the same terms we score with:

• The objection is real, and it is already priced in. The competition area sits at 6 out of 10 with medium confidence, not higher, precisely because a depth advantage has to be re-earned every quarter.
• The change's direction of travel favored buyers: failed lookups became free, data prices dropped, and Clay publicly accepted an expected revenue hit of around ten percent to make it happen.¹² That is the opposite of quiet extraction.
• The commentary we can cite settled within weeks into guides on how to adapt, not exits.¹¹
• The version that would move the verdict, a second meter change landing on builders with no offsetting give, is exactly what the watch signals under the moves below are there to catch.

Two places Clay is selling itself short

Both are copy rather than product, and both are entirely in Clay's hands.

1. The homepage hides the good line.

The headline reads "Build systems to grow revenue", and Apollo, ZoomInfo, Salesforce and forty other companies could publish that sentence tomorrow. "GTM development environment", the line nobody can copy, sits one click deeper, so the front door competes on the most crowded promise in GTM while the category Clay owns outright waits on page two.

2. The path from user to budget is accidental.

Nothing in the packaging tells a Head of RevOps why the tool their engineer loves should become a company-level system, so Clay gets there on luck rather than design.

Neither one is a repositioning, because Clay is already in the right place and only needs to say so where people can see it.

Three moves to own the category outright

Before the moves themselves, here is where they land.

In twelve months, "GTM development environment" is the phrase buyers use when they describe what they are shopping for, so Salesforce and HubSpot have to argue on Clay's terms instead of their own. The renewal conversation happens with a Head of RevOps who has already budgeted for an operating layer, rather than with an engineer expensing credits. And when a platform ships the next free enrichment bundle, Clay's customers already know why it is a different thing, because Clay told them a year earlier.

Clay does not need new capability to get there, and every move below is copy, packaging or content.

Move 1. Put the good line on the front door.

Make "GTM development environment" the headline and move "Build systems to grow revenue" underneath it as support.

What it takes: a copy change and no product work, and Clay's own activation and demo-request numbers would show inside a quarter whether it worked.

Watch: activation and demo requests over the following quarter, and whether buyers start arriving already saying the phrase. Not working looks like flat demo requests and the phrase appearing only in Clay's own copy.

Move 2. Answer Salesforce in public.

Clay has a real case against free bundled credits and keeps it in internal conversations, but a buyer comparing the two needs it written down in their own terms, on a page they can forward to their boss.

What it takes: content and sales enablement built on proof Clay already has, including the 200+ providers, the community, and customers such as OpenAI and Anthropic.⁷

Watch: whether the comparison page starts showing up in deals, forwarded by champions to their bosses. Not working looks like reps still building the argument by hand, and deals against bundled credits stalling at the same rate two quarters in.

Move 3. Design the path instead of inheriting it.

Name the company-level value that turns a $2K engineer's tool into a RevOps system, because that sentence is what moves a renewal from a personal card to a budget line.

What it takes: packaging and motion design on top of the free-to-enterprise structure Clay already has, done last because it needs moves 1 and 2 to carry the story.

Watch: renewals moving from personal cards to budget lines, with RevOps named on the order form. Not working looks like renewal conversations still landing on individual engineers a year from now.

Update, August 2026: this page gained sources and footnotes for every external figure, watch signals under the three moves, notes on what evidence would settle each medium-confidence area, the strongest case against our own verdict, and per-section check counts. None of it changes the measurement: the score and verdict stand as signed, and the score's own history is recorded in the pricing correction note above.

The GTM eval · engine cpa-system-v1-4 · snapshot Q3 2026. Built only from public sources: clay.com, its pricing and customer pages, plus dated market and buyer records reviewed by Charlotte Han in July 2026. Every figure carries the date it was taken, and scores held stable across repeated scoring runs. Anything only a founder could answer was left blank rather than estimated, which caps the confidence we claim and is marked wherever it happens. Reviewed and approved before publication.