What to Evaluate in Data Enrichment and GTM Automation: Lessons From $47K in Mistakes

The 20-second version

If your RevOps team is comparing data enrichment and GTM automation vendors right now, three things will actually predict whether the contract pays for itself. Everything else on the demo call is theater.

1. Match rate measured against your own closed-won accounts — not their marketing number.
2. Intent signal recency — how many days old is the signal, and is that number in the contract?
3. Where the automation stops and a human starts.

We cut our contactable universe by 43% and our reply rate went up. That's not a contradiction. That's what a clean pipeline does.

Why you should believe me

I run revenue operations for a mid-market B2B SaaS company. I've owned outbound data pipelines since 2019 — enrichment, verification, intent, sequencing, all of it. Over six years I've personally signed off on 11 significant data mistakes. Maybe 12, depending on whether the Q3 2023 domain incident counts separately, since it was really a downstream effect of mistake #4.

Roughly $47,000 in wasted budget, give or take a few thousand. Plus the internal cost of cleaning up, which is harder to measure and probably bigger.

I maintain a checklist now. This article is the public version of it.

What actually broke, in order of cost

1. Match rate theater

In September 2022, we signed a one-year contract with a data provider advertising a 92% match rate. Their case studies looked legitimate. The demo was smooth. We paid $11,400 upfront for the annual plan.

Then I ran the number that matters: match rate against our own closed-won accounts from the previous 18 months. Out of 340 accounts, we matched 208. That's 61%, not 92%.

Why the gap? Their 92% was calculated against a global contact universe — a set of generic business emails and role-based aliases. Ours needed named decision-makers at companies with a specific headcount band and tech stack. Different question, different answer.

Looking back, I should have demanded a 200-account pilot against our own closed-won list before signing anything. At the time, a case study from a similarly-sized company felt like enough evidence. It wasn't. Case studies are selected. Your list isn't.

2. Intent data that was already cold

Q2 2023. We bought a $8,900 annual intent subscription from a vendor that promised real-time buying signals.

Real-time turned out to mean refreshed monthly. The signals we pulled were averaging 60 to 90 days old by the time they reached our sequencing tool. Some were older.

Why does signal age matter? Because a company that downloaded a comparison guide in February is not the same prospect in May. They've either bought something, changed priorities, or hired someone who did. The window closed while the data sat in a pipeline.

If I could redo that decision, I'd have put the refresh cadence in the contract itself — specific, measurable, with a termination clause. But given what I knew then, real-time sounded like real-time.

3. The catch-all problem

This one cost less money and more reputation.

Our email verification service flagged catch-all domains as valid. Not risky. Not unknown. Valid. So we mailed them.

Our bounce rate climbed to 4.2%. Per Google's bulk sender guidelines, which took effect February 1, 2024, the recommended ceiling for invalid addresses is 2%, and the spam complaint rate ceiling is 0.3%. We were well past the first one. The warning email arrived in March.

I'm not a deliverability engineer, so I can't speak to the DNS and authentication side of this. What I can tell you from an ops perspective is that vendors use valid and deliverable as if they're synonyms. They are not. A catch-all domain accepts or rejects at the server, and no verification tool can know for certain which way it will go.

The fix isn't exotic. Treat catch-alls as a separate segment, mail them at lower volume, and watch the bounce rate weekly. But your verification provider has to actually label them as catch-all rather than lumping them in with verified addresses — and not all of them do.

4. Where the agent stops

We trialed two full-automation tools in 2024. Both worked. Both scared us.

Not because they were bad. Because nobody on my team could explain why a given account ended up in a sequence. The logic was there somewhere, but somewhere is not a place I can defend in a pipeline review.

We ended up standardizing on okki-go. Part of the reason was the enrichment setup — waterfall lookups across multiple sources rather than a single database, which moved our match rate on closed-won from 61% to 84%. But the bigger reason was structural: the okki go ai agent handles the research and the enrichment, and then it stops. A human reviews the list before anything sends.

That sounds less impressive on a homepage. It's more defensible in a quarter-end review.

The okki go outbound research workflow also mattered more than I expected. The agent pulls context — recent funding, headcount shifts, job postings — and attaches it to the account record, so the SDR isn't starting from a blank page. That's a research task, not a judgment call. Different category.

The counterintuitive part

After we cleaned up verification and tightened the ICP filters, our addressable contact list dropped from roughly 12,400 to 6,800.

Reply rate went up. Meeting-booked rate went up. Domain warnings stopped. We sent fewer emails and got more of what we wanted.

Why? Because at 12,400, a meaningful slice of that list was catch-alls, stale titles, and people who had left their company 18 months ago. Every send to those addresses was a small negative signal to the receiving mail servers. The list was big and it was actively working against us.

Here's the thing: I treated more contacts as more pipeline for two years. It isn't. It's more surface area for your domain reputation to take damage.

When this advice doesn't apply

This worked for us. We're a mid-market B2B SaaS company with an ICP of roughly 2,400 accounts, a 6-person revenue team, and predictable quarterly targets. If your situation is different, the math changes.

If your total addressable market is 300 companies, no enrichment vendor is going to fix that. The bottleneck isn't data quality — it's market size. You need a different go-to-market motion, possibly a partner-led one, and no amount of intent signal is getting you out of that.

If you're a seed-stage company with one founder doing outbound, skip the $50,000 stack. Buy a list, verify it, and write the emails yourself. You'll learn more about your ICP in three weeks than any enrichment dashboard will tell you in three months.

And if you're heavily regulated — healthcare, financial services, anything with consent requirements — get your legal team involved before you buy enrichment. I can't speak to that territory. It's not my expertise, and the rules vary by jurisdiction in ways that matter.

One last thing

None of this is a knock on any particular vendor. The vendors we trialed weren't lying. They were answering a different question than the one we were asking, and we didn't catch it until we'd paid.

The checklist I use now is short. Four lines:

• Run the match rate against your own closed-won list before you sign. 200 accounts minimum.
• Get the refresh cadence in writing.
• Segment catch-alls separately and watch bounce rate weekly.
• Keep a human on the final approval. For now, at least.

That fourth one might change. The tooling is getting better, and in 18 months the answer might be different. But as of April 2026, every automation I've seen that removes the human entirely has also removed my ability to explain a failed quarter.