The Day My Planning Doc Corrected Me

Altitude: learning from my mistakes

I keep a master planning document for ka-do: the same content register that's also central to why I don't think I've mastered search yet. Strategy, content register, personas, numbers. One long file that everything folds into. And for months, quietly, one assumption sat underneath a lot of it: that my users would look like me.

I'm a contractor. Twenty years of it. Ka-do exists partly because of contractor problems: the meeting-heavy weeks, the context you carry alone, the record-keeping that protects you. So when I modelled the user base, contractors were the centre of gravity. Churn assumptions shaped around bench time. Revenue stability shaped around contract cycles. Landing copy angled toward the person I know best: me.

Then I actually counted my own content register. Roughly eighty pieces aimed at general productivity. About ten aimed at contractors.

The traffic engine I'd spent months building was pulling a general audience, eight to one, while my planning was modelling a contractor majority that my own strategy wasn't even trying to attract. The evidence wasn't hidden. It wasn't in some analytics tool I hadn't checked. It was in my own document, in a table I wrote, sitting a few hundred lines above the assumptions it contradicted.

The class of error

This is worth naming precisely, because "I was wrong about my users" is too vague to learn from. The actual error: I let founder-market fit quietly become a demographic forecast. "I am the ideal customer" is true and useful. It's why the product is honest, why the contractor content has teeth, why the story works. But somewhere along the way "I am the ICP" slid into "most users will be like me," and nothing in between ever got checked, because it never felt like a claim. It felt like background.

That's the signature of this error class: it doesn't live in your conclusions, where you'd audit it. It lives in your defaults. And it distorts everything downstream while looking like nothing at all. My churn model was inheriting a contractor-shaped risk profile, my revenue assumptions a contractor-shaped instability, my landing page a contractor-shaped front door that would have quietly filtered out the majority of the people my own content was attracting.

The test that catches it

The fix, embarrassingly, was already in the file: count what you're actually doing, and check it against what you're assuming. My content register was a census of my real strategy (where the effort was going, who it was aimed at) and the census disagreed with the model. Any founder can run this test in an afternoon: list your assumptions about who your users will be, then inventory what your marketing actually targets. If the two disagree, one of them is lying, and it's usually the assumption, because the inventory is made of work you actually did.

The correction, once seen, took minutes. Contractor stays as the hero story: the credibility wedge, the founder-market anchor, the sharpest segment to message. It stops being the hero demographic. The models get rebuilt on the mix the content actually attracts. The landing page gets a wider front door with a contractor-shaped side entrance, not the reverse.

Why I'm writing it down

Because the data that corrected me was mine, in a file I maintain, and I'd read past it for months. The lesson isn't "gather more information." I had the information. The lesson is that assumptions that arrive as identity (my users are like me) never present themselves for review. You have to drag them out by force, by checking them against the boring inventory of what you're actually building and writing.

The doc knew before I did. I just had to read it as evidence instead of filing. Reading my own evidence instead of filing it is the same discipline behind When Is a Product Ready to Launch?