Main idea

Anticipation is often treated as a faculty of prediction — hence as a quality some teams have and others do not, and that can be neither organized nor verified. Described as a procedure, it loses that character: spot weak signals, draw a stated hypothesis from them, test it early, and learn before others do.

Each of these gestures is verifiable. A weak signal can be named — a recurring friction, a workaround, an isolated but coherent request, an advanced usage among a particular profile, an expectation observed in another tool. A hypothesis is written down. A test is dated and produces a result.

What distinguishes a company that anticipates is therefore not the accuracy of its intuitions, but the delay between a signal's appearance and the moment it knows what to make of it.

Why it matters

This makes anticipation something you can arbitrate. As long as it is divination, it can be neither budgeted nor discussed in committee: nobody defends a prediction. Reduced to a learning loop, it can be compared with other uses of the same team capacity.

It also bounds what is asked of it. A hypothesis tested and invalidated is a result, not a forecasting failure — and that is what makes it possible to launch several.

Nuances and limits

Not every weak signal announces anything: most remain isolated cases, and an organization that treats every friction as a premonition scatters itself. The sorting remains a judgment, which the procedure does not automate.

And learning early is not enough: a company can know before the others and do nothing with what it knows, for want of having decided which risk it accepted.

Open questions

• At what point does a set of isolated signals become coherent enough to justify writing a hypothesis and testing it?