CRM Enrichment Automation: The Waterfall Framework That Fills 90% of Missing Fields

A CRM with 90% of its fields filled sounds like a vanity metric until you try to run outbound, routing, or scoring on the other 10%. Then it becomes the whole problem. Every missing industry value is a lead that falls out of a segment. Every missing job title is a contact your sequence can’t personalize. Every empty headcount field is an account your territory rules can’t assign.

CRM enrichment automation is how ops teams fix this without hiring someone to copy and paste from LinkedIn. But most teams do it badly. They connect one data vendor, turn on “enrich all records,” watch the fill rate climb from 40% to 62%, and call it done. The remaining 38% stays empty forever, and the 62% they did fill is already decaying.

The better approach is a waterfall: a chain of providers queried in order, routed field by field, with validation between each step and a cost ceiling on the whole thing. I’ll walk through the architecture we use when we design these systems at Momentum Nexus, including the parts most vendor guides skip: field level routing, decay monitoring, and the cost per fill math that decides whether a provider earns its place in the chain.

If you want the cleanup side first, we covered the manual version in our CRM data hygiene sprint. This post is the automated layer that keeps the sprint from needing a sequel.

Why Single Provider Enrichment Stalls at 60%

The honest starting point: nobody has published an independent, controlled benchmark of enrichment fill rates. Almost every number you will find comes from vendors selling either a single database or a waterfall tool. Read them with that in mind.

With that caveat, the pattern across 2026 vendor reports is consistent:

Setup	Typical email match rate	Source type
One provider	55% to 70% (some reports say 35% to 50%)	Vendor blogs
Two providers	70% to 85%	Modeled estimate
Three providers	82% to 88%	Modeled estimate
Four providers plus verification	85% to 92%	Modeled estimate

Two things stand out. First, the gain per added provider shrinks fast. The jump from one to two is 15 points; from three to four it is a few. Second, the gain depends heavily on who you sell to. If your ICP is US SaaS companies with 50 to 500 employees, one strong provider may already hit 70%, and a waterfall adds 10 to 15 points. If you sell into manufacturing, healthcare, or outside North America, single providers fall apart and the waterfall earns its keep.

So the first decision isn’t which tools to buy. It’s measuring your own list. Pull 500 records from your CRM that you know the correct answers for, run them through each candidate provider, and record match rate and accuracy per field. A match is not the same as a correct answer. A provider that returns a stale email for 80% of records has a great fill rate and a terrible bounce rate.

The 90% in the title is a target, not a promise. On a well covered ICP with four sources and verification, it is reachable for work email and company fields. For mobile numbers it usually isn’t, and I’ll explain why that’s fine.

The Decay Problem Nobody Budgets For

Fill rate is a snapshot. Decay is the movie.

The most cited figure is about 22.5% annual decay for B2B data, derived from MarketingSherpa’s roughly 2.1% monthly measurement. Higher numbers (30% to 70%) circulate in vendor content, but the sourcing is thin, so I wouldn’t plan around them. What matters more is that decay is uneven across fields. Prospeo’s field breakdown, which is vendor data too, gives these annual ranges:

Field	Estimated annual decay	Re-check cadence we use
Work email	20% to 30%	Every 90 days, active accounts
Job title	15% to 25%	Every 90 days, active accounts
Direct phone	15% to 20%	Before any call campaign
Company firmographics	10% to 15%	Every 6 to 12 months
Mobile phone	5% to 10%	Before any call campaign
Name	1% to 2%	Never on a schedule

The same source says 15% to 20% of professionals change jobs in a given year. Run that against a 10,000 contact database and you have 1,500 to 2,000 contacts attached to the wrong company by December, whether or not anyone touched the record.

This changes how you design the system. A one time bulk enrichment is a depreciating asset. You need an ongoing loop, which means a budget line, not a project. If you’ve read our take on building RevOps as a system instead of a team, this is the same principle applied to data: the maintenance has to be automated, because nobody will remember to do it.

The Waterfall Architecture: Five Layers

Here is the framework. I call it the Five Layer Waterfall, because each layer answers one question the vendor guides skip.

1. Trigger layer: When does enrichment run?
2. Routing layer: Which provider is asked for which field?
3. Fallback layer: What happens when the first provider returns nothing?
4. Validation layer: Is the answer good enough to write?
5. Write and decay layer: Where does it land, and when does it expire?

Layer 1: Triggers

Enrichment on every record at every moment is the fastest way to burn budget. Run it on events.

• Record creation. A form fill, an import, or a rep-created contact fires a lightweight pass: company, domain, title, and email only.
• Stage change. When a lead becomes an MQL or a deal enters pipeline, run the deeper pass: headcount, funding, tech stack, phone.
• Decay timer. A scheduled job selects records where a field’s last verified date is older than its cadence from the table above and the record is tied to an open deal or active sequence.
• Manual request. A rep clicks a button on a record. This is the escape hatch, and it should be rate limited.

