Claude Data Enrichment: Waterfall Enrichment From 50+ Providers (2026)
By Kushal Magar · July 6, 2026 · 11 min read
Key Takeaway
Claude data enrichment works best as a waterfall: SyncGTM's MCP queries 50+ providers in a tuned order, stops the moment one returns a verified result, and only falls deeper on the contacts cheaper sources miss. You pay per hit — not per attempt — and match rates climb from 40–60% single-source to roughly 87%.
Every enrichment provider has holes in its data. Apollo misses contacts that Prospeo holds. Prospeo misses contacts that a fourth source has. Run your list through any single provider and 40–60% of it comes back blank.
Waterfall enrichment fixes the coverage problem by cascading. It queries one provider, and if that provider returns nothing, it falls through to the next — and the next — until a verified result surfaces or the sources run out.
Claude data enrichment is where that cascade meets a conversation. Connect SyncGTM to Claude through the Model Context Protocol, and one plain-English prompt runs a 50+ provider waterfall over your whole list. This guide is about the mechanics — how the cascade is ordered, when it stops, and how to run it without burning credits.
What is Claude data enrichment?
Claude data enrichment is a workflow where Claude connects to SyncGTM's MCP and enriches B2B contacts through a waterfall — querying 50+ data providers in sequence, stopping the moment one returns a verified email, phone, or company field, and falling through to the next provider only when the current one comes up empty. One prompt enriches a list; you pay one credit per verified result, not per provider queried.
TL;DR
- A waterfall cascades, it doesn't broadcast. Providers are queried in a tuned order. The cascade stops at the first verified hit and only falls deeper on the contacts earlier sources miss.
- Coverage jumps from 40–60% to ~87%. No single provider covers every contact. Cascading across 50+ sources fills the gaps one provider alone leaves blank.
- You pay per hit, not per attempt. SyncGTM bills one credit per verified data point regardless of how many providers ran behind the scenes.
- Each data type has its own waterfall. Work email, personal email, mobile, direct dial, and verification each run a provider order tuned for that field.
- Claude orchestrates it in plain English. No provider API keys, no SDKs, no code. Point Claude at a CSV and describe the fields you want.
- Cost: free tier covers 100 enrichments/month; paid plans start at $49/month for 2,500 credits.
What This Guide Covers
This is a mechanics-first tutorial. It covers what a waterfall actually does under the hood, why cascading beats any single provider, how SyncGTM orders 50+ sources, and the exact six-step flow to run a waterfall inside Claude — from a single contact to a full CSV.
If you want the broader picture of every field Claude can pull per contact, read the companion guide on Claude enrichment. This post zooms in on the waterfall engine that makes the coverage numbers work.
Written for GTM engineers, SDR leaders, and RevOps teams who already live in a terminal. If you can open Claude Code and edit a config file, you can follow along.
What Is Waterfall Data Enrichment in Claude?
Waterfall data enrichment is an ordered cascade of data providers queried one after another until a verified result is found. The term comes from the visual: data "falls" from the first source to the second to the third, each catching what the one above it dropped.
Inside Claude, that cascade is exposed as a single MCP tool. When Claude asks SyncGTM for a contact's work email, it does not pick a provider — SyncGTM does. The waterfall sends the lookup to its highest-confidence source first, checks the result, and either returns it or falls through. Claude sees one clean answer; the 50-provider cascade happens beneath the tool call.
This matters because the alternative — wiring up Apollo, Prospeo, Hunter, and a dozen other APIs by hand — means managing that many keys, rate limits, and response formats. For a fuller definition of the pattern, see what is waterfall enrichment. The short version: waterfall trades the illusion of one perfect provider for the reality of many imperfect ones, queried in the right order.
Why Waterfall Beats Single-Provider Enrichment
The math is simple. If provider A covers 55% of your list and provider B covers 50%, neither alone is enough. But their gaps rarely overlap perfectly — the contacts A misses are often the ones B holds. Query both and combined coverage climbs well past either number.
