Of the products or services a customer could reasonably buy from you, the share they actually do. A customer buying three of nine lines is at 33 percent penetration.
Penetration = products the customer buys ÷ products the customer could buy
"Could buy" is the judgement. For a store it is the SKU list for its channel. For an institutional client it is the strategies or products that similar clients hold. For a hospital it is the product lines a facility of that type uses.
Unweighted penetration counts products. Weighted penetration weights each by its typical value, so missing the largest line counts for more than missing the smallest. Report which one you use.
Per customer first. Then per salesperson as the sum of products held over the sum of products possible across their customers, not the average of their customers' percentages.
Measuring every customer against the full catalogue, so a convenience store is penalised for not stocking the 24-pack. Counting a product as held on one small order a year ago. Not updating the possible set when the catalogue changes.
Products a customer buys divided by the products relevant to customers of its kind, where relevant means bought by more than half of comparable customers.
A customer buys three of the six lines that most similar customers buy: 50 percent. The three missing lines are valued at the median spend of those who do buy them.
The whole catalogue used as the denominator. Every customer scores low and the gaps include things they could never use.