Waterfall Enrichment: Stacking Data Providers Without Paying Twice

Waterfall enrichment is the practice of querying data providers in sequence rather than picking one, passing each provider only the records the previous pass failed to resolve. It exists because no vendor covers your list, and the gaps are not random. They are structurally different by provider.

This is for anyone who has exported 5,000 contacts from a single provider, found 2,600 emails, and assumed the other 2,400 people do not have email addresses.

Coverage Is Not a Vendor Quality Score

Providers build their data differently, so they miss differently. One is strong on US mid-market tech because that is where its contributor network lives. Another is strong on European enterprise because of a licensing deal. A third is strong on anyone who has changed jobs recently because it watches profile updates.

This means running provider B over provider A's misses returns far more than provider B's headline coverage would suggest. You are not sampling randomly from B's database; you are sampling from exactly the population where A is weak, and B's weaknesses are somewhere else.

Realistic shape for a mid-market US B2B list: first pass resolves roughly half, second pass adds around twenty points, third adds ten more, and verification removes a few. Call it 85 to 90% coverage where one provider gave you 55%. The exact numbers move with your ICP, but the curve is consistent, each pass adds less than the one before.

The Third Pass Is Where Money Goes to Die

That diminishing curve is the whole economic story. The first provider resolves half your list at its standard rate. The fourth provider is being asked to solve the hardest 10%, charged per attempt, and succeeding on a fraction of them.

So the effective cost per incremental contact climbs steeply, often ten times by the fourth pass. Most credit models charge for the attempt, not the hit, which means you are paying the most for the pass with the worst hit rate.

The correct stopping point is not "maximum coverage." It is the pass where cost per incremental contact exceeds what a contact is worth to you. Compute that number once (take your average deal value, multiply by the rate at which a sequenced contact becomes a deal) and stop the waterfall there. For most teams that lands at two or three passes, not six.

Verify at the End, Not in the Middle

Providers return guesses. Many return pattern-derived addresses, first.last at the company domain, flagged with a confidence score that is, in practice, a statement about the pattern rather than about the person.

So verification is a separate step, run once at the end against the merged set, not per provider. Running it per pass means paying to verify the same address twice when two providers return it, which they often do.

Drop anything that comes back invalid. Segment catch-all domains separately or exclude them from cold campaigns entirely: they accept everything, so they tell you nothing while contributing to complaint rates. Keeping hard bounces under 2% is the whole reason this step exists, and it is the load-bearing connection between enrichment and deliverability.

Data Rots Faster Than People Expect

B2B contact data decays at roughly 2 to 3% a month through job changes, title changes and domain migrations. Over a year that is a quarter to a third of your list.

Two consequences. Verify immediately before sending, not at import, a list enriched in March and sent in September is meaningfully rotten. And treat a job change as a buying signal rather than a data problem: someone who just moved into a role has budget, a mandate, and no incumbent loyalty. That is the single highest-converting trigger in B2B outbound, and the enrichment layer is where you detect it.

Mobile Numbers Are a Different Market

Direct dials and mobile numbers have far lower coverage than email (typically 20 to 40% on a mid-market list) and cost several times more per record. They also carry compliance weight that email does not.

So do not run phone enrichment across your whole list. Enrich phone only for accounts that have already engaged, or for a tier where the deal size justifies it. Blanket phone enrichment is the most common way teams triple their data bill for a channel they then do not staff.

Build or Buy the Waterfall

Running this yourself means contracts with three or four providers, each with its own minimum commitment, plus the orchestration to route misses between them and merge results without creating duplicates. The minimums are usually what kills it: four providers with annual floors is a large fixed cost to solve a variable problem.

This is what our enrichment waterfall inside grobot Command Center is for: the provider relationships and routing sit underneath, and you spend credits against results rather than signing four contracts. Credits meter enrichments, verifications and AI actions; sends are included in the license and never metered.

It is worth saying what this does not do: it does not conjure data that no provider has. If your ICP is 40-person manufacturers in rural markets, the waterfall improves your coverage and does not fix it, and no vendor claiming otherwise is being straight with you.

The Working Rules

• Order providers by hit rate on your ICP, not by their marketing claims. Test on 500 known records before committing.
• Stop when cost per incremental contact passes what a contact is worth to you.
• Verify once, at the end, immediately before send.
• Re-enrich quarterly, and treat job changes as triggers rather than corrections.
• Enrich phone selectively, never across the whole list.