The constraint
Peoplelogic builds performance management software. Beyond the standard review processes, the team is working out how to bring AI into HR tech responsibly: to improve the output, not to replace people's jobs. At the time it was a 15-person company with about 100 clients, growing every month.
Sales had a working guess at who the buyer was. It started conversations, but nothing repeated. Finding a pattern takes volume and fast A/B testing, and Peoplelogic was not set up to do either at speed.
I was hoping to just better understand our ICP. We had a guess, but we had not found one that had been very, very repeatable in that we could get some conversations going, but we weren't really finding anything common thread.
What we built
An outbound program designed to answer one question: who actually buys.
Work started in early November 2023, and leads were coming in that same month. The campaigns were cut so they could be compared against each other: managers against directors, and a culture, engagement and happiness angle against an HR operations and systems angle. Lists were pulled from Apollo, every campaign ran a five-email sequence, and new segments went live as results came back, in December, late January and late February.
Austin measured two conversions: responses to booked meetings, and meetings to opportunities. His feedback on which conversations were working went straight back into the targeting.
In March 2024 we analyzed the positive replies: 171 of them, from 18,099 people contacted.
The title was the signal
86% of positive replies came from people with culture, employee experience or employee engagement in their job title, a group that made up 46% of the list. "Culture" alone accounted for 75%.
That group replied nearly twice as often
One positive reply every 56 contacts, against one every 105 across the whole list.
Size and seniority narrowed it further
77% of positive replies came from companies with 10 to 250 employees, and managers and heads accounted for 70% of them.
Two emails did the work of five
82% of positive replies landed on the first two emails, so the recommendation was to cut sequences to three touches.
Industry did not matter
No industry stood out.
Day to day, both teams worked in one shared Slack. Across very different time zones, Austin's estimate is that 85 to 90% of the time the answers came as fast as if the team sat in his own.
You treated it like, I don't know, like I was your manager and you guys were my demand gen team working for me in a company.
- Apollo
Results
| Metric | Result |
|---|---|
| Meetings a month | 100 |
| Responses booked as meetings | 30 to 40%, in Austin's estimate |
| Positive replies by March 2024 | 171 from 18,099 contacted (1 in 105) |
| Share from culture and employee experience titles | 86%, from 46% of the list |
| Positive reply rate in that group | 1 in 56 |
| Positive replies across the engagement | 314 from 49,550 people reached (1 in 158) |
So we were just hammering meetings which was teaching us a lot about who these people were and what they are and helping us to narrow down and give you guys feedback on which ones were working and which ones didn't.
What this means for you
If your ICP is still a guess, sending more of the same guess will not fix it. Run enough volume, split cleanly enough, that the replies can tell you something. Peoplelogic's answer turned out to be a job-title keyword, a company size band, and the finding that two emails did the work of five. None of that was visible from a handful of conversations.
