SALES
Pipeline you can plan around
Outbound, inbound handoff and expansion, run as one motion.
GTM BrainFOR SALES
SEND 01SEND 05
It gets sharper with every reply and every call it reads.
GTM STACK AND DATAPart of: The GTM tech stack by company stage: the layers, the tools and what to build

Waterfall enrichment: how it works, and how we build ours for sourcing and verification

A short definition, the mechanics behind each step, and the exact sourcing and verification orders we run for clients, with the rule that ends each one.

Dark header with the theme Cheapest check first above white tiles: the Supabase and Prospeo logos, a funnel icon, the Findymail logo and a double check mark.

Waterfall enrichment fills a missing field, such as a contact at an account or a work email, by asking several data providers one after another. Each provider only receives the records the providers before it could not fill. A record leaves the chain as soon as one result passes your quality check, so nothing gets bought twice.

The term spread through Clay, where a waterfall is a column that tries providers in order until one returns an answer. The same logic runs just as well in code. Our own email-finding chain started life as a Clay table before we rebuilt it as part of the engine our GTM tech stack runs on. Below is how the mechanics work, the two chains we run for clients, and how to build your own.

How a data waterfall works

Every chain, whatever tool it runs in, has three parts. Get any one of them wrong and the chain either misses people or pays for the same person twice.

  • The order. Which provider is asked first, second and third. This is what decides the cost of the chain.
  • The check. What a result has to pass before you accept it. A found email that nobody verified is a guess with a confident label on it.
  • The stop. What ends the chain for a record, and what happens to the records that reach the bottom with nothing.

The order follows the cost model of each source. Data you already hold costs nothing to reuse. A flat-rate source costs the same whether you run one record or ten thousand. A finder that charges only on a hit costs nothing when it misses: Prospeo, for example, charges 1 credit per verified email it finds. A metered search charges per lookup. Prospeo lists 1 credit per verified email on its pricing page (checked 22 Oct 2026). Put them in that order and the expensive steps only touch the records that are still empty.

In practice, the method covers two separate jobs. Sourcing answers who should be on the list. Verification answers whether you can safely send to an address. They need different providers and different stop rules, so we run them as two chains rather than one long one.

The sourcing chain we run

Our lead sourcing skill builds one deduplicated list from a client's ICP. The order is fixed, and nothing paid runs until a person has approved the numbers.

  1. Our own Supabase. Anyone already in our database is reused before anything is bought. That record was paid for once already.
  2. BlitzAPI, domain-first. It takes the list of target companies and returns the people at each one. It runs on a flat subscription, so it goes ahead of every metered source.
  3. Metered providers, cheapest first. Prospeo and the other per-lookup sources only see the people the first two steps did not cover. Prospeo searches by job title, which catches people a domain search misses.

Two rules sit around those steps. Before any paid call, a preview shows the real match count and where credits will be spent, and a person approves it. And every candidate is checked against the database before its email is revealed, so a paid source never charges us for someone we already hold. A source that fails mid-run is logged, and the chain carries on with the next one.

A person found without a usable email is kept. They move to the second chain, which exists to repair exactly that.

The verification chain we run

The email verification skill is the repair pass for anyone without an address we would send to. It starts by checking what is already there, because an address that passes costs one check and no finder credits at all.

  • Verify every address first. Role addresses such as info@ or sales@ are ruled out, since nobody owns a shared inbox personally. A clean result ends the chain for that person.
  • Second check on catch-all and unknown. A catch-all domain accepts any address, so a standard check proves little. BounceBan settles those results without a message going out.
  • Finders, one at a time. If the address still is not clean, up to four finders run in turn. Findymail, LeadMagic and Prospeo are among them.
  • Same rule for every candidate. Each address a finder returns goes through the same verification as the first one. The first candidate that passes wins, and no later finder is called.

Each provider is asked at most once per address. BounceBan is only called for catch-all or unknown results, since a second check on a result that is already settled wastes a credit. Findymail sits early among the finders because a miss costs nothing there.

People who reach the bottom with nothing clean end in one of two states. Held means nobody could tell either way. Not deliverable means the address is confirmed dead. Both stay on file and neither gets sent to. Reporting them separately matters, because the held ones are often still worth a human look.

Both chains, step by step
StepWhat it checksWhat stops the chain
Own databaseWhether the person is already on fileFound: reused, nothing bought
BlitzAPIThe people at each target company domainEnough matches for the list size
Metered providersJob-title search for who is still missingThe limit is reached, or the spend is not approved
First verificationValid, invalid, catch-all or unknownValid: done, no finder called
BounceBanCatch-all and unknown results onlyDeliverable: clean. Undeliverable: go to the finders
Finders in turnA new candidate address, re-verifiedThe first candidate that passes
End stateNothing clean after every finderHeld or not deliverable, and a human decides

The same chain inside Claude

We published a lighter version of this as a workflow that runs inside Claude. It began with a frustration: every CSV pushed into Clay for email enrichment had to be reformatted first. An Apollo export, a client CSV and a BlitzAPI pull each arrived in a different structure.

