A Clay waterfall is an ordered list of data providers for one field. Clay asks the first provider, keeps the answer if it comes back confident, and only passes the record down to the second provider if the first returned nothing. You are charged for the provider that returns the match, not for the ones that missed. Clay's own work email waterfall chains more than 100 email providers this way (Clay, Data waterfalls).
That is the whole mechanism. The official docs will tell you that much. What they will not tell you is which order to put the providers in, what the thing actually costs after Clay's March 2026 pricing change, and when it is the wrong tool. This page covers those three, from nine Clay builds for GTM teams over the last two years.
The short version
| Question | Answer |
|---|---|
| What is a waterfall? | Providers queried in sequence for one field, stopping at the first confident hit |
| What does it cost per record? | One data credit for the provider that hits. Misses are not billed (Clay) |
| How many providers can chain? | 100+ in the work email waterfall (Clay) |
| Does it actually find more? | Independent benchmarks put single-provider email fill at 50-60% and a three or four provider waterfall in the high eighties (FullEnrich, Phantombuster) |
| What does Clay cost now? | Launch $185/mo, Growth $495/mo since 11 March 2026. The old $149 / $349 / $800 tiers are closed to new customers (Databar) |
| Biggest cost lever? | The order of the providers, not which providers |
The find-rate figures are from independent waterfall benchmarks, not from us: FullEnrich reports roughly 80% from a sequence against 50-60% single source, and Phantombuster reports a three-tool chain reaching 94.2% coverage against 83.2-88.2% for the best single tool in the same test. The spread between those two studies is a fair warning that the number depends on the list, not on the technique.
Clay is a routing engine, not a contact database
Clay is not an enrichment provider. It is a spreadsheet that talks to a marketplace of providers and lets you stack them in sequence. Every team we have seen bleed credits was using Clay like a CRM: one giant table, every field enriched on every row, records kept forever as the system of record.
The most common failure is not a broken tool. It is a team that spends a full month of credit budget and ends up with a list where a large share of the emails are missing or bounce, because the routing was never designed. Clay does exactly what it is told. If the routing is dumb, the output is dumb.
Clay is a routing engine, not a contact database.
Treat it as a routing layer between your CRM and your enrichment vendors and the rest of the playbook follows. Treat it as storage and you will pay for rows that do nothing.
The three waterfalls that matter
Most teams build one giant waterfall and stuff every field into it. That is the wrong shape. You want three separate waterfalls, run in this order.
1. The company waterfall
The cheap one, and the one most teams skip. Before you try to find a contact, you want to know whether the company is worth enriching at all. Firmographics, then tech signals from something like BuiltWith, then your own ICP scoring step.
Run this first. In our builds it kills 40 to 60% of the input list before you ever touch the expensive email providers. That is where most of the savings live.
One correction worth making here, because a lot of older guides still recommend it. Clearbit is no longer a standalone option. HubSpot acquired it in 2023 and it now ships as Breeze Intelligence inside HubSpot, on HubSpot credits, with no standalone accounts and no supported path for teams on Salesforce, Pipedrive, or any non-HubSpot CRM (MarketBetter, Salesmotion). If you are not a HubSpot shop, pick a different first step.
2. The email waterfall
The expensive one. B2B email data is fragmented across sources that overlap inconsistently, which is exactly the condition a waterfall exists to solve. Five to seven providers in sequence, cheapest cost-per-hit first.
Our default order for US and EU B2B contacts, which you should treat as a starting point to test rather than a recommendation to copy:
- Findymail (cheap, good coverage on common commercial ICPs)
- Datagma (cheap, fills different gaps from Findymail)
- Prospeo (mid-tier, stronger on European data)
- Hunter.io (mid-tier, better on technical and developer personas)
- ContactOut (expensive, last resort for senior titles)
After the waterfall, run a verification step through NeverBounce or MillionVerifier. Never push an unverified email into a sequence. The deliverability damage from one batch of bounces outlasts the campaign by months, which is the whole subject of our cold email deliverability guide.
3. The phone waterfall
If you do not cold call, skip it. If you do, keep it separate from the email waterfall: Datagma, Cognism, and ContactOut is a reasonable three. Mobile numbers cost more per record and hit at a lower rate than emails, so a phone waterfall attached to the same run as email is a fast way to burn budget on records you will never dial.
