My onboarding was a lie
Photo by Oleksii Rozanov on Unsplash
Hey friends,
Luca here.
I recently decided to change my approach to my SaaS, BlackTwist.
For the past year, Mattia and I have been heads down building features, shipping updates, and growing the user base. And that worked for a while. But the numbers started telling a different story. Signups were steady. Revenue was flat. Something in the middle was broken.
So I stopped building new things and started reading the data. I wanted to understand where users were getting stuck, which steps they were skipping, and what I could actually fix to move the numbers that matter.
What I found changed how I think about our product.
A few weeks ago, I opened PostHog and looked at a number I’d been avoiding.
312 people signed up for BlackTwist in the last 30 days. 64 of them created their first post. That’s 20%.
Eighty percent of new users signed up, looked around, and left.
I knew activation was a problem. I just didn’t know where the wall was. So I sat down, built funnels, watched session recordings, and let Claude Code inspect our onboarding page through the DevTools MCP. What I found was uncomfortable. And useful.
Let me walk you through it.
The first thing I did was build a funnel in PostHog. From signing up to the social account connected to the first post created. The results told me everything.
312 signups. 129 connected their Threads account. 64 created a post.
The biggest drop was from step one to step two. 59% of users never connected their social account. They signed up, saw the onboarding, and bounced before doing the one thing that makes the product work.
I dug deeper. I built a second funnel just for the onboarding flow itself. Goal selection, posting goal, reply goal, reminders, done. 348 users started it. 318 finished. 91% completion rate. The onboarding flow was fine. People were clicking through in about 10 seconds per step.
The problem was what came after.
After completing onboarding, users needed to connect their Threads account. That’s where 59% disappeared. The flow asked too much too fast. “Give us access to your account” right after you just met the product.
So the onboarding was easy to complete but it was doing nothing useful. People were answering questions about their goals and setting reminders before they ever connected a social account or scheduled a single post. The flow felt productive. It was actually a dead end.
Here’s where it gets interesting. I pointed Claude Code at our onboarding page using the Chrome DevTools MCP. Claude could see the actual page, the DOM, the layout, the buttons. I asked it to analyze the flow and tell me what a new user would struggle with.
It confirmed what the data already showed. Too many steps before the first moment of value. The user sets goals and reminders before connecting their account. By the time they reach the connection step, they’ve spent their patience on steps that feel like paperwork.
I rebuilt the onboarding.
Fewer steps. I cut the flow down so users get to the important part faster. The “connect your Threads account” step now happens earlier, when motivation is still high.
Then I added post templates. After connecting, users see templates they can use for their first scheduled post. Instead of staring at a blank text box, they pick a template, edit it, and schedule it. The friction from “I signed up” to “I just scheduled my first post” dropped significantly.
After publishing, I added a celebration modal. A small moment that says “you did it.” It sounds silly. It works. Then the modal sends them to the calendar view where they set up their time slots. So by the end of their first session, they have a connected account, a scheduled post, and configured time slots.
The old onboarding was completed in 91% of cases and produced nothing. The new onboarding has more resistance in it, but every step produces something real.
I haven’t been running this long enough to share before/after conversion numbers. That’s a future newsletter. What I can share is the process, because it applies to any product.
Open your analytics. Build a funnel from signup to the one action that makes your product useful. Find the step where people disappear. That’s your wall. Everything else is a distraction.
I used PostHog for the funnels and session recordings. I used Claude Code with the DevTools MCP to get a second pair of eyes on the actual page. The combination of data and AI review caught things I missed after staring at the same onboarding for months.
If you’re building something and your activation rate is low, the answer is probably in your data already. You just have to look.
Still here, still building.
Luca
