How to Set Up Waterfall Enrichment in Clay, Step by Step

Every contact database is incomplete, and each one is incomplete in a different way. One provider knows the startup founders and misses the enterprise directors. Another has broad coverage but returns a guess where you wanted a certainty. If you buy from a single source you inherit that source's blind spots.

The waterfall is Clay's answer to that, and it is the feature the platform is best known for. It is worth learning properly, because a waterfall built the wrong way round can cost you several times what the same coverage should have cost.

This guide covers what a waterfall actually is, how to build one in Clay, how to order the providers inside it, and the settings that decide whether you end up with a clean list or a bounce problem. Every step below comes from Clay's own documentation or their published guidance, and the sources are listed at the end.

What is a waterfall enrichment in Clay?

A waterfall is a fallback chain inside a single column. You line several data providers up in a fixed order, the record goes to the first one, and if that provider returns a confident result the chain stops there. If it does not, the record falls through to the next provider, and so on until something lands or the chain runs out.

Clay's documentation frames this as using multiple data providers in a predetermined sequence so you do not duplicate tasks or spend extra credits. The important consequence is that the work shrinks at every level. Your first provider sees every record. Your second sees only the records the first one missed. Your fourth may see a handful.

Clay's own waterfall guide is explicit about the billing shape that follows from this: a contact found by the first provider never touches the second, so you never pay the second provider's credit on that contact. You are paying for hits, not for attempts.

How do you build a waterfall in Clay?

Clay documents a short path for building a waterfall from scratch. The steps below follow their documentation. If a label has moved by the time you read this, the shape of the flow is still the same: create the column, choose what kind of data it produces, then arrange the providers inside it.

1. Add the column — Click Add column at the far right of your table. Clay's docs also describe reaching enrichments through the Add enrichment button in the top right corner of the table, and a named waterfall such as Work Email is found by searching there.
2. Select Waterfall and name it — Choose Waterfall, then click the pencil icon to rename the column to something you will recognise later. Naming matters more than it sounds, because waterfalls generate their own supporting columns and a table full of untitled ones gets confusing fast.
3. Set the Data Type — Change the Data Type to match what you want the column to produce. This is what decides which providers Clay offers you inside the sequence.
4. Arrange the providers — Insert your actions in the waterfall sequence and configure each one. Clay's docs state you can reorder, add, or delete the providers in the sequence, and that a toggle switch next to a provider lets you skip that step without deleting it. That toggle is how you test an order cheaply.
5. Configure the run settings — Open Run settings to set auto-update and any conditional trigger. Clay's guidance elsewhere is to keep auto-update off while you are still building, so half-finished configurations do not start running themselves.
6. Save, and save as a template if you will reuse it — Click Save. Clay also offers a Save as template option so a waterfall you tuned once can be dropped into the next table instead of rebuilt.

Two optional settings are worth turning on straight away. Clay documents an option to output the name of the successful provider, and an option to hide the provider columns. Together they give you a tidy table that still tells you which source is actually earning its place.

The Work Email waterfall, in more detail

Clay ships a purpose-built Work Email waterfall, and it is the one most teams set up first. According to Clay's documentation you click Add enrichment in the top right corner of your table, search for Work Email, select it, and then choose between Quick setup and Full configuration before mapping your input columns and saving.

Two settings inside it change the economics more than the provider list does.

• Infer Email. Clay documents a toggle for including an infer-email enrichment as the first step, and states that this step is free. It takes an Email Pattern, which defaults to [email protected], the company Domain, and at least one name input. If the inferred address passes validation the waterfall stops there and you have paid nothing for the find. If it fails, the chain continues to the paid providers.
• Validation. An optional validation provider can run after each step, with a toggle for whether validation success is required. Clay documents four strategies: Conservative, Balanced, Aggressive, and Advanced for manual control. Their guidance describes Conservative as the safest, and Balanced as the middle ground that includes catch-all domains.

There are two more settings worth knowing. Clay documents a Threshold for duplicate results, which defaults to zero, and their tips suggest pairing it with Conservative validation to stop credits going on an address that keeps failing. And Hide provider columns is on by default, which is why your table may look simpler than the work happening inside it.

What order should the providers go in?

Cheapest reliable provider first, most expensive specialist last. Clay's waterfall guide gives three ordering factors in priority order: price per lookup first, quality at your confidence threshold second, and coverage third.

The logic is straightforward once you see the volume curve. Early levels run on every record, late levels run on almost none. So the expensive provider sitting at the bottom only ever gets asked about the difficult contacts nobody else could resolve, which is exactly what you want to pay a premium for.

Quality still constrains the order. Clay's guide is clear that a provider only earns an early slot if it meets your accuracy bar at the confidence threshold you have set, because everything the first provider returns goes straight into your list without a second opinion. Cheap and wrong at position one is worse than expensive and right at position four.

How much does a waterfall cost to run?

Clay bills enrichment in two separate currencies, and understanding the split explains most of what looks strange on a usage report. Clay's documentation states that each record enriched or exported counts as one Action regardless of the data source or provider, while Data Credits are the variable amount you pay for the data itself.

That means a waterfall does not multiply your Action count by the number of providers in it. It is still one enriched record. What varies is the Data Credit cost, which depends on which provider ended up supplying the answer.

One more detail from Clay's documentation is worth planning around. If you connect your own API keys for a provider, you remove the Data Credit cost while still consuming one Action. For a team that already pays for a contact data subscription, that is often the single biggest change you can make to a Clay bill.

Gotchas the setup screen does not warn you about

None of these are flaws in Clay. They are the ordinary consequences of a tool that will happily run three thousand lookups the moment you tell it to.

• Your domain column is the whole game. Independent setup guides consistently flag this: email finders want a root domain, not a full URL with protocol and path. A column full of long URLs quietly fails across every provider in the chain.
• Empty inputs still cost you. Rows missing a domain will still be sent through the waterfall unless you filter or condition them out first. Clay's credit guidance recommends conditional runs through the Only run if box in Run settings for exactly this.
• Coverage and accuracy trade off by design. Clay's guide states plainly that no single provider optimises both at once. A waterfall does not remove that trade-off, it just lets you take the good half from several sources.
• Test small, then scale. Clay's credit conservation guidance recommends testing on a small sample of about ten rows before running a whole column. On a waterfall this matters twice over, because a bad input mapping burns through every level of the chain on every row.

Once a waterfall is running, the useful habit is to look at which provider is actually finding things. If you turned on the option to output the successful provider's name, you can group by it and see whether the provider you put at position three has ever contributed anything. Sequences drift out of date as providers change their coverage.

Chain you tune, or employee you brief

Clay's waterfall is a genuinely good piece of design. It takes a real problem, the fact that no data vendor covers everyone, and turns it into a configuration you control down to the individual lookup. For a team with someone who enjoys that control, the results speak for themselves.

Sistava approaches it from the other end. You hire an AI Employee, describe the outcome you want in a sentence, and it works out how to get the contacts and check them. Neither approach wins outright. If you want to own the lookup chain and tune it, build the waterfall. If you want the verified list without owning the machinery behind it, brief the employee.

Sources

• Clay documentation, Waterfalls
• Clay documentation, Work Email waterfall
• Clay, The Complete Guide to Waterfall Enrichment
• Clay documentation, Actions and Data Credits
• Clay documentation, Ways to save Clay credits
• Clay documentation, Enrichments
• LeadMagic, Clay Waterfall Enrichment guide