Decide what the number is before you rebuild anything

Every stalled growth project we have inherited has the same shape: a site or a campaign that four people have opinions about, and no agreed way to tell which opinion is right. Iteration turns into taste, and taste does not converge.

The fix is unglamorous. Before anything gets rebuilt, write down the one number the work is meant to move, where it sits today, and what a month of it is worth in money. Take it from the analytics you already have rather than from someone’s benchmark deck.

Once that number exists, changes stop being debates. You ship, you read the delta, and you keep the version that scored better. The argument moves from what looks good to what measures better, which is the only argument that terminates.

It also gives you something to put in front of whoever signs the invoice, and a tripwire for the week a platform quietly changes how it sends you traffic.