Waterfall enrichment means asking several data providers for the same contact in sequence and stopping at the first verified result. It exists for one reason: no provider covers every market, and a single-provider setup loses whatever its supplier is weak at without ever telling you a gap was there.
Why one provider is never enough
Every B2B email provider is built on a particular acquisition strategy, and every acquisition strategy has a shape. One is strong on US tech because that is where its contributor network sits. Another is strong on European companies because it built a business around a data-protection position that made European buyers comfortable. A third is strong on large enterprises and thin below 50 employees.
None of them will tell you this, and not out of dishonesty. They publish an aggregate accuracy figure, which is an average across their whole database, and an average is precisely the statistic that hides which segments fail. A provider at 90% overall can be at 97% in its core segment and 55% in yours, and both numbers are inside the average.
The gap is invisible because a miss looks like an absence. The provider returns nothing, you conclude the person has no findable email, and you move on. Nobody tells you a different provider had it.
What the mechanism actually is
Simpler than the word suggests. Ours is about forty lines:
- Order the providers.
- Ask the first one for an email.
- If it comes back verified, stop and return it.
- If it comes back empty, unverified, or throws, treat that as a miss and go to the next rung.
- If no rung verifies, return nothing, and do not charge.
Point five is the one that matters commercially and it is worth stating separately: a provider exception is treated as a miss rather than an error. A vendor having a bad afternoon should cost you one rung, not your whole job.
Ordering is a cost decision and a compliance decision
The first rung answers most of the time, so whatever you put there sets your bill.
But the ordering rule we actually use is not about cost. Any provider whose coverage genuinely suits an input gets moved to the front for that input. In practice that means EU and EEA contacts go to Dropcontact first, and everyone else goes to Prospeo first, because Dropcontact's whole product position is built around European data protection and it is the more appropriate first port of call for a European person.
That is a data-protection decision that happens to also improve the hit rate. If you build your own waterfall, decide which of those two things is driving your order before you tune it, because optimising a fixed order purely on cost will eventually route data somewhere you would rather it did not go.
Diminishing returns are steeper than the marketing suggests
The intuition is that each rung adds coverage. The reality is that providers overlap heavily on the easy part of the market, so rungs two and three spend most of their calls confirming what rung one already found.
The useful mental model is that each rung only earns its place on the contacts the previous rungs missed, and misses are concentrated: hard-to-find people are hard for structural reasons that affect every provider similarly. Small company, no web presence, a name that collides, a domain that is a catch-all. Adding a fourth vendor licensing similar data does not help with any of those.
Two providers with genuinely different coverage beat six with overlapping coverage, and it is not close. When a vendor advertises a large number of waterfall sources, the useful question is which of them have coverage the others do not.
What a waterfall costs, and who pays for the misses
Here is the part that decides whether the model works for you rather than for your vendor.
Most providers charge per lookup, hit or miss. So a waterfall that walks three rungs and finds nothing has cost three payments and produced zero value. Somebody absorbs that.
| Model | Who pays for the miss | What it does to your incentives |
|---|---|---|
| Credits per lookup | You | You avoid hard segments, because searching them is expensive whether or not it works |
| Credits per verified result | The vendor | You search freely, and the vendor is motivated to keep the miss rate down |
The second model is the one we run, and it is worth being clear that it is a deliberate constraint on us rather than generosity. Because misses are ours to absorb, we meter enrichment attempts separately from credits, published on the fair use page. Credits track what you received; attempts track what we paid for. A vendor charging you for misses does not need that second number and generally does not publish one.
What "verified" has to mean for any of this to work
A waterfall stops at the first verified result, so the definition of verified is the whole control system. If a rung returns a pattern-guessed address and your waterfall accepts it as verified, the walk stops early, at the worst possible answer, every single time.
That is the failure that turns a waterfall into an expensive way to generate bounces. A rung must only halt the walk on a live deliverability check, not on a plausible-looking string. See what "verified email" should actually mean for how to test what your vendor means by the word.
Should you build your own?
Occasionally, and less often than the tooling market implies.
Build if you have real volume, direct contracts with two or more providers, and someone who will own coverage testing as a recurring job. The orchestration is genuinely easy. The maintenance is not: contracts, rate limits, changing response shapes, and the periodic re-measurement that tells you whether your rung order is still right.
Do not build if you are wiring together the same three vendors everyone else uses through a general-purpose automation tool. You get a system to maintain, the same coverage, and per-lookup pricing on every rung including the failures.
The honest test is whether you would still run it if nobody could see the diagram.
What this does not fix
Waterfalls improve the chance of finding an address. They do nothing about whether the address should be contacted, whether the account is worth contacting, or whether your domain can send to it without damaging your sender reputation. Those are three separate problems and enrichment solves none of them.
Related reading
- The bounce tax on what per-attempt pricing actually costs at real match rates.
- What "verified email" should actually mean on the definition the whole mechanism depends on.
- Verified email for the short version.