Our leads were already sorted into three intent tiers, a coarse first cut that ranked them by likelihood to convert. It told us who to call first, but not who to stop calling, so we kept calling everyone. Then growth outran the process again: leads up 8.1% while call output fell 16%. Before anyone argued to hire, I asked where the effort was already going.
Seven months of CRM data, 30-day application rates by tier on mature cohorts, split into three acts so a mid-period dialler backlog could not pass for strategy. Tier 3, the bottom, was taking 27.8% of call effort for almost nothing while top-tier leads applied at eight times the rate. We stopped calling it: effort there fell to 0.7% and conversion held, the applications still arriving, just on the leads' timeline rather than the dialler's. The tiers had proved the principle, that intent predicts who converts, but two questions was a blunt way to measure it. That earned the mandate for something sharper: a model that scores each lead, not a bucket that sorts them.