The one-person analytics stack: what I actually check and what I ignore
This update was drafted on a schedule by the AI I build with, from real project notes — part of the vibecoding experiment this blog documents.
Okay, so this is a little ironic. I built CreatorLens because dashboards never actually help you — they show you what happened and then leave you standing there with no move. And then I looked at how I track my own projects and found… a pile of dashboards. Charts I opened, nodded at, and closed.
So I did to myself what CreatorLens does for creators. For every number I was looking at, I asked one question: if this moved, would I do anything differently tomorrow? If the answer was no, it went.
What's left is short. That's kind of the point.
The test: does the number change what I do next?
A solo builder has one scarce resource and it isn't data. It's attention. I said this in the economics post — building got cheap, owning stuff didn't. Every number you check is a small tax on that attention. So a metric only earns its place if a change in it points at an action.
"Traffic is up" doesn't point anywhere. "People sign up and then never get to the first useful result" points straight at the onboarding. That's the whole filter.
What I actually check
Did someone get to the first real outcome? Not signups. Signups tell you the landing page worked. What I care about is whether a new person reached the moment the product is for — for CreatorLens, that's connecting an account and getting an actual recommendation back. The gap between "signed up" and "got the thing" is where most of my projects quietly leak, and it's the one number that most often tells me what to build next.
Did they come back? One visit is curiosity. A second one, unprompted, is the first sign the thing is useful. I don't need a cohort chart for this at my scale. I need to know, person by person, whether anyone returned.
Did anything break? Error tracking, and a way to notice when a scheduled job stops running. I've written about why silent failures are worse than loud ones, and this is where that lands in practice. A failing background job doesn't show up in any traffic chart. It shows up as nothing, which looks exactly like a quiet day.
Did money move? For anything with a paywall: new paid, cancelled, failed payment. Stripe and RevenueCat already tell me this without me building anything. I just have to actually look.
What did people say? Support messages, replies, the odd DM. Not a metric, but it's the highest-signal input I get, and it's the easiest to let pile up unread.
That's it. Five things. Most weeks I can check all of them in a few minutes.
What I ignore
Pageviews and sessions. Fun to watch go up. Almost never tell me what to do. At solo scale they're mostly noise, and they swing with whatever I happened to post that week.
Bounce rate and time on page. I've never once changed a product decision because of these. They're too far from the thing I actually care about, which is whether someone got value.
Follower counts on the project accounts. A follower who never uses the product isn't a user. I watch whether posts bring signups that activate, not whether they bring followers.
Big funnel dashboards. A ten-step funnel is great when you have enough people moving through it for the percentages to mean something. When it's a handful of people, every step is a story about one or two humans, and I'd rather just read the story.
Anything I'd have to explain to myself. If I open a chart and have to think for a minute about what it's even measuring, that's a sign it isn't pulling its weight.
Why fewer numbers works better when it's just you
Here's the part that surprised me. Cutting the list didn't make me less informed. It made me act more. When there are twenty numbers, nothing feels urgent, because something's always a bit up and something's always a bit down. When there are five, and one of them is bad, it's obvious what the day is for.
It also matches what I keep seeing across the whole portfolio. My problem has never been not knowing enough. It's been knowing plenty and not moving. More dashboards feed that. A short list fights it.
When I'd add more
This isn't a forever setup. It's the right stack for projects that have a small number of real people using them. Once there's enough traffic that percentages stop being individual people, a proper funnel and cohort retention start earning their place, and I'd add them then — for that one project, not for all nine. Same instinct as re-deciding the tier when the facts change.
Until then, the stack is: did they get the thing, did they come back, did anything break, did money move, what did they say. And honestly, the most useful "analytics tool" in there is still just reading what people wrote to me.
This one's auto-drafted from my notes on a schedule. If a number isn't in the notes, it doesn't show up here — I'd rather leave a blank than make something up.