Waterfall enrichment chains data providers in a fixed order and stops the moment one returns a verified answer, so position one decides most of your bill. The saving comes from the order, not the number of providers. Here is the credit math, and where it sits in the rest of the data enrichment stack.
By Rahul Bageria, co-founder · Updated August 2026 · 14 min read
The word covers three different jobs with three different economics. Everything below is priced in credits, because plan prices move and credit costs do not.
| Waterfall type | Cost per hit | Coverage spread | What actually limits you | Build one? |
|---|---|---|---|---|
| Work email | 0.2 to 6 credits | 40.12% to 90.26% | Coverage. Quality is uniformly high, so you are shopping for reach | Yes. This is the case a waterfall was built for |
| Mobile number | 5 to 25 credits | 8.52% to 86.42% | Cost first, then coverage. Both are far worse than email | Only if one dial is worth twenty-five emails |
| Company attributes | 0 to 8 credits | 68.51% to 87.36% | Quality. The sources disagree with each other, not with reality | Rarely. Pick one source and move on |
| Verification | $0.002 to $0.008 | Not applicable | Catch-all domains, which no verifier can resolve | Always. Once, at the end, for everything |
Where these came from. Credit costs and spreads are read off Clay's published provider data tests, Global view, checked August 2026. That is vendor data, and Clay publishes no method for those tables. Verification prices are Bouncer's own published rates, checked August 2026.
The cascade is the easy half. The half that decides your results is the test the chain applies before it calls a record done.
A stopping rule is what a chain checks before it stops. Stop on first returned keeps whatever came back. Stop on first valid keeps only what passed a check. Vendors set this differently, and it decides what reaches your send.
FullEnrich documents the stricter version of the rule: it continues past a provider that returns an invalid email rather than stopping there, and flags landlines at no credit cost while it keeps hunting for a mobile (fullenrich.com, checked August 2026). Two chains that look identical on a diagram can behave completely differently here.
You never pay the providers a record skipped. Whether you pay for one that ran and missed is that provider's own policy, not the waterfall's.
Clay's FAQ states the pass-through plainly (checked August 2026): if a provider charges Clay for an API call, Clay charges you, and if the provider refunds Clay for failing to deliver valid data, Clay refunds you. No marketing page mentions it.
Clay publishes a separate provider table for each of these. Read them side by side and they are not variations on one problem, they are three problems.
| Waterfall type | Credits per hit | Coverage spread | Quality spread | The binding constraint |
|---|---|---|---|---|
| Work email | 0.2 to 6 | 40.12% to 90.26% | 81.78% to 97.15% | Coverage. Every credible provider is accurate, so a chain buys reach. |
| Mobile number | 5 to 25 | 8.52% to 86.42% | 87.43% to 92.50% | Cost and coverage together. The cheapest mobile hit costs 25x the cheapest email hit. |
| Company employee count | 0 to 8 | 68.51% to 87.36% | 68.88% to 84.84% | Quality. One provider sits at 0 credits and 86.72% coverage, and everyone disagrees. |
Vendor data, checked August 2026, from Clay's work email, mobile phone and employee count data tests, Global view. Clay publishes no methodology on those pages, so read them as a vendor measuring its own marketplace. Each test also ships North America, Europe, APAC and LATAM tabs, which admits the right order changes by region.
An email waterfall buys you reach. A mobile waterfall buys you reach at twenty-five to a hundred and twenty-five times the price. A company-data waterfall mostly buys you an argument between sources.
A worked example on Clay's published figures, not a forecast. What to copy is the shape of the result, which holds whatever coverage rate you go on to measure on your own list.
We are not printing a plan price, because plan prices move and only yours matters. Take your monthly bill, divide by the credits it includes, and multiply through. The ratios below hold at any rate.
Enrow 0.2 credits, Findymail 0.5, BetterContact 1.5, SMARTe 6, on Clay's Global work-email table, August 2026. Same job, thirty times the price.
Clay publishes Findymail at 90.26% coverage and 0.5 credits, measured on Clay's population rather than yours. Run 1,000 records through that and the arithmetic returns about 903 emails for 451 credits. Half a credit each.
Enrow at 70.94% and 0.2 credits clears about 709 records for 142 credits. Send the 291 leftovers to Findymail and, at its published rate, the same sum returns 972 emails for 274 credits.
Clay's rates come from Clay's population, and the 291 survivors are the hard records, where coverage always runs lower. Measure yours with a 200-row pass of your own list: our post on how many providers a chain needs walks that test through.
Half a credit per found email against roughly 0.28, on those published figures. Put your own measured rates in and both numbers move, but the gap does not close: the two-step chain is cheaper per result at every residual rate including zero, because step one costs 60% less per hit. What you are copying is the order, not the answer.
Every chain review I get asked for is about the tail. It is the wrong end of the chain. Position one runs on every record and position four runs on a handful. Fix position one and stop fiddling.
Not between a waterfall and a single provider. Between the chain you have and the chain with one more step in it.
Nobody actually faces this choice. You already have a provider. The decision in front of you is whether to put a cheaper one in front of it, or a rarer one behind it, and those are different decisions.
Ask it once per step, in order, and the answer changes as you go down. A provider that is obviously wrong at position one can be obviously right at position four, and the reverse is just as common.
The residual size cancels out of both sides of the comparison, so it drops out of the sum. What is left is simple: the step pays whenever a found record is worth more to you than that step's credits per hit.
Step four can be perfectly economic per hit and still pointless, because it is running on forty records. That is the real ceiling, and the full argument sits in how many providers a chain should actually have.
