60 Signups, 74% Activation, 8% Week-One Retention: What Our Own Data Actually Said

We Went Looking for a Signup Problem and Found Something Else

At the end of July we had about 60 new accounts for the month. That is a small number, and small numbers are the best time to look closely, because you can still reconcile every single row by hand.

So we did. We queried production for aggregate account metadata — creation timestamps, referrer, whether a note was ever created, what kinds of objects it contained, and when it was last edited. We did not read note titles or note content. The whole point was to answer a product question, not to browse anyone's notes.

We expected to find a leaky signup form. Instead we found that the signup form was one of the healthiest parts of the funnel, and that the real problems were on either side of it.

Reconciling the Number First

The dashboard said 60. The database said 58 rows. After removing one automated health-check account, we could reconcile 57 real signups in Pacific time, or 58 if you draw the boundary in UTC.

This sounds pedantic. It is not. If you cannot explain the gap between two of your own numbers, you cannot trust any conclusion you draw from either of them. Timezone boundaries and monitoring accounts are the two most common reasons a growth dashboard disagrees with a database, and both were in play here.

Working number: 57.

Activation Was Genuinely Good

Of those 57 accounts:

• 45 signed up with Google, 12 with a password
• 42 created a real note beyond the welcome canvas — 74% activation
• 38 of those 42 did it within five minutes of signing up
• 23 created at least two notes; seven created five or more
• The median user created exactly one note

Seventy-four percent activation is a number most products would be happy with. And the five-minute detail matters: people who activate at all do it almost immediately. There is no slow-burn cohort quietly warming up. You either get someone in the first five minutes or you do not get them.

What did they actually do on the canvas?

• 26 used canvas objects — mostly text, rectangles, images, and freehand drawing
• 23 used markdown blocks
• 12 used sticky notes
• Only seven touched the AI or media features
• Four used folders, two used tags, one collaborated, one saved a recording

That last group is the interesting one. The features we talk about most in marketing — AI meeting notes, recording, collaboration — were used by almost nobody in their first session. New users draw boxes and write text. Everything else is a second-week behavior, if it happens at all.

Retention Was Not Good

Here is the part that actually mattered.

• 7 of 41 eligible activated users edited a real note after day one — 17%
• 2 of 25 eligible users edited one after day seven — 8%
• Only two people created notes on more than one calendar day

These are deliberately conservative, lower-bound figures. They measure saved content changes, not opening the app, not reading an existing note, not scrolling around a canvas. Someone could have opened OmniCanvas every day for a week to reread their meeting notes and would not appear in these numbers at all.

Even granting that, the shape is unmistakable. People arrive, they get value once, and they do not come back to build on it. A note-taking app that gets used once is not a note-taking app. It is a scratchpad.

Where Signups Came From, and Which Ones Were Worth Having

Google sends the most people. ChatGPT sends the best people. Ten signups from ChatGPT produced a 90% activation rate, and 60% of them created multiple notes, against 39% for Google traffic.

The sample is tiny and we are not going to over-read it. But the mechanism is plausible: someone who asks an assistant "what should I use for spatial note-taking" and gets a recommendation arrives with intent already formed. Someone who lands on a listicle from a search result is still shopping.

The Actual Leak

Then we looked at the marketing site. About 11,668 distinct visitors over the period.

• Homepage: 433 sessions, 59 sessions with a CTA click — 13.6% clickthrough
• Blog posts: 11,001 sessions, 21 sessions with a CTA click — 0.19% clickthrough
• Signup screen: roughly 70 recorded actors produced 57 accounts
• Password signup: 12 completions out of 13 attempts

Read those numbers next to each other. The homepage converts at seventy times the rate of the blog. The signup form converts at over 80%. And the blog — which is where essentially all of our traffic lives — converts at almost nothing.

Individual posts made it worse:

• /best-whiteboard-apps/ — 1,824 sessions, zero CTA clicks, zero attributed signups
• /omnicanvas-vs-obsidian-canvas/ — 866 sessions, zero clicks, zero signups
• /best-free-infinite-canvas-apps/ — only 365 sessions, but nine signups and eight activations

Our single highest-traffic article produced nothing. An article with a fifth of the traffic produced nine accounts. The difference is not traffic quality in some mystical sense. It is that one of those articles asks the reader to do something they were already trying to do, and the other one ends with a paragraph about how there are many good options.

There is one hopeful signal: blog CTR reached 1.21% during July 22–28, up from 0.07% earlier in the month. Recent CTA work may already be helping. That needs another two weeks of clean measurement before we believe it.

What We Concluded

Three things, in priority order.

1. The blog-to-product transition is the whole ballgame. If the 11,001 blog sessions converted at even the homepage's 13.6%, the arithmetic gets absurd. We do not need more traffic. We need the traffic we have to be able to do something at the end of the article.

2. The signup form is not the problem, so stop optimizing it. Twelve of thirteen password attempts completed. There is nothing left to win there, and every hour spent on it is an hour not spent on retention.

3. Retention is the real product problem. Activation at 74% and week-one retention at 17% describes a product that demonstrates its value and then fails to become a habit. No amount of funnel work fixes that. It is a product question about what brings someone back on day two.

What We Did About It

The immediate change was to stop sending blog readers to a signup wall. If an article is about whiteboarding, the link at the end should open a whiteboard — the real one, no account — and ask for an account later, once there is something worth saving.

We shipped that on July 28, along with intent-specific CTAs on the highest-traffic whiteboard and Obsidian articles, meaningful-work detection, save-your-work prompts, and a second-note loop that waits until someone has actually done something before suggesting what to do next.

We also wrote down the whole funnel and a reusable cohort query so this analysis can be rerun in one command instead of one afternoon. That turned out to be the most valuable artifact of the day.

We will publish the after-numbers. Including if they are bad.

If you want to see what the reader now lands on instead of a signup form, open a blank canvas with no account required.

30 days free. No credit card required.