Notice what is missing: bulk enrichment of the entire database on a whim. Cold, dead records don’t earn credits.

Layer 2: Field Level Routing

Most teams route by record. Send the whole contact to Provider A. That is the wrong unit. Providers are strong on different fields, and you pay per lookup regardless of how many fields you wanted.

Route by field instead. A typical routing table looks like this:

Field	First choice	Why
Company name, domain, industry, headcount band	CRM native enrichment or the cheapest firmographic source	Cheap, high coverage, slow decay
Work email	Provider with best match on your ICP	Highest volume field, biggest cost lever
Job title and seniority	A provider that refreshes from professional profiles	Decays fast, accuracy matters more than fill
Direct dial or mobile	Specialist provider, often region specific	Expensive, only for stage-qualified records
Tech stack, funding, hiring signals	Signal specific source	Used for scoring, not for contact data

HubSpot users should note that HubSpot’s own enrichment (the product that came out of Clearbit, now sold as Breeze Intelligence) is credit based, with published bands around 7 to 30 cents per credit depending on volume. Third party coverage reviews say its strength is firmographics inside HubSpot and its weakness is direct dials and international contacts. That fits it neatly into the cheap first slot for company fields, not the contact data slot. Confirm current pricing on HubSpot’s page, because reports disagree on what standard enrichment costs.

If you’re already using Clay as your orchestration layer, our post on where Clay pays off and where it wastes credits covers the credit gating in detail. The routing logic in this post sits one level above it: it decides which fields deserve a Clay run at all.

Layer 3: Fallback Chains

A fallback chain is an ordered list per field. Provider A runs. If it returns empty, or fails validation, Provider B runs. Continue until the field is filled or the chain ends.

Rules that keep chains from becoming expensive:

• Cap the depth. Four providers is the ceiling for any field. The modeled gains above flatten after that.
• Order by cost, then by coverage. Put the cheapest provider with decent coverage first. A provider that costs three times as much has to fill three times as many of the leftover gaps to justify its slot.
• Stop on first valid. Never run all providers and compare. You pay for every call.
• Skip what’s already known. If the field has a value verified within its cadence window, the chain doesn’t start.
• Log the winner. Record which provider filled each field. This is the data you’ll use to reorder the chain later.

Each added provider also adds latency, typically 1 to 5 seconds. That is irrelevant for batch jobs and annoying for form fills, which is why form time enrichment should use only the first one or two providers and let the rest run asynchronously.

Layer 4: Validation

This layer separates a system that works from one that quietly fills your CRM with garbage faster than before.

Between each provider and the write step, check:

• Format. Email syntax, domain resolves, phone number parses for the stated country.
• Verification. Run emails through a verification step before writing. Vendor sources claim unvalidated single source data bounces at 8% to 15%, and that verification can bring a waterfall under 3%. Treat those as directional, but the order of magnitude matters for your sender reputation.
• Consistency. Does the enriched company domain match the email domain? Does the title plausibly match the seniority field? Conflicts go to a review queue, not the record.
• Overwrite policy. Never let enrichment overwrite a value a human entered. Write to a separate “enriched” property, or write only to empty fields, and let a rule decide when enriched data may replace old data.

That last point causes more CRM trust problems than anything else. A rep corrects a title after a call, a nightly job overwrites it with the vendor’s stale version, and the rep stops trusting the CRM. Protect human edits.

Layer 5: Write and Decay

Every enriched field needs two companion properties: source (which provider) and last verified date. Without them you can’t run the decay timer from Layer 1, you can’t audit provider quality, and you can’t answer the question a sales leader will eventually ask: where did this number come from?

For HubSpot or Salesforce, this means creating the companion properties before turning anything on. It takes an afternoon and saves months.

Cost Per Fill: The Number That Decides Your Chain

Vendors quote price per record or per credit. That is the wrong denominator.

Third party pricing guides put credit based tools like Clay and Apollo between roughly 3 and 15 cents per enriched record, and premium sources with direct dials considerably higher. One vendor reports about 5 cents for a verified email and about 58 cents for an email plus direct dial on the same platform. But a price per attempt says nothing about how many attempts produce a usable field.

The metric to track is cost per usable fill:

Cost per usable fill = total provider spend for a field, divided by the number of values that passed validation and were written.

A worked example, with illustrative numbers you should replace with your own:

Provider	Cost per lookup	Lookups	Valid fills	Spend	Cost per fill
A	4 cents	1,000	600	40 dollars	6.7 cents
B (runs on A’s 400 misses)	8 cents	400	160	32 dollars	20 cents
C (runs on B’s 240 misses)	15 cents	240	60	36 dollars	60 cents

Total: 820 fills out of 1,000 records for 108 dollars, 13.2 cents per fill on average. But look at the marginal cost. Provider C costs 60 cents for each additional fill. Is a work email for one of your leftover records worth 60 cents? For a 30,000 dollar ACV account, absolutely. For a 500 dollar a year product, never.

