A tier nobody can explain gets ignored by the person meant to act on it. A tier with a reason attached gets worked, because the rep can see whether they agree.
Missing data is not a zero
The quiet failure in most scoring is a dimension the enrichment did not populate being scored as an absence rather than as an unknown. An account with no data on four dimensions then lands in tier C, and looks identical to an account that genuinely scored badly on all four.
So the run checks coverage per dimension first and re-normalises the weights over what it can actually populate, and the reason line says which dimensions were unavailable. Tier C for a real reason and tier C for missing data are different answers.
Fit and signal do different jobs
Fit is durable: size, industry, structure, stack. Signal is perishable: a raise, a hire, a leadership change. Averaging them into one number hides the only thing you actually need to know, which is whether this is a good account you should work patiently or an ordinary account you should work this week.
Signal can promote an account, but only by one step. That cap exists because an unlimited signal promotion turns the model into a signal alert with extra steps.
It applies the model, it does not write it
Defining the disqualifiers, the dimensions and the bands is a separate piece of work, done once and calibrated against a real sample. This skill only applies what that produced, and never adjusts it mid-run. Scoring that tunes itself against the list it is scoring is not scoring.
Runs well with
Skills compound. These are the ones we usually install alongside it.