Stack four or five providers and the curve flattens near the ceiling of what is findable. That is why SyncGTM's waterfall reaches roughly 87% verified email match rates in testing, while a typical single source lands at 40–60%. According to G2's sales intelligence category, coverage and accuracy are the two metrics buyers cite most when a data tool fails them — and waterfall is the structural answer to both.
| Approach | Email Match Rate | Mobile Match Rate | Key Limitation |
|---|---|---|---|
| Single provider | 40–60% | 15–30% | One database's blind spots become yours |
| Manual multi-tool | 70–80% | 30–45% | You manage every key, credit, and format |
| Claude + SyncGTM waterfall | ~87% | 50–65% | One tool, one credit per verified hit |
The gap widens on hard data. Mobile numbers are scarce in any one database — this is exactly where cascading earns its keep, and why waterfall contact providers exist as a category at all.
How the Waterfall Sequences 50+ Providers
A waterfall is only as good as its order. Query providers in the wrong sequence and you either pay too much or miss too much. SyncGTM tunes the order per data type using three signals.
1. Accuracy first, then breadth
The provider most likely to return a verified result for this contact type goes first. A high-accuracy source that also verifies its output stops the cascade early and cleanly — no second guess needed.
2. Cost-aware ordering
Because the cascade is stop-on-hit, the cheapest reliable providers sit near the top. Most contacts resolve in the first one or two hops, so the expensive specialist sources are only ever touched for the hard 10–15% that everything else missed.
3. Data-type specialisation
The provider that is best for work email is rarely best for mobile. So there is no single global order — each field runs its own waterfall with its own ranked list. SyncGTM exposes these as distinct MCP tools Claude can call:
- find_work_email — cascades B2B email databases, verifies, and returns a deliverable address.
- find_mobile_number — a separate cascade tuned for cell coverage, where single-source hit rates are lowest.
- find_work_phone — direct-dial and switchboard numbers from a distinct provider set.
- verify_email — an SMTP-level check that runs before you ever send, so bounced sends never eat a credit downstream.
- enrich_person / enrich_organization — layer contact and company fields on top of the found identifiers.
The stop-on-hit rule
A verified hit ends the cascade immediately. Providers below it are never queried, never billed. This is what makes a 50-provider waterfall cheaper than it sounds — depth is insurance you only draw on when the shallow end fails.
Competing MCP servers run the same idea at smaller scale. FullEnrich, for example, cascades across roughly 20+ providers. The mechanics are identical; the difference is how many sources sit in the waterfall and how tightly the order is tuned per data type.
What You Need Before You Start
- Claude Code or the Claude desktop app with MCP support enabled.
- A SyncGTM API key (the free tier includes 100 enrichments/month).
- A list of leads with at least a full name and company domain, or a LinkedIn URL — the seed the waterfall enriches from.
That is the whole stack. No provider accounts, no separate credits with Apollo or Hunter — the waterfall wraps all of them behind one key.
Step 1: Connect the SyncGTM MCP
The SyncGTM MCP is a Model Context Protocol server that exposes every waterfall as a tool Claude can call. Add it to your MCP config:
# ~/.claude/mcp.json
{
"syncgtm": {
"command": "npx",
"args": ["syncgtm-mcp"],
"env": {
"SYNCGTM_API_KEY": "your-api-key"
}
}
}Restart Claude and confirm the tools are live. For the official walkthrough of adding servers, see the Claude Code MCP documentation. Once connected, Claude can run the full waterfall from the terminal — no REST client, no Python.
Step 2: Watch a Single Waterfall Run
Start with one contact so you can see the cascade in action. Ask Claude to narrate which hop returned the result — useful for understanding your list before you spend at scale.
Prompt example — single-contact waterfall
"Find and verify the work email for Jane Doe, VP Marketing at Acme (acme.com). Use the SyncGTM waterfall, and tell me the verification status and how confident the match is."
Claude calls find_work_email, the cascade resolves, and verify_email confirms deliverability. One verified address, one credit — even though several providers may have been checked to get there.
Step 3: Control Waterfall Order and Depth
You do not hand-order 50 providers — that is SyncGTM's job. But you do control two levers that shape the cascade: which data types to chase, and how strict the verification bar is.