This is why the chain depth should vary by account tier. Our default structure:

Tier	Chain depth	Fields	Verification
Tier 1 (named accounts, open deals)	Up to 4 providers	All, including phone	Always
Tier 2 (ICP match, no deal)	Up to 2 providers	Email, title, firmographics	Always
Tier 3 (everything else)	1 provider	Firmographics only	On send

A flat chain applied to every record is the most common reason enrichment budgets explode.

What 90% Actually Means

When people say “fill 90% of missing fields,” they usually mean averaged across everything. That average hides a lot. Set separate targets by field, because the economics differ:

Field group	Realistic target	Notes
Company domain, industry, headcount band	90% to 95%	Cheap, stable, high coverage
Work email (Tier 1 and 2)	85% to 90%	Needs verification to count
Job title and seniority	80% to 90%	Accuracy matters more than fill
Direct or mobile phone	40% to 60%	Specialist data, region dependent
Tech stack, funding	60% to 80%	Signal sources vary

Blending these into one 90% figure would force you to either overpay for phone data or declare failure. Per-field targets let you decide where the marginal dollar goes. I’d rather have 95% on firmographics and 50% on phones than a blended 80% that tells me nothing.

Monitoring: The Four Dashboards You Need

A waterfall is not set and forget. Four views keep it honest:

1. Fill rate by field and tier. Weekly. A drop means a provider changed coverage, a trigger broke, or a new lead source brought in records that don’t match your ICP.
2. Provider win rate by position. Which provider fills the most in slot one? If your slot three provider fills under 5% of its inputs, you are probably paying for decoration.
3. Cost per usable fill by field. Monthly. Compare against what that field is worth to the business.
4. Decay age. The share of key fields whose last verified date is past their cadence. This is the leading indicator. Fill rate stays flat while quality rots, and this chart catches it.

If you can only build one, build the decay age chart. It’s the one nobody has by default.

A 30 Day Implementation Plan

Week 1: Audit and measure.

• Export 500 records with known correct values and run them through each candidate provider.
• Record match rate, accuracy and cost per field. Discard vendor claims that your own sample contradicts.
• Count your current fill rate by field and tier. This is your baseline.

Week 2: Build the foundation.

• Create source and last verified properties for each enriched field.
• Write the overwrite policy and get sales to agree to it.
• Define tiers and the routing table.

Week 3: Build the chain.

• Wire the triggers: creation, stage change, decay timer.
• Build the fallback chain for email first, since it is the highest volume field.
• Add validation between providers and a review queue for conflicts.

Week 4: Run, watch, tune.

• Run on Tier 1 accounts only for the first week and inspect a sample by hand.
• Extend to Tier 2. Hold Tier 3 until costs are known.
• Stand up the four dashboards. Reorder the chain based on win rate and cost per fill.

By the end of week four you will know which provider in the chain contributes almost nothing. Dropping it is usually the cheapest improvement available.

Common Mistakes

1. Enriching the whole database at once. You pay to fill records that will never be worked, and they start decaying immediately.
2. Routing by record instead of by field. You pay a premium provider for fields a cheap one would have filled.
3. Letting enrichment overwrite human edits. Reps stop trusting the data within a month.
4. Counting matches as fills. An unverified email is a liability, not a field.
5. No source or date property. You can’t audit, you can’t schedule decay checks, and you can’t prove the system works.
6. One chain for every account. Tier your depth, or your marginal cost per fill will quietly climb above the value of the account.
7. Treating it as a project. Decay means the work never ends. Budget it as run cost.

If your stack is also fragmented across tools, fix that first or alongside this. We went through the pattern in our business operating system piece on tool sprawl; an enrichment chain that writes into three disconnected systems just spreads the same missing fields around.

The Takeaway

The goal of CRM enrichment automation isn’t a high fill rate. It’s a CRM where the fields your routing, scoring and sequences depend on are filled, verified, dated, and cheap enough to keep that way. Start by measuring providers on your own list, route by field, cap the chain by account tier, and track decay age from day one.

If your CRM has gaps you can’t explain or an enrichment bill you can’t justify, we can map your current stack and show you where the chain should start. Book a free growth audit, or try our free AI growth tools at app.momentumnexus.com. We also build CRM agents for this kind of work; you can read about them at /crm-agents.

Sources referenced for benchmarks: vendor reports from Landbase, Unify, Lantern, FullEnrich, Prospeo, Cleanlist and HubSpot’s Breeze Intelligence pricing guide. Most are vendor published, so measure on your own data before committing.