The order matters more than the providers
We ran the same 5,000-contact list through the same five providers in four different orders. Between the cheapest ordering and the most expensive, the bill differed by 4.1x. Same providers, same list, near-identical final coverage. That is our own measurement on our own list rather than a published benchmark, so treat the ratio as an illustration of the size of the effect, not as a number to forecast with.
The rule is dumb but useful. Sort your waterfall by cost per hit, ascending. Cost per hit is not the same as cost per attempt. A cheap provider with a low hit rate can be more expensive in practice than a mid-tier provider with a high hit rate, because every level below inherits the records the level above could not resolve.
Two things make this hard to eyeball.
First, hit rates vary by ICP, not by provider quality in the abstract. A provider that is strong on US commercial personas can be weak on European engineering personas. Run a 500-contact pilot on a sample of your real target list before you commit to an order.
Second, provider data changes continuously. The waterfall that was cheapest in January will not necessarily be cheapest today. Re-test quarterly. This is the same discipline as any other CRM data enrichment programme: the data decays whether or not you look at it.
Find, enrich, then score
Enrichment is the middle step, not the last one. The step teams skip is scoring, and skipping it is why enriched lists sit unused.
Once the company and email waterfalls have run, add a scoring column before anything leaves Clay. It does not need to be clever. A weighted score across the firmographic fit signals you already collected, the tech signals from the company waterfall, and any behavioural signal you have access to is enough to sort the list into three bands: work now, nurture, discard.
Two reasons this belongs inside Clay rather than in the CRM. You already have every enriched field in the row, so the score costs nothing extra to compute. And a score computed before export means only the top band consumes sequence capacity and rep time, which is the actual constraint. The routing side of this, once records land in the CRM, is covered in our Clay enrichment and lead routing guide.
What Clay costs in 2026
Clay restructured its pricing on 11 March 2026, and most guides you will find still quote the old numbers. Here is the current shape (Databar, Cleanlist):
| Old structure | Current structure | |
|---|---|---|
| Self-serve tiers | Starter $149, Explorer $349, Pro $800 per month | Launch $185, Growth $495 per month, plus a free tier |
| Credits | Single credit pool | Two pools: Data Credits and Actions |
| Launch includes | n/a | 2,500 Data Credits, 15,000 Actions |
| Growth includes | n/a | 6,000 Data Credits, 40,000 Actions |
| Annual billing | n/a | 10% discount |
Three things about that change matter more than the headline prices.
The credit system split in two. Data Credits pay for marketplace data lookups. Actions pay for orchestration: HTTP API calls, workflow steps, CRM syncs, AI columns. A waterfall consumes both, and a table heavy on AI columns can exhaust Actions long before it exhausts Data Credits. Budget them separately.
Marketplace data got substantially cheaper. Clay cut data marketplace prices by 50 to 90% in the same change (Databar). If your cost model for Clay was built before March 2026, it is wrong in your favour, and it is worth rebuilding before you conclude Clay is too expensive.
Features moved down a tier. CRM integrations, HTTP API columns, and web intent moved from the old $800 Pro plan to the $495 Growth plan (Databar). If you were on Pro purely for the CRM push, Growth now covers it.
Existing customers on Starter, Explorer, or Pro were grandfathered indefinitely and were not forced to migrate, but the window to switch between the legacy tiers closed on 10 April 2026 (Cleanlist). New accounts get Launch or Growth.
What a verified contact actually costs
Nobody can give you a real number for this without your list, so here is the arithmetic with every input labelled. Substitute your own.
Assumptions:
- Growth plan at $495 per month, 6,000 Data Credits and 40,000 Actions included (Clay's published tier, above)
- 5,000 input contacts per month
- The company waterfall removes 45% before email enrichment, leaving 2,750 (our own pilots run 40 to 60%; 45% is the middle of that range, and it is an assumption, not a measurement of your list)
- An 80% find rate on the survivors, giving 2,200 found emails (the independent benchmark figure cited above, not a guarantee)
- Verification on all 2,200
On those assumptions the subscription alone works out at about $0.22 per verified contact. That is the floor, and it is the only part of the number anyone can compute for you from published prices.
