Fit and intent, scored separately.

The lead scored 87.
Nobody knew what to do.

Firmographic and technographic criteria weighted by what closed-won accounts actually look like, with the disqualifying attributes named explicitly.

Behavioural signals weighted by what genuinely preceded closing — not every tracked action, only the predictive ones.

Which data sources fill which fields, at what cost per record, and what happens when enrichment fails or conflicts.

What a rep does at each combination of fit and intent, with routing and SLA attached so the score drives behaviour.

The model tested on held-out closed deals before launch, so its predictive value is measured rather than assumed.

A quarterly refresh against outcomes, with drift reported, so the model tracks your business as the mix moves.

Volume exceeds capacity, so the question is no longer whether to follow up but in what order.

A model exists, reps do not use it, and nobody has checked whether it predicts anything.

SDRs working leads that were never going to buy, which is expensive and demoralising in equal measure.

Usage signals available and unused, which are usually far more predictive than firmographic fit alone.

Paying for data enrichment with no evidence it improves conversion or routing.

Moving upmarket, and the old scoring criteria now describe accounts you no longer want.

Every signal tested,not the ten someone guessed.

Dozens of candidates, measured

Every available firmographic and behavioural signal tested for correlation with closing, so the weights come from evidence rather than from a room full of assumptions.

Enrichment reconciled at scale

Records matched and deduplicated across enrichment sources, with conflicts surfaced rather than silently overwritten.

When the model stops working

Ongoing comparison of predicted against actual outcomes, so decay is reported the quarter it starts rather than noticed a year later.

What you refuse to sell to is a commercial decision.

A model finds correlation. Which segments you want, and which you decline regardless of intent, is strategy — and it is set by your leadership rather than inferred from last year’s wins.

When scoring
adds nothing

If capacity exceeds volume, prioritisation saves nothing. Call everyone and spend the effort elsewhere.

Below roughly a hundred closed-won deals the correlations are noise. Firmographic rules based on judgement are more honest.

A score with no agreed action per band is ignored within a month. The routing commitment matters more than the model.

Without behavioural tracking or reliable firmographics, a model scores data completeness rather than account quality.

Whether high-scoring leads actually convert better. If the bands do not separate, the model is not working.

Whether the agreed action per band is actually taken. A model nobody follows predicts nothing useful.

How much better the model is than working leads in arrival order. This is the honest measure of its value.

Whether declined leads genuinely would not have closed, checked periodically so the model is not silently excluding good accounts.

Share of records successfully enriched and the cost per record, against the conversion lift it produces.

Predicted against actual over time. Rising divergence means recalibration is due rather than optional.

What scoring
sits alongside