What Is Waterfall Enrichment? The Cascade, Explained

By Priya Raman, AI SEO, GEO & Martech Analytics. Last updated: 2026-10-22

Every enrichment vendor sells coverage. None of them has all of it. A provider that finds work emails for 70% of your list misses the other 30% in a pattern that is not random: it skews by company size, region, and how recently someone changed jobs. Waterfall enrichment exists because stacking providers in sequence covers more of the list than any one of them covers alone, as long as you stop paying once a record is resolved.

The term runs through almost every Clay and enrichment product page, usually as a feature checkbox. It is better understood as a budgeting discipline. This explainer covers how the cascade works, where it breaks, and how to judge one. The commercial companions live in Clay alternatives, which splits replacement tools by the job they do, and Clay pricing, which models what the cascade costs inside Clay's two-meter billing.

What is waterfall enrichment, in one paragraph?

Waterfall enrichment is a method of data enrichment where each record passes through an ordered chain of data sources until one of them returns the field you need. The chain runs per field, not per record: your email cascade and your phone cascade are different chains with different providers in a different order. Platforms like Clay built their product around this pattern, chaining more than 150 data providers behind one table, and specialist tools such as FullEnrich and Databar sell the cascade itself as the product. The alternative is a single database lookup, where one vendor either has the record or does not.

How does a waterfall actually run, step by step?

A production cascade has five stages, and the order is the whole game.

First, you clean the input. Names, domains, and company names get normalized, duplicates get merged, and records that fail basic sanity checks (no domain, a personal email where a work email is required) leave the queue. Every provider downstream charges for attempts or returns, so dirt is a cost multiplier.

Second, the record hits the cheapest broad provider. This is usually a large compiled database with wide coverage and mediocre precision. It resolves the easy majority of records for the lowest unit cost in the chain.

Third, unresolved records fall through to the next provider, typically a specialist: a provider strong in one region, one company-size band, or one field such as mobile numbers. Each fall-through step sees fewer records, which is what makes an expensive per-hit price survivable. You pay it only on the residue.

Fourth, the last provider in the chain is the most precise or the most expensive, sometimes an AI research agent of the kind Clay ships as Claygent, which assembles an answer from the live web instead of a static database. It sees the smallest and hardest slice of the list.

Fifth, verification gates whatever comes back. Emails get checked for deliverability, numbers get checked for line type and activity, and anything that fails goes back into the record as a miss, not a match. Skipping this stage is how waterfalls quietly poison CRMs: the cascade reports a 90% match rate, and a fifth of those matches bounce.

Want to put this into practice?

Reachium automates LinkedIn outreach, content publishing, and inbox management in one platform.

See Reachium →

Why not just buy the biggest database instead?

Sometimes you should. If one database covers your segment well, a single contract with a database-first tool is simpler and cheaper to run than a cascade, and the alternatives post weighs that case directly. The waterfall earns its complexity in three situations: your list spans regions or company sizes that no single provider covers evenly; the field you need is scarce, as mobile numbers usually are; or you enrich continuously, so the marginal coverage compounds month after month. Where a single provider already finds nearly everything you need, each extra cascade step buys fractions of a point of coverage at full price.

What does a waterfall cost per verified contact?

Cost per verified contact is total cascade spend divided by contacts that passed verification. It is the only waterfall metric that connects to pipeline, because unverified and bounced contacts produce no meetings.

Here is a worked example with stated assumptions, so you can rerun it with your own numbers. Take 1,000 records that need a work email. Assume provider A, a broad database, finds 55% of them. Provider B, a specialist, then finds 40% of the 450 records A missed, adding 180. Provider C, the premium last step, finds 30% of the remaining 270, adding 81. Verification then rejects 6% of all matches as undeliverable, leaving roughly 763 verified contacts from the original 1,000.

The shape matters more than the exact figures. Provider C touched only 27% of the list, so even at several times A's unit price it adds a modest share of total spend, while lifting verified coverage from about 517 contacts (A alone, after verification) to about 763. Whether that lift is worth it depends on what a verified contact is worth in your motion, which is a pricing model question, not a tooling one. We built a free waterfall planner worksheet that runs this exact arithmetic with your own hit rates and credit prices, including the fall-through at each step.

Where do waterfalls go wrong?

Four failure modes show up repeatedly. The cascade runs in price order rather than value order, so a cheap provider with sloppy data fills fields that a better provider would have filled correctly, and the stop-at-first-hit rule locks the sloppy answer in. The team measures match rate instead of verified rate, so the dashboard improves while bounce rates climb. Nobody owns the residue: records no provider can resolve pile up unexamined, even though they often share a fixable cause such as a bad domain field. And the waterfall runs on the whole CRM on a schedule, re-buying data the team already owns, when it should run on net-new and decayed records only. Data decay is the quiet argument for the cascade: people change jobs constantly, and a record that was verified last year is a hypothesis today. Treating enrichment as a stack decision with an owner, rather than a one-off purchase, is what separates teams that get value from it.

Want to put this into practice?

Reachium automates LinkedIn outreach, content publishing, and inbox management in one platform.

See Reachium →

FAQ

What is the difference between waterfall enrichment and a single database lookup?

A single lookup asks one provider for the field and accepts its coverage as final. A waterfall asks several providers in sequence, stopping at the first verified hit, so coverage compounds across providers while the expensive ones only process unresolved records.

Does waterfall enrichment improve match rates?

It improves coverage, because each additional provider resolves some records the earlier ones missed. The gain shrinks with every step: the first provider resolves the easy majority, and later steps fight over a residue that gets harder and more expensive per record.

What should I measure a waterfall on?

Cost per verified contact: total spend divided by contacts that passed verification. Match rate alone flatters the cascade, because it counts matches that bounce, belong to the wrong person, or fail a line-type check.

Do email and phone enrichment use the same waterfall?

No. They run as separate cascades with different provider orders. Email coverage is broad and cheap at the top of the chain; mobile numbers are scarcer, cost more per hit, and usually justify a different specialist order.

Is waterfall enrichment a Clay feature or a general method?

A general method that Clay popularized as a product pattern. Clay chains more than 150 providers, and dedicated tools including FullEnrich, Prospeo, and Databar sell waterfall enrichment on its own, while database platforms such as Apollo run waterfalls behind their own data.

Sources

GTMStack publishes original audits of the AI marketing stack. If you want the next one when it lands, the newsletter is the place to get it.