Set these once in your project's CLAUDE.md so every run behaves the same way:
## Waterfall Enrichment Config Provider: SyncGTM MCP (waterfall mode) Data types to resolve, in priority order: 1. work_email (required, must be verified) 2. mobile_number (required) 3. work_phone (optional fallback if no mobile) Rules: - Only accept work_email with verification = valid - Stop the waterfall on first verified hit (default) - If mobile_number is missing, do NOT re-run — mark "not found" - Never overwrite a value that is already verified in the source file
The last two rules matter for spend. "Do not re-run on miss" stops Claude from burning credits re-querying contacts that genuinely have no findable mobile — the deepest, most expensive part of any waterfall.
Step 4: Waterfall Mobile and Direct Dials
Phone data is where waterfall matters most. Mobile coverage in any single database is thin — cascading is often the difference between a 20% and a 60% hit rate.
Run mobile as its own pass so you can see the true cost of phone data separately from email:
Prompt example — mobile waterfall
"For every contact in leads.csv that already has a verified work email, run the mobile-number waterfall. Where no mobile is found, fall back to a direct-dial work phone. Add both columns and flag which came from mobile vs. work phone."
Chaining email-first then phone keeps spend tight: you only pay to find a mobile for contacts already worth calling. This is the same tiering waterfall email finders use, applied to phone data.
Step 5: Verify and Dedupe Before You Spend
The cheapest credit is the one you never spend. Two guardrails cut waterfall waste before it starts.
Dedupe first. Have Claude collapse duplicate rows by email or domain + name before enriching. Two rows for the same person means two waterfall runs and two credits for one contact.
Verify at the edge. Run verify_email on any addresses your list already contains. A clean existing address skips the email waterfall entirely — you only cascade for the gaps. Bad ones get flagged so they do not poison your send.
Prompt example — pre-flight cleanup
"Before enriching leads.csv: 1) dedupe by domain + full name, 2) verify any emails already in the file and keep the valid ones, 3) only run the email waterfall on rows with no valid email. Report how many credits step 3 will cost before you run it."
Step 6: Run the Waterfall Across a Full List
Once the config and guardrails are set, the full run is a single prompt. Claude orchestrates every waterfall, applies your rules, and writes an enriched file.
Full waterfall prompt
"Take leads.csv and enrich it with SyncGTM: 1) dedupe by domain + name, 2) waterfall a verified work email for every row, 3) for rows with a verified email, waterfall a mobile then a work-phone fallback, 4) leave misses marked 'not found' without re-running, 5) export leads-enriched.csv with source and verification columns, and give me the final match-rate summary."
The output is a send-ready file with a match-rate summary — how many emails verified, how many mobiles landed, and where the waterfall came up empty. That summary is your coverage report, generated in the same run. To push the enriched list into your CRM or fold company-level data on top, continue with Claude Code firmographic data enrichment.
Tips to Maximise Coverage and Cut Credit Waste
- Enrich email before phone. Emails are cheaper and more findable. Qualify on email, then only spend deep-waterfall phone credits on contacts worth calling.
- Feed the waterfall a clean seed. A correct company domain lifts match rates more than any config tweak. Fix obvious domain typos before you run.
- Never re-run a genuine miss in the same session. If the full waterfall found nothing, the next-day retry usually finds nothing too. Mark it and move on.
- Ask for a cost estimate first. Claude can count how many rows will actually hit the waterfall after dedupe and verification — approve the spend before the run, not after.
- Log the source column. Keeping which hop resolved each contact tells you when your seed data — not the waterfall — is the bottleneck.
These are the same disciplines behind well-run B2B lead enrichment programs — the waterfall just makes them a one-prompt habit instead of a spreadsheet chore.
Conclusion
The reason single-provider enrichment disappoints is not bad data — it is one database's blind spots becoming your whole list's blind spots. Waterfall enrichment removes that ceiling by cascading across 50+ sources, and stop-on-hit billing keeps the depth affordable.
Run inside Claude, it collapses to a conversation. Connect the SyncGTM MCP once, set your rules in CLAUDE.md, and one prompt takes a raw list to a verified, send-ready file with a coverage report attached — no provider keys, no code, no spreadsheet gymnastics.
Start free — 100 enrichments/month, no credit card required. Connect Claude, point it at your next list, and watch the waterfall cascade across 50+ providers.