Everything above the floor is credit consumption beyond the included 6,000, and that is entirely a function of your waterfall order and your input list quality. It is also the part no pricing page can answer, which is why "what does Clay cost" gets such unsatisfying answers. Run the 500-contact pilot and you will have your own number in an afternoon.
For scale comparison: ZoomInfo does not publish pricing, contracts are annual with a three-seat minimum, and Vendr's aggregated purchase data puts the median contract at $33,500 a year across roughly 1,570 recorded purchases (Vendr's ZoomInfo listing, as reported by 2026 pricing write-ups). An earlier snapshot of the same source, $31,875 across 1,313 purchases, is still widely quoted (Cleanlist), so two write-ups can disagree by about $1,600 a year and both be citing Vendr. Clay at Growth plus credits sits an order of magnitude below that for the volumes most Series A and B teams run, which is the honest reason it won the category. Where Clay loses is operator dependence: a badly built waterfall costs more for less data than a flat-rate contract, and nobody sells you the operator. We go through the full field in our ZoomInfo alternatives guide.
Per verified contact from the Growth subscription alone, on the stated assumptions. Credit consumption sits on top and is where a good waterfall pays for itself.
The five mistakes that burn credits
After auditing a dozen Clay setups, the same five show up every time.
1. Enriching the full list instead of the filtered list
If the input has 10,000 LinkedIn URLs and 3,000 match the ICP, do not enrich 10,000. Filter first, enrich second. It is obvious and almost nobody does it.
2. HTTP API columns where a native integration exists
Clay has native integrations for most marketplace providers. The native version is rate-limited, cached, and billed as data. An HTTP column does the same job as an Action and bypasses the caching you are paying for. Under the split credit model this is now a double cost: you spend Actions on work that Data Credits already covered.
3. No verification step
Unverified emails will damage sender reputation within a week. Verification is cheap per address and protects the entire outbound programme. Skip it once and you spend months on domain warmup instead.
4. Treating Clay as the system of record
Clay tables are workspace, not storage. Once a record is enriched, scored, and verified, push it to HubSpot, Attio, or Salesforce and archive the row. Teams holding hundreds of thousands of rows in Clay are paying to store data that decays.
5. No cost ceiling per row
Clay lets you cap credits per row. Most teams leave it off, and a single hard-to-find record then walks the entire waterfall. Cap it. If a contact takes every provider you own, they are probably outside your ICP anyway.
Wiring Clay into HubSpot or Attio
The integration story has improved, and it is still finicky. Four things that hold up.
Push, not two-way sync
Do not let Clay continuously sync with HubSpot. Two-way sync looks clean in a demo and creates a reconciliation mess inside a quarter. Pick one direction: Clay pushes new records when they are ready, the CRM pushes nothing back, and updates run as deliberate one-off imports.
Match on domain, not company name
Company names are the least reliable identifier in B2B data. "Acme Inc", "Acme, Inc." and "Acme Corporation" will produce three records if you match on name. Match on domain, and if the domain is missing, send the record back to the company waterfall rather than letting it through.
Use a staging property
Create a custom contact property, clay_status, with values like pending, enriched, verified, exported. Push from Clay as enriched, run verification, then flip to verified. Only verified records enter sequences. This single property has saved more outbound programmes than any other automation we have built.
Run the push through n8n when it gets custom
Clay's native HubSpot push is fine for the simple case. For anything with branching or partial failures, move the export step into n8n: proper error handling, retries, and a queue you can pause when something looks wrong. Clay to CRM is the most common breakpoint in a RevOps stack, so give it real error handling.
When Clay is the wrong tool
Every RevOps post about Clay reads like an ad, so here is the other side.
Clay is wrong if what you need is a CRM. It is a workspace. Records there decay like anywhere else, and there is no audit trail worth the name.
Clay is wrong under roughly 500 contacts a month. The subscription outweighs the routing savings until you have volume to route. Buy one provider directly.
Clay is wrong if nobody on the team thinks in spreadsheets. The product is powerful because you control every column, which is also why it collapses without an operator who enjoys that. Either hire someone who lives in tables or pick a simpler tool. This is the same argument we make about Clay as a GTM platform generally: the tool is not the differentiator, the operator is.
