Eureka Moments
Case Study  ·  Sales Strategy

How Much Could Each Account Be Worth?

A B2B distributor needed reliable revenue potential for 230,000 accounts. The answer required two models working in tandem.

For any sales organization, prioritization is everything. The question of where to focus hinges on knowing what each account could realistically spend — not what they're spending now, but what they would spend if fully developed. That number is unobservable by definition, and the accounts where you most need it are usually the ones where you know the least.

Two Paths, One Score

We built a two-path opportunity scoring framework covering 230,000 accounts. The first path applied to accounts with transaction history: a multiple-imputation model that filled in each account's purchasing matrix based on the patterns of similar customers — what they would likely buy across categories they hadn't yet purchased. The second path handled the 55% of accounts that lacked sufficient history: a firmographic model using third-party practice data to estimate potential from size and characteristics, calibrated against accounts where both data sources existed.

230,000 ACCOUNTS PATH 1 · 85K ACCOUNTS Cross-product imputation PATH 2 · 145K ACCOUNTS Firmographic model One score PER ACCOUNT
Two paths, reconciled into a single defensible score for every account in the universe.

Multiple imputation runs many models and simulations, then averages the results — producing a stable, uncertainty-aware estimate rather than a single brittle prediction.

Guardrails Make the Score Defensible

Raw model outputs are rarely usable as-is. Imputation in particular can produce outliers at the tails. The final scoring logic applied percentile-based guardrails — estimates below the 5th percentile were floored, above the 99th were capped — and historical sales for each account provided an absolute floor. The score could never fall below what the account had already shown it was willing to spend.

230K
Accounts
scored
85K
Scored via
cross-product model
2
Independent models
reconciled into one score

The scores fed directly into territory design, segmentation, and rep-level quotas. Accounts that looked small on current revenue but scored high on potential could be elevated for outreach. Accounts with high current revenue but limited remaining upside could be managed more efficiently. The model didn't replace judgment — it gave judgment something rigorous to work from.

Building a sales potential model for your account base? Say hello.