One of our first eleven trials scheduled a post. What we changed.

BuzzyX went live on 2026-10-14. By 2026-10-17 eleven people had started a trial. I looked at what they had done with it and found that exactly one of them had scheduled a post. Not one a day. One, total, across eleven accounts.

That number is in our decision log for that day, and this post is about what it meant and what we did. No results yet, because the change is three days old. I will write the follow-up when there is something real to say.

What eleven trials looked like

Most of the eleven had done the first thing the product invites you to do: sign in with X, look at the overview, open Compose, try Assist on a draft. A few had published a post straight away. Several had looked at Analytics. That is normal first-session behaviour, and it is what the landing page sells.

What almost nobody had done was queue anything on the calendar. The Schedule page existed, it worked, and the week calendar was right there in the sidebar. Ten of eleven never used it.

That matters because scheduling is the habit that keeps people posting. Publishing one post in a session is a test. Queuing three for the week is a decision to show up. A trial that ends with an empty queue has nothing waiting for the person on day eight, and nothing pulling them back on day two.

Where the lesson came from

I did not work this out alone. Nevo David, who built Postiz, moved its onboarding from "connect a channel" to "schedule your first post" specifically to cut churn, and that is cited in our decision log as the reason we took this seriously. Connecting a channel is setup. Scheduling a post is the actual behaviour the product exists for. Onboarding should end with the behaviour, not with the setup.

BuzzyX's version of "connect a channel" is signing in with X, which everyone does because there is no other way in. So our funnel stopped exactly at the equivalent step: everyone connected, almost nobody scheduled.

Change one, a card on the overview

The obvious fix was a wizard. I did not want a wizard. They are long, people click through them, and they make the first session about the product instead of about posting.

Instead, the overview now shows one card to any account that has never scheduled a post: "Schedule your first post", with a button that opens Compose in schedule mode. The card has a "Not now". If you press it, the card goes away for seven days, and that choice is stored on the server so it holds across your devices. If you still have not scheduled anything after seven days, the card comes back once.

The condition is "no scheduled post ever", in any status. Cancelling a scheduled post does not bring the card back, because at that point you already know how scheduling works.

Change two, a Schedule button inside Compose

The second change was bigger than it sounds. Before 2026-10-17, writing with Assist happened on the Compose page and scheduling happened on the Schedule page. If you wrote something good with Assist and wanted to queue it for tomorrow morning, you had to copy it across. Of course nobody did.

Compose now has a Schedule button for everyone. It reuses the same time field as the Schedule page and the same API, preset to tomorrow at 9:00 in your local time. "Post to X" is still the default, so nothing changed for people who want to publish now. But the path from "Assist gave me a good draft" to "it goes out tomorrow" is one click.

The measurement

We record a first-post event the moment an account schedules its first post. It is decided on the server, under the same lock that protects the post counters, so two parallel requests cannot both count as first and the event fires at most once per account.

Around that we track four things:

• The card was shown.
• The card's button was clicked.
• "Not now" was pressed, after the server saved it.
• A first post was scheduled.

Those four numbers are what I will be reading over the next few weeks. The question is not "did scheduling go up". It is "what share of trials schedule at least one post in their first session, and does that share predict who stays past day seven". If it does, the card earned its place. If it does not, the card goes and we try something else.

What I am not claiming

I am not claiming the card works. Three days and a handful of trials is not data. I am also not claiming that scheduling causes retention. It might simply be that people who were going to stay are the ones who schedule. The measurement above is meant to tell those two apart over time, and I would rather write "we do not know yet" than dress up a guess.

The number I would have wanted

If I had read a post like this before launch, the number I would have wanted is a baseline: what share of trials schedule something in their first session on a product where scheduling is a core feature. I do not have that number from anyone else, and I am not going to borrow one from a blog post that made it up. So ours is the baseline now: one in eleven, on 2026-10-17, before the card and the button. Whatever the next cohort does, it gets compared with that.

Why write this up now

Because the honest version of building in public is the part before the result, not after. Eleven trials and one scheduled post is a small, slightly embarrassing number, and it is the kind of number that most product pages never show you. It changed what we built within a day. That seems worth sharing on its own.

If you are on a trial and you have not scheduled anything, the card will find you. If you want the reasoning behind a posting rhythm before you queue anything, how often to post on X is the guide I would send a friend. And if the blank composer is the problem, what to post when you have nothing to say is where I would start.