Main idea

The questions that land on a Product Manager's Slack all look alike in form — "can we do this?", "what's the limit?", "can we promise this to the customer?" — and yet they split into two families that have nothing in common.

Some ask you to understand a market, a positioning, a direction; they call for judgment, and sometimes for a deliberate no. The others are about a rule, an exception, a constraint, an interaction between two features: they have a factual answer, already written somewhere in what is running. The first family gets decided, the second gets observed.

Confusing the two costs in both directions. A PM's judgment is mobilized where reading would have been enough — and the PM becomes a bottleneck for facts. Conversely, a strategic direction is sometimes treated as a simple check of what the product already does, and the answer comes back as "no, that isn't planned" where the real task was to decide whether it should be.

"Getting observed" doesn't mean "easy to read". The answer exists in what is running, but rarely in a form you can query: a rule such as "an invoiced order takes no further line" may be written nowhere, its enforcement scattered across five files, each carrying only a piece of it. Observing then means reproducing the case in a test environment. So the second family of questions isn't the easy one: it is the family whose answer is determined, and whose access stays expensive as long as the product's model is written nowhere.

Layer added by "I wrote a product's ontology. Three times, I thought I was done." (2026-08-11).

Why it matters

This gives a sorting criterion to apply before answering, and that sorting decides who should handle the question. What gets observed can be delegated to a system, provided it reaches the source; what gets decided cannot be, whatever the tooling.

It also changes how a PM's workload is read: a queue of requests isn't a sign of being at the center of decisions — it may be nothing but a volume of readings nobody else knows how to perform.

Nuances and limits

The split isn't always clean on arrival. A behavior question sometimes hides a strategic one — "can we promise this configuration to this prospect?" is a question of fact as long as the configuration exists, and becomes a judgment call the moment it would have to be built.

And a factual question with no answer in the source isn't thereby a strategic question: it is a hole, and it calls for a test or a decision about coverage.

Open questions

• Who, in an organization, performs this sorting at the moment the question is asked, before it lands on the PM by default?