In the rebuilt version, Claude reads whatever file it is given. With no CSV, it searches our Supabase first, then BlitzAPI, then Exa, until the leads are found. Any email already on the list is checked by DeBounce. A finder runs on the rest, with LeadMagic as the fallback, and any catch-all result gets a final check through BounceBan, without sending a single message.

Enrichment only fills in records. Deciding which accounts deserve a lookup this week is the job of buying signals. The two work best together: the signal picks the account, and the chain finds and checks the person.

How to build a waterfall enrichment chain

The steps below work in Clay, in n8n or in code. Write them down before you connect a single provider.

  1. 01Define the field and what done meansName the one field the chain fills, then the standard it has to meet. For email that is an address that passes verification, with a rule for catch-all results written down.
  2. 02Check what you already ownPoint the first step at your own database or CRM. Anything found there is reused, and it also tells you which records the paid steps can skip.
  3. 03Order the rest by cost modelFlat-rate sources go next, then finders that charge only on a hit, then anything metered per lookup. Within each group, start with the provider that covers your market best.
  4. 04Deduplicate before you payMatch every candidate against what you hold before revealing its email. A duplicate caught at this point costs nothing; one caught after the reveal has already been charged.
  5. 05Set one check and one stop ruleUse the same verification for every candidate, whichever provider returned it. Stop at the first result that passes, and never call a provider twice for the same record.
  6. 06Put a cost gate before the first paid callPreview the real match count and the expected spend, and have a person approve it. Run 10 rows first, read the summary, then scale.
  7. 07Decide what happens to the leftoversKeep anyone who ends without a clean result, mark why, and send nothing until someone decides. Deleting them loses people you already paid to find.
  8. 08Log which step filled each recordRecord the provider that produced every result. That log shows which tiers earn their place and which ones to move or drop.

Tools that run a waterfall for you

Clay is where most teams build their first waterfall, with waterfall columns across a long list of providers and a credit system to manage. FullEnrich is a ready-made chain. It runs one contact through several vendors in turn for a verified email or mobile number, and bills nothing when nothing turns up. Apollo covers search, verification and sending under one login, which suits an early team better than a multi-provider setup.

The alternative is to build the chain yourself, which is what we do. It costs engineering time, and in return you choose every provider, every check and every stop rule. Clay vs Claude Code is the question that decides where that line falls, and our Clay alternatives guide maps the other options. You do not need all of these tools. One chain that is ordered well beats three that overlap.

We run both chains for clients as part of our outbound service, so each list gets the coverage of every source in the order above. If you want that built around your own ICP, book a call.

Want this built for your team?One call with a founder. We tell you which parts apply to your GTM, and what we would build first.Request a call
What is waterfall enrichment?

A method for filling missing contact data by querying several providers in a set order. Each provider only sees the records the earlier ones left empty, and a record leaves the chain once a result passes verification.

What are the best waterfall enrichment tools?

Clay is the most common place to build one, with waterfall columns across many providers. FullEnrich runs a ready-made chain for emails and mobile numbers. Teams with an engineer can build their own in code and choose every provider and stop rule themselves.

Where does the data in an outbound list come from?

From several sources in sequence. Data you already hold comes first, then databases that search by company domain or job title, then finders that look up one address at a time. LinkedIn is one input among them. Each provider covers some people well and misses others, which is why they run in order.

Where does the enrichment spend actually go?

In a chain ordered by cost, owned data and flat-rate sources add nothing per record. The spend sits in the metered searches and the paid finders, and they only run on the records the earlier steps could not fill.

FURTHER READING
Matteo Fois
WRITTEN BYMatteo FoisCo-founder and COO, Kinetyca

Lawyer by training, then operations at Amazon. Runs delivery: process, documentation and the detail corporates expect.

LinkedIn ↗YouTube ↗
KINETYCA GTM NEWSLETTERGet this a week early.Everything we publish here goes to newsletter readers first, along with the cases and numbers that never make it onto the site. Tuesdays and Thursdays, written by David and Matteo.
TWICE A WEEK · NO SPAM · ONE-CLICK UNSUBSCRIBE
Kinetyca GTM NewsletterNews, cases and what is actually working in GTM right now.
© 2026 KinetycaPrivacyTermsImprint
Follow the foundersDavidCEOMatteoCOO