Waterfall enrichment: where to stop adding data sources
Every bundled prospecting tool sells waterfall enrichment. Here is how to measure what each added source really adds, and where the chain should end.
Waterfall enrichment stops paying off when the next source's cost per usable record is higher than what that record is worth to you. For most teams that happens at the second or third provider, and you find the exact point by measuring each source's marginal match rate, deliverable rate and credit cost on your own ICP.
The idea is simple, and every bundled prospecting tool now advertises it: ask provider one for an email, ask provider two if provider one misses, and keep going. The part that doesn't get advertised is the shape of the curve. Let me walk through an example, and then tell you what I'd do.
How waterfall enrichment adds cost faster than coverage
The first provider does most of the work. Each one after it only sees the records the earlier ones failed on, so it is fishing in the leftovers. Those leftovers skew toward the hardest contacts: small companies, unusual titles, regions the vendors cover thinly. They also skew stale, which is why the match rate falls and the bad-record rate rises together.
That's the mechanism. The sales page shows cumulative match rate going up. What you need is the marginal match rate: of the records that reached source N, how many did it resolve, and how many of those were usable?
Here is a worked example. These are my assumptions, not benchmarks, so swap in your own. Say you pilot 500 ICP records through a four-source chain:
- Source one resolves 275 records (55%) at 1 credit per match, with 90% deliverable. That is 247 usable records for 275 credits, about 1.1 credits each.
- Source two sees the remaining 225 and resolves 70 (31%) at 2 credits, 82% deliverable. That is 57 usable for 140 credits, about 2.5 each.
- Source three sees 155 and resolves 22 (14%) at 4 credits, 70% deliverable. That is 15 usable for 88 credits, about 5.9 each.
- Source four sees 133 and resolves 7 (5%) at 6 credits, 60% deliverable. That is 4 usable for 42 credits, about 10.5 each.
Usable coverage climbs from 49% to roughly 64%. The last two sources add about 4 points of that and burn close to a third of the credits. I'll call that stretch the long tail of the chain: where each added point of coverage costs several times the one before. In a waterfall enrichment setup, the long tail is where the money goes.
Finding the break-even point for waterfall enrichment
Break-even is where the cost per usable record from source N equals the expected value of one usable record. Most write-ups say "judge by cost per record" and stop there. The expected value is the half people skip.
Let's say a credit costs $0.10 and, in your funnel, a usable contact is worth about $1 in expected pipeline value (a small chance of a meeting, times what a meeting is worth). Source two costs about $0.25 per usable record, so it pays. Source three costs about $0.59, so it pays, narrowly. Source four costs about $1.05, so it doesn't.
Then subtract the hidden costs. A bounced email costs sender reputation, and a wrong-person match costs a rep a few minutes. Late-chain sources have the lowest deliverable rate, so their true cost per usable record is worse than the credit math shows. My feeling is that most teams who run this honestly cut the chain at three, and sometimes at two.
I don't have a great answer on the value per contact for your business. It depends on your deal size and reply rates, and it is the input that moves the stopping point most.
Provider overlap is why more sources help less than expected
Many data vendors license from the same upstream pools or scrape the same public profiles. Two sources can look independent and return the same record, or the same stale one. Then source two's "match" is mostly source one's data repeated.
You can test for it. Take 200 records that source one missed and send them to each candidate. Count how many each resolves, then count how many of those a different candidate also resolves. If two candidates resolve nearly the same set, you only need one of them. Run the test per segment, because overlap shifts by region, industry and seniority. A source that adds nothing for US software accounts can be the only useful one for European manufacturing.
Order matters as much as count
Put the cheap, high-volume source first and stop the chain the moment a record resolves. Put the source with the best deliverable rate next, even if its match rate is lower, because a verified record is worth more than a raw one. Reorder by segment if your data says to. A single global chain is convenient, for better or worse, and it is almost never optimal.
Check the charging rule too. Some tools charge on a miss, or for the lookup rather than the match. That changes the per-usable math a lot, so get the rule in writing before you build around it. Every waterfall enrichment chain I'd trust has a documented charging rule per source.
How often to re-audit the chain
Data decays, and vendors change their pricing and coverage. My recommendation is to rerun the pilot on a few hundred fresh ICP records every quarter, and any time a provider changes how it meters credits. Drop any source whose marginal usable rate has fallen below your break-even. It takes an afternoon, and it is the cheapest cost control in the stack. If you automate that audit, keep the rule the same: a person approves the change before the chain is rewritten.
Where approval fits
Enrichment spend is spend, and the writes it produces land in your CRM, whether that's HubSpot or Salesforce. At Coco, spend and CRM writes go through an approval gate, so a person sees what a bulk waterfall enrichment run will cost and what it will overwrite before it happens. That is a control, not a claim about accuracy.
If you are working out the economics of contact enrichment, the pricing estimator is a decent place to model credit burn, though estimates carry some error. More RevOps mechanics like this live on the Coco blog.
What I'd do this week
- Pull 500 ICP records and run each source on the full set, not just the leftovers, so you can see overlap.
- Record match rate, deliverable rate and credits per source.
- Compute cost per usable record and compare it to your value per contact.
- Cut the chain where the two lines cross, and put the next audit on the calendar.
Waterfall enrichment is a good idea with a diminishing return. The work is finding your own curve instead of trusting the vendor's.
Sources
- HubSpot developer documentation: API overview — HubSpot
- Salesforce developer documentation — Salesforce
