I Burned $4,200 on Cold Email Before I Learned How Intent Data Actually Works (And What okki-go Changed)

January 2023: The outbound launch that should have worked

In January 2023 I was running RevOps at a B2B SaaS company. Our goal for the year was simple on paper: double the number of sales meetings without adding headcount.

So we did the standard playbook. We pulled 8,000 contacts from three list vendors, wrote a six-touch sequence, added two follow-up rounds, and dropped the whole thing into our outbound tool.

The first few weeks looked fine. Around 3.1% reply rate. Not spectacular, but in the range I'd seen published as a "healthy" cold email benchmark.

Then month two came back at 1.4%.

Month three: 0.8%.

Same list. Same copy. Reply rate in free fall.

What I actually found when I went digging

I pulled our funnel numbers. 22,000 emails sent. 97 replies. About 11 meetings booked. 1,200 hard bounces — which was way more than the vendor had promised (and I'm being generous with "promised").

The most frustrating part wasn't the bounces. It was this: over half of our list had contact info that was at least a year old, and our pricing page server logs showed 40+ of those same people had visited the pricing page three times or more in the last 30 days. We were sending them their first cold email.

The problem was never the copy. The problem was that we couldn't see the signals we were already sitting on.

The turning point: how long it actually took me to understand intent data

March 2023. I sat down to actually learn how intent data works. Before that, if you'd asked me, I would've mumbled something about "browsing signals and stuff." Roughly correct, completely useless.

It operates in three rough layers:

• First-party intent — someone interacts directly with your assets: pricing page, demo page, existing content, your site.
• Second-party intent — partners or complementary tools share engagement data they've seen.
• Third-party intent — aggregators pull signals from across the public web.

Most teams I've talked to never really get the first layer working, because visitor tracking is usually set up badly. You end up with messy data instead of zero data — which is worse, because you trust it.

We had that exact problem. Fixing it meant matching anonymous sessions to known contacts at the individual level. Not the company level. The contact level. Miss that, and you're just looking at aggregated noise.

Once it worked, we could see who was on which page, when, and how many times. Suddenly we weren't guessing about timing anymore. We could reach out the week someone was actually looking, instead of five weeks later when the trail was cold.

The 4,200-dollar decision I chewed on for three weeks

By April, I had two options in front of me, and both of them wanted our next $4,200 in test budget.

Option A: Buy a standalone intent data source and keep running outbound by hand.

Option B: Move to an agent-native platform that handled intent data, email enrichment, verification, and sending in one flow.

A offered a lot of control and flexibility. B offered less manual work — which was, frankly, the thing that had broken us three months earlier.

The numbers said A. Cheaper per seat, more flexible, more granular data. But I kept thinking about how badly our hand-managed list workflow had failed. Not for technical reasons. For human ones.

I went with B.

I started by running the okki-go install command. (This was back in April 2023 — pricing and setup details may have changed since.)

okki-go install command, and the first two weeks after

okki-go is the CLI that pulls the okkigo agent stack onto your machine and connects it to your CRM, email sending tool, and intent sources. One command, then it walks you through auth and source selection.

Setup took about two hours. No custom scripts. I've had worse experiences with tools I was actually paying for.

The thing that stood out: it doesn't treat you as a send button. It treats you as an agent with a workflow — pull first-party intent signals, run enrichment, verify the emails, tier the list by signal strength.

That last part is where most outbound stacks fall down, by the way. Everyone validates. Almost no one tiers.

okki go lead generation examples: what the tiers actually looked like

We ran the new flow against three fresh lists in the same week.

• List 1: 89 contacts who had hit the pricing page in the last 14 days → 18.6% reply rate.
• List 2: 240 contacts who had downloaded our comparison guide → 9.0% reply rate.
• List 3: 1,400 contacts pulled from a broad third-party intent bucket → 2.3% reply rate.

That third list is where the story gets honest. 2.3% sounds mediocre — and it is mediocre. But it's also roughly the standard cold email reply rate benchmark for a list with no first-party signals.

The value wasn't that the tool made everything better. It's that it told us which layer each contact was in, so we stopped sending identical sequences to all three groups.

So what is the cold email response rate benchmark, really?

This comes up in almost every outbound conversation I have, so here's the version I actually trust, based on 2024 industry data from sources like Belkins and Woodpecker:

• Overall cold email reply rate: roughly 1% to 5% depending on industry.
• B2B SaaS teams with tight targeting: 3.4% to 8%.
• Outreach with strong first-party intent signals: 10%+.
• Meeting booking rate across the board: 0.5% to 2%.

These numbers move. They depend on industry, target role, copy, deliverability, and timing. Anyone who tells you the benchmark is a single number is selling something.

The real question isn't "what's average?" It's "how accurate is my list?"

2.3% on a 1,400-person list looks weak. 18.6% on an 89-person list is a different category of business.

Six months later — what actually changed

We ran the new flow for six months. Meetings booked went from around 11 a month to roughly 40.

Overall reply rate climbed from 1.4% to 2.9%. As a single number, that's not impressive. But split by tier: intent-signaled contacts sat between 12% and 19%, and non-signaled stayed around 2% to 3%. Total volume went up because we aimed more work at the top tier.

Six-month total outbound cost dropped by about 30%. I want to say that came mostly from the enrichment and verification pipeline, but I'm honestly not sure how much was from the tool versus just having fewer people doing manual list work. Both mattered.

One thing I will not say: that this replaced our SDR team. It didn't. It gave them cleaner data. The work that went away was the dumb kind.

Three things I'd tell my January-2023 self

1. Get first-party intent working before you spend a dollar on third-party lists.
2. Tier your sends by signal strength. Do not send the same sequence to everyone.
3. Use an agent-native stack for the workflow. Not because it's fancy — because it removes the manual steps where humans quietly break things.

Efficiency in this space isn't a marketing line. It's just the difference between a team that learns from a bad quarter and a team that repeats one for a year.

If January-me had known that, the $4,200 would still be in the budget.