In the 2026 Anymail Finder benchmark, a chain of providers found 87.1 % of addresses against 86.4 % for a single source, and got more of them wrong. The waterfall is usually sold on coverage. Below: the mechanism, how the order of providers is chosen, what a cascade costs compared with parallel lookups, and the limit most descriptions leave out: a longer chain finds more contacts, not more correct ones.
What is waterfall enrichment?
Waterfall enrichment is a fallback chain. For each contact, the system asks provider one. If provider one returns a valid result, the chain stops. If not, it asks provider two, then provider three, until one succeeds or the list runs out. Clay's guide describes it as "a fallback chain, not a single lookup", and its course material notes that the sequence stops at the first valid result.
The reason it exists is simple. Every provider has gaps. Each one builds its data from different sources, so the people one cannot find, another often can. Chaining them fills the gaps without paying all of them for every contact.
The same method applies to any field: work emails, direct phone numbers, company data. Emails and phone numbers are where it is used most, because they are where single-provider gaps are largest.
How does the cascade decide the order?
The order is a trade-off between hit rate, accuracy and price per lookup. A common rule is to put the cheapest provider with acceptable accuracy first, so most contacts resolve at low cost, and the more expensive, deeper providers last, so they only see the hard cases.
- Accuracy first: a provider that often returns wrong data should not be early in the chain, because the chain stops at its answer and never asks the better provider.
- Then price: among providers with similar accuracy, the cheaper one goes first.
- Then segment: the best order differs by region, company size and seniority. A provider strong on European executives may be weak on American engineers.
- Always end with verification: an email found by any provider is checked before it is used. The verification step is what makes "first valid result" mean valid.
The first point is the one that hurts. A waterfall trusts the first answer it gets. A confident wrong answer early in the chain blocks every right answer after it.
There is no standard for waterfall enrichment; the definitions come from vendors. Clay's guide and course lesson describe the category: a fallback chain that stops at the first valid result. The Anymail Finder benchmark (2026) is the clearest public test of what the chain does to coverage and accuracy.
What does a waterfall improve, and what not?
A waterfall can only improve coverage, the share of contacts for which you get any result. It cannot improve accuracy, the share of those results that are correct, and in the only published benchmark it barely moved coverage either. The 2026 Anymail Finder email finder benchmark shows the difference.
| Approach | Coverage | Accuracy |
|---|---|---|
| Waterfall (several providers chained) | 87.1% | 95.7% |
| Single source | 86.4% | 98.9% |
Anymail Finder, Email Finder Benchmark 2026. Published by an email finder vendor that sells a single-source product; the test's definition of a waterfall is not published. Treat as one benchmark, not a market average.
In that test, the chain found 0.7 points more addresses and got 3.2 points more of them wrong. Every extra provider adds its own error rate to the pool. Coverage and accuracy are two different scales, and a cascade moves only the first one.
The practical consequence is that a waterfall needs a verification step at the end, not just at the start. Judge a cascade on two numbers: how many contacts it returns, and how many of those bounce when you send. A chain that adds a few points of coverage and a few points of bounces has not helped your sending domains.
What does it cost compared with parallel lookups?
A waterfall costs less than asking every provider at once, because most contacts stop at the first or second step. With three providers queried in parallel, every contact triggers three lookups. With a waterfall, a contact found by provider one triggers one lookup, and only the hard cases reach provider three.
| Parallel lookups | Waterfall | Single provider | |
|---|---|---|---|
| Lookups per contact | Always all providers | One up to all, stops at first hit | One |
| Coverage | Highest | Close to parallel | Lowest |
| Accuracy | Needs a rule to pick between answers | First answer wins | That provider's accuracy |
| Cost driver | Number of providers | Hit rate of early providers | That provider's price |
The real cost depends on how each provider bills. Some charge per attempt, so a miss still costs money. Others charge only for a result; Clay's guide describes a model in which only the hit is billed. Before you compare cascades, ask which of the two applies at each step.
When is one good provider enough?
One provider is enough when it already cascades internally, when accuracy matters more than coverage, or when your targets are easy to find.
The hit rate of a cascade is bounded by who is findable at all. Phone numbers of engineers and IT staff are rarely published, so even a long chain finds a minority of them, while senior managers are found by most of the chain's early steps.
- The provider cascades itself: some premium providers already query many sources and resolve the result. Wrapping a second waterfall around them adds cost, not coverage.
- Accuracy is the constraint: if bounces hurt your domains more than missing contacts hurt your pipeline, a single accurate source is the safer choice.
- Senior, visible targets: senior management in active industries is found around 70% of the time, against around 30% for individual engineers and IT specialists. For senior targets, the gaps a cascade would fill are small.
- Your targets are senior and visible: the ceiling is set by the role, not by the number of providers. Leadspicker's phone cascade across roughly 21 providers reaches around 70% of senior managers in active industries and around 30% of individual engineers and IT specialists. Below that ceiling, adding providers adds cost faster than contacts.
Leadspicker tested around 30 data providers for phone numbers and standardised on one premium provider that cascades the lookup across roughly 21 further providers, because it worked out considerably cheaper than cascading providers itself.
For phone numbers, Leadspicker uses one premium provider that cascades the lookup across roughly 21 further data providers, chosen after testing around 30. For email addresses, Leadspicker draws on several third-party providers in a chain. The provider names and their order are listed on the pricing page.
Frequently asked questions
Does a waterfall find many more contacts than one provider?
Only a little in the one published test. In the 2026 Anymail Finder benchmark the waterfall reached 87.1% coverage against 86.4% for a single source, with lower accuracy (95.7% against 98.9%). Each provider has different gaps, so the gain depends on your segment.
Do you pay for every provider in the cascade?
Only for the providers the contact reaches, and only if they bill per attempt. Providers that bill per result charge nothing for a miss. Check the billing model at each step before you compare costs.
When is one good provider enough?
When the provider already cascades across many sources itself, when accuracy matters more than coverage, or when your targets are senior and publicly visible. In those cases a second chain adds cost but little coverage.
