What I argued in 2024
The claim, in short, was that self serve onboarding would be cheaper and would let us scale without growing our customer success team. I had data from a handful of SaaS companies I admired. Linear, Vercel, etc. They had no humans in their onboarding loop. Why couldn't we be like them.
I was pretty confident. I told the team we should do this. We spent Q4 2024 and Q1 2025 building it. Beautiful flow. Sandboxed test data. Inline checklists. Help articles right next to the relevant screens.
What actually happened
Time to first value, which had been about 11 days under the old guided onboarding, went to 31 days under self serve. Then customers started churning at month four at a rate we hadn't seen before.
I want to be careful here because I don't have a clean explanation. The thing I told myself for about three months was that the customers were the problem. They didn't engage. They didn't watch the videos. They weren't following the checklists.
Sarah, our head of CS, finally told me directly in a 1:1 in April that the customers were not the problem. The product was the problem. Specifically, the product was complex enough that it needed a human to walk a real property manager through it the first time. Not because the property manager was dumb. Because the product had too many cross dependencies and the order in which you configured things mattered and a flow chart could not adequately explain it.
She was right. I had been making the argument from the wrong direction for about six months.
What we did about it
We brought back guided onboarding in May 2025. We kept the new flows for the simpler accounts (under 500 units). For everything above that, a real human walks them through. Time to first value is back to about 9 days. Churn at month four normalized in about 60 days.
The lesson, which I think is the actual lesson, is that "you should be more like Linear" is almost never the right answer when your product, customer base, and team are different from Linear in every way that matters. Linear sells to engineers who are paid to figure things out. We sell to property managers who are running ten other systems and don't have an hour to watch our intro video.
I should have known this. I think I did know it. I just liked the version of the world where we could scale without humans, and I went looking for evidence to support that version.
What's still true from the original post
I think the part about self serve documentation was right. Our help center traffic has tripled and customers do find their own answers more often. So the documentation work was not wasted. Just the part about removing humans from the onboarding loop.
Anyway. Wanted to say so publicly because I told a lot of people this opinion confidently in 2024 and some of them probably remember.