The category sells provider count because it is the number that fits on a badge. The arithmetic points somewhere else.
"Twenty-five sources in one call. More providers, higher find rate."
"Position one runs on 100% of records. Position four runs on the leftovers."
The two best vendor guides contradict each other. Clay's (14 April 2026) says put the cheapest trustworthy provider first, since early levels run on every record. Hunter's (20 July 2026) says start with the strictest finder even if it costs more, because forgiving providers win a bigger share of your list and their guesses come back as bounces.
Both are right, and the disagreement is only apparent. Clay's rule carries the word trustworthy; Hunter's objection is entirely about what happens when a cheap provider is not. Run them as two steps and they stop fighting.
Set a quality floor and delete everything under it. Then order what survives, cheapest first. A cheap provider below your floor is not a cheap provider. It is a bounce generator with a discount.
Want your enrichment chain built and run by the two of us?
Book a Fit CheckTwo work-email chains built from the same table. Same providers available, opposite results, and the only difference is the order.
SMARTe 6cr, then Surfe 2.5cr, then LeadMagic 0.3cr, then Enrow 0.2cr, each step's own verdict is final
Enrow 0.2cr, then Findymail 0.5cr, then Dropcontact 1cr, then BetterContact 1.5cr, verify once at the end
Floor: drop every provider under 90% quality on your own sample
Step 1 Enrow, 0.2 credits per hit, runs on all 1,000 records
Step 2 Findymail, 0.5 credits, runs on whatever Enrow missed
Step 3 Dropcontact, 1 credit, runs on what those two missed
Step 4 BetterContact, 1.5 credits, runs on the last few percent
Then one verifier, once, across everything all four returned
Credit costs from Clay's published Global work-email table, checked August 2026. Re-rank on your own segment before you run it.
Which tool you build it in is a separate question, and we keep that read on the ten waterfall enrichment tools worth running.
Every vendor sells one enrichment product. The credit sheet says it is two, and the second one is expensive.
Published range
5 to 25 credits per verified mobile
What moves the number
Read off, not estimated
Clay's Global mobile provider table plus each vendor's own pricing page, all checked August 2026.
Four ways a chain hurts you, and one that you cannot take back once it has happened.
Hunter's guide (20 July 2026) puts it plainly: inside a chain the forgiving providers claim a bigger share of your list, and the more eager the finder, the more guessed addresses reach your outreach. You meet them again as bounces.
Every employee-count provider in Clay's table sits between 68.88% and 84.84% quality, well under any credible email finder. Chaining more sources that disagree does not settle the disagreement, it gives you more of it to reconcile.
Published coverage is measured on a whole list. Step three sees only the records two providers already failed on. Nobody publishes that second number, and one run on your own data hands it to you free.
Hunter names the cases for skipping it: a bounce can hurt your sending domain, the list is small enough to fill by hand, or one provider already covers your segment. Measure your own coverage first, or you cannot tell which case you are in.
On an accept-all domain the mail server accepts messages to any address, real or not, so as Hunter states (20 July 2026) no verifier can fully confirm the mailbox exists. A forgiving provider calls that address valid anyway. A stop-on-first-valid chain stops right there, the address enters your main send, and the bounces land on a domain you cannot un-burn.
Be careful what you attach that to. Google's Gmail sender guidelines, in force for bulk senders since 1 February 2024, say keep the spam rate in Postmaster Tools below 0.10% and never reach 0.30% (checked August 2026). Those are spam complaint rates, not bounce rates. Google's bounce instruction is separate: cut sending volume when messages start bouncing.
Every page on this topic quotes a find rate. Read enough of them and the numbers stop agreeing, including inside a single publisher.
Across the pages ranking for this term, the single-provider baseline is quoted four incompatible ways, and the waterfall result four more. Not one names a dataset, a sample or a segment. We are not repeating any of them here.
Clay's 19 April 2026 test ranked Prospeo first. Clay's Global data table has Prospeo at 40.12% coverage, last of fifteen. Same publisher, same year, opposite verdicts.
That test ran 4,238 addresses from people who had manually replied to campaigns, all HR leaders at companies up to 10,000 staff, each validated through Debounce as non-catch-all. A narrow population, honestly declared, different answer.
A coverage figure without a stated population is not a measurement, it is a mood. Ask any vendor which segment produced the number. The good ones answer, and Clay does, on the page that carries the method.
We run enrichment on several of these tools every week, and we still do not publish a house find rate. It would carry exactly the same defect: true on our segments, meaningless on yours.
Clay publishes both of those results openly, which is the only reason anyone can see the gap. The vendors worth reading are the ones that leave enough on the table for you to catch them.
Co-founder of Real Good GTM. He has been the first business hire and Chief of Staff at seed-stage B2B startups, building outbound pipeline before any playbook existed. This post comes from building enrichment chains every week and watching the same ordering mistake cost people money at position one.
Connect on LinkedInThe two managed waterfalls head to head, the single sources that make up a chain, and the service where we run the whole thing.
Two managed waterfalls with near-identical credit mechanics, split on integrations, phone coverage and price.
Read the comparisonThe single sources a chain is assembled from, and what each one is actually good at.
See the toolsWhere we build the list, run the waterfall and verify it, on your ICP rather than a template.
See the serviceBook a fit check. We'll look at the coverage you actually have on your own list, put the providers in the right order for your segment, and tell you straight if a waterfall is not the thing holding your outbound back.
Book a Fit CheckNo hard sell. No fake numbers. Real good work speaks for itself.