Waterfall Enrichment, Explained

What waterfall enrichment is

Waterfall enrichment is the practice of querying several data providers in sequence for the same field, and stopping as soon as one returns an acceptable answer. Instead of one vendor contract that covers perhaps 60% of your target market at a fixed price, you run a chain — provider A, then B, then C — and pay only for the calls you make.

The technique exists because B2B data coverage is uneven and non-overlapping. No provider is best across every geography, company size, and seniority. One is strong on US mid-market, another on European contacts, another on technical roles. A waterfall exploits that variance rather than paying a premium to ignore it.

How the sequence works

1. 01
   Normalise the input

   Clean and standardise the identifiers you have — domain, company name, full name, LinkedIn URL — before any call. Providers match on what you send them, and inconsistent input is the largest single cause of apparent low coverage.

2. 02
   Query provider 1

   The cheapest provider with the highest hit rate for your specific market. Not the best-known one; the one your own data says wins most often on your records.

3. 03
   Evaluate the result against a threshold

   A returned value is not a good value. Check confidence score, format, and plausibility. If it fails, treat it as a miss and continue.

4. 04
   Fall through to the next provider

   Repeat for each provider in the chain, typically three to five deep. Beyond five the marginal coverage is usually not worth the added latency and complexity.

5. 05
   Validate the winner

   Verify the final value independently — an email verification service for deliverability, a format and pattern check for phone numbers. This step is what separates a working waterfall from an expensive bounce generator.

6. 06
   Write back with provenance

   Store the value, the provider that supplied it, the confidence, and the timestamp. Without provenance you cannot audit quality or reorder the chain later.

Ordering the chain

The ordering is the entire technique, and the metric that governs it is cost per successful match, not list price.

A provider charging $0.02 per record that resolves 20% of your list has an effective cost of $0.10 per match. A provider at $0.12 that resolves 70% costs $0.17 per match but removes far more volume from the expensive tail. Which should go first depends on your data, which is why you have to measure rather than copy someone else's chain.

Realistic coverage benchmarks

What a well-built waterfall achieves on a clean, well-normalised list. Coverage falls sharply outside North America and Western Europe, and for companies under about 20 employees.

Two honest caveats. First, coverage numbers are only meaningful after verification — an unverified 95% is often an effective 70%. Second, these assume the input identifiers are clean. On a messy list, normalisation alone typically recovers more coverage than adding a fourth provider.

The validation layer

Most of the damage done by enrichment happens here, or rather in its absence. A waterfall that returns a value for 95% of records and validates none of them produces a bounce rate that damages deliverability for every campaign you run afterwards, including the ones that had nothing to do with the enriched list.

• Verify every email before it enters a sending system. Treat catch-all domains as a separate risk tier rather than valid, and set your own policy for whether to send.
• Reject implausible values. Role addresses, obvious placeholders, and mismatched domains should fail before write-back.
• Score, do not just flag. A grade per record lets you route the safe tier to automated sequences and the risky tier to manual review.
• Suppress against your own history. Prior bounces, unsubscribes, and complaints should never be re-enriched back into a sending list.

Caching and re-enrichment

The most common silent cost leak in GTM stacks is re-buying data you already own. Two controls fix it.

Cache with provenance. Every enriched value stored with its provider, confidence, and timestamp. Before any call, check the cache. This alone commonly cuts spend by a third in teams that have been running enrichment for a year without it.

Set a TTL per field type, because fields decay at very different rates. Firmographics change slowly; job titles change constantly.

Instrumenting the waterfall

You cannot order a chain you do not measure. Track four things per provider, continuously:

• Hit rate — share of queries returning an acceptable value, segmented by geography and company size.
• Cost per successful match — the number that determines position in the chain.
• Validation pass rate — how much of what a provider returns actually survives verification. This is where cheap providers often stop being cheap.
• Downstream conversion — reply and meeting rate on records sourced from each provider. The only measure that connects data quality to revenue.

Review quarterly and reorder. Provider performance shifts as their datasets and your ICP both change, and a chain ordered eighteen months ago is almost certainly no longer optimal.

Common mistakes

• Ordering by brand rather than by measured hit rate on your own list.
• No confidence threshold, so a low-confidence match blocks a better provider from ever being called.
• Enriching the whole database instead of the active segment. Enrich what you will contact this quarter.
• No monitoring, so a provider's API failing silently looks identical to genuinely low coverage.
• Treating enrichment as a project. It is a standing job with a budget, an owner, and a quarterly review — see what a GTM engineer owns.