← All work

PostHog · 2026

Redesigning and automating PostHog’s customer onboarding

PostHog’s onboarding team gives new paying customers a person to talk to. I redesigned the old automation system behind it, bringing the median wait for a first touch back down from 22 days to 7.

Drawing of three boxes, Sign up, First insight and Paying and staying, joined by arrows
My role

Technical onboarding engineer and owner of the team’s onboarding automation

Team

Onboarding team of three, with a data analyst and the engineer behind PostHog’s customer platform

Timeline

April to October 2026

Outcome

Redesigned the system that onboards PostHog’s new paying customers, bringing the median wait for a first touch back down from 22 days to 7

22 → 7

days a new customer waited for a first touch, June to August, once automated outreach took over

3×

monthly outreach, from 274 conversations in March to 930 in August

1,448

new customer accounts entered the onboarding queue in eight months

Why the first touch mattered

PostHog’s onboarding team gives new paying customers a person who looks at their setup and their bill and helps them get value before the invoices start to hurt. That first touch is where we catch expensive mistakes, like session replay recording every visit, while they’re still cheap to fix.

1,448 accounts entered our onboarding queue in the first eight months of 2026, and three of us worked it almost entirely by hand, researching each account and writing each first email ourselves. My team lead’s advice at my first meetup was to think bigger, because onboarding touches a huge share of PostHog’s customers.

The system behind it had grown by patches: five overlapping workflows and seven to nine account rules decided who got onboarded, and accounts fell through the gaps between them. An account that joined in September 2025 waited a median of 58 days for its first email, and only 19 of 296 heard from us inside a week.

And I had to fix it next to a full queue of my own accounts. In May the company set a mid-October deadline to leave our old CRM, so everything had to be built twice, the second time on PostHog’s own tools, which were still being built. I also had a month of leave in September, so it had to run without me.

What I redesigned

Before: five entry workflows and every step done by hand. After: one entry workflow feeding an automated, tracked email sequence.BEFORE: A MANUAL PIPELINEFive entryworkflows,7 to 9 rulesSpecialistresearcheseach accountFirst emailwrittenby handStage movedby handAFTER: ONE AUTOMATED SYSTEMOne entryworkflow andone ruleAutomatedfirst email,AI audit linkThree trackedfollow-upemailsFailed paymentsand close-outautomated
Before the redesign every step needed a person. After it, one entry workflow feeds an automated, tracked email sequence.

One way in instead of five

On 8 April I mapped every workflow, and the next day I replaced the five entry workflows with one and the seven to nine rules with a single computed one. One mistake now hits every account, which is why the map came first.

A first touch nobody has to write

On 24 July I launched automated first outreach: an email with a link that opens PostHog AI inside the customer’s own project, with an account audit prompt already written.

I launched it only to accounts forecast at $500 to $550 a month, 42 on day one, trading weeks of coverage for a batch small enough to read through and fix. The first weekday batch sent 33 emails and closed out 20 accounts with nobody touching them, and the day after working 17 failed-payment accounts by hand, I automated those too.

Measuring what customers actually do

Calls booked would have told us almost nothing, since in September we ran about one call for every 90 outreach conversations. So in early August I shipped a three-email follow-up sequence with tracking on every link and a dashboard per email.

The first version undercounted, because links into the app lose their tags at the login redirect, so I rebuilt the matching and wrote the method into the team handbook.

Proving impact without fooling ourselves

My first dashboards showed onboarded accounts with a great revenue lift over self-serve ones. Working through it with Abhishek, one of PostHog’s data analysts, we traced that to selection bias in how the two groups were built. What held up was product adoption, 4.68 products on average for onboarded accounts, so that became the team’s provable impact.

Moving the whole system onto PostHog

From August I rebuilt the whole pipeline on PostHog’s own tools with the engineer who owns them. A pre-launch audit of all 14 workflows caught one that would have marked every account as done the moment it was switched on, and on 15 October we switched the old system off, a day before the deadline.

What happened

The wait for a first touch is a team result, and it moved up and down before landing at 7 days for August entrants, the first full month of automated first outreach. Across the year, the team brought it from 58 days to 7. 12.9% of accounts that entered from January to August were never contacted at all.

Median days from entering the onboarding queue to a first touch: 58 for September 2025 entrants, then 2.5, 4, 4, 6, 7, 22, 18 and 7 for January to August 2026.MEDIAN DAYS FROM JOINING THE QUEUE TO FIRST TOUCH582.5446722187SepJanFebMarAprMayJunJulAug20252026Redesign startsFirst full monthof automated first touch
Median days from joining the onboarding queue to a first touch, by month of entry. A team result that moved up and down along the way.

Monthly outreach went from 274 conversations in March to 930 in August while the reply rate mostly held at 8 to 10%. Automated emails got replies less often, 6.7% against roughly 10% for my teammates’ mostly hand-written ones, so automation bought reach at some cost per email.

Week of 17 August 2026: 65 automated onboarding emails and 13 by hand. Reply rate from June to mid-September: 6.7% for automated email, 10.3% and 10.1% for two teammates writing mostly by hand.EMAILS I SENTWEEK OF 17 AUG 2026REPLY RATE BY SENDERJUNE TO MID-SEPTEMBER65136.7%10.3%10.1%AutomatedBy handAutomatedTwo teammatesmy addressmostly by hand
By August automation carried most of the onboarding email I sent, and it got replies less often than email written by hand.

The graduation email didn’t work. It got close to zero replies and its links carried no tracking, so the team replaced it in August.

“Dan is a true executor… a role model for less tenured members.” Magda Olszewska, onboarding team lead

What I’d do differently

I’d set up a holdout group on day one. We only agreed a 10% random holdout in October, after a dashboard that flattered us had already answered the first impact question.

I’d count automated sends from the first email. For a month our weekly update counted only hand-written mail and under-reported outreach about five to one.

And I’d pair much earlier. In September Magda pointed out that if I broke my leg, nobody else on the team could run the setup, and she was right.