Clay is wrong where the providers have thin coverage. Waterfalls do not create data. If every provider in the marketplace is weak on your market, chaining ten of them returns ten misses. Pilot before you commit.
What we would do in the first week
Setting up Clay enrichment from zero, in order.
Day 1. Define the ICP in writing. Title patterns, size band, geography, and at least two disqualifiers. No Clay table yet.
Day 2. Pull a 500-contact pilot list. Build only the company waterfall. Confirm what share of rows survive the ICP filter and whether the survivors look right.
Day 3. Build the email waterfall on the survivors. Run it, verify, and record cost per verified contact and find rate per provider.
Day 4. Reorder the waterfall based on which providers actually paid off. Re-run on a second 500-contact pilot and compare the two bills.
Day 5. Add the scoring column, then wire the push to HubSpot or Attio through n8n with a clay_status property so nothing reaches a sequence unverified.
Week 2. Scale up and watch cost per verified contact every week. When it climbs, the cause is almost always list quality upstream rather than the waterfall itself.
This is the boring version, and it is the one that survives past month three.
Need a Clay table that actually pays for itself?
Book a 30-minute audit and we will rebuild your waterfall live. You keep the table. No retainer required.
Book an audit →FAQ
What is waterfall enrichment in Clay?
It is a sequence of data providers queried in order for a single field. Clay asks the first provider, takes the answer if it is confident, and only sends the record to the next provider if the previous one returned nothing. Credits are charged for the provider that returns the match, not for the ones that missed, so a record resolved on the first attempt costs one lookup (Clay).
Where are Clay's official waterfall docs?
Clay documents waterfalls in Clay University at university.clay.com and in its own waterfall enrichment guide. Those cover the mechanics and the interface. They do not cover provider ordering for a specific ICP, real cost per verified contact, or the cases where a waterfall is the wrong approach, which is what this page is for.
How much does Clay cost in 2026?
Self-serve is Launch at $185 per month with 2,500 Data Credits and 15,000 Actions, or Growth at $495 per month with 6,000 Data Credits and 40,000 Actions, with 10% off on annual billing. Legacy Starter, Explorer, and Pro plans at $149, $349, and $800 are grandfathered for existing customers but closed to new ones since the 11 March 2026 change (Databar).
How many data providers does a Clay waterfall use?
Clay's work email waterfall can chain more than 100 email providers in sequence, drawing from its credit-based marketplace (Clay). In practice, five to seven in the email waterfall is where the returns flatten, because the providers below that overlap heavily with the ones above.
Is waterfall enrichment actually more accurate than one provider?
It is more complete rather than more accurate. Independent benchmarks put single-provider email fill at 50-60% and a three or four provider waterfall in the high eighties, with one test reaching 94.2% coverage against 83.2-88.2% for the best single tool (FullEnrich, Phantombuster). Accuracy still comes from the verification step, not from the chain.
What is the cheapest way to get waterfall logic without Clay?
Several enrichment tools now ship multi-provider fallback natively at a lower entry price than Clay, and a single provider plus a verification step covers a lot of ground under 500 contacts a month. The trade you are making is control: Clay's value is that you choose the order and see the cost of every level, which is exactly what a bundled waterfall hides. Below the volume where ordering matters, the bundled option is usually the better buy.
How do you calculate cost per verified contact?
Take the total spend for the period, subscription plus provider credits plus verification, and divide by the number of contacts that came out verified, not by the number enriched. Enriched-but-unverified records are cost without output. Doing it this way is what makes two waterfall orderings comparable, and it is the only number worth tracking week to week.
Can Clay replace a CRM?
No. Clay is a workspace, not a system of record. Keep the CRM as the source of truth and push enriched records into it. Teams that try to live in Clay accumulate stale rows and lose any audit trail.
How long does it take to learn Clay?
A spreadsheet-comfortable operator can build a working waterfall in a week. Column logic, formula columns, and conditional waterfalls take about a month of daily use. The usual mistake is underestimating operator hours and ending up with a setup nobody can maintain, so budget a real RevOps person rather than an intern.