Pricing and Margin Model

Estimated difficulty: 3/5 | Estimated time: 90 min

Find the usage level at which your pricing stops working. Put a seat price, a credit price and a per task price against a cost model that scales with use, and print gross margin per customer across the whole distribution of how much people actually use the thing.

Why it matters#

Software pricing evolved where the marginal cost of a user was zero, so a seat price was safe. With a model in the product, your most engaged customers are your most expensive, which is the opposite of the correlation every seat price assumes. The failure is slow: margin erodes toward your best customers while product strategy pushes engagement up, and the two effects hide each other for several quarters.

What to build#

1. A usage distribution. Not an average. Use something long tailed, because that is what usage is, and the average customer is not where the money or the cost lives.
2. Three pricing models: a flat seat price, a credit bundle with overage, and pure per task.
3. A cost model per customer that uses your cost per task from project 02.
4. Margin per customer across the distribution, printed as a table by usage decile. Find the decile where margin goes negative under each pricing model.
5. The blended answer. Print overall gross margin for each pricing model, and the share of customers who are unprofitable under each. These two numbers frequently point in opposite directions, which is the whole lesson.

What good looks like#

• You can name the usage percentile at which a seat price stops covering cost.
• Credits with overage look better on blended margin and worse on customer experience, and the program makes that tradeoff numeric rather than aesthetic.
• Changing cost per task by a third visibly moves which pricing model wins.

Stretch#

Add a fall in model prices over time, since token prices have fallen every year so far, and ask who captures that: you, through margin, or the customer, through a price cut. Then add a cap on the heaviest users and see how much of the margin problem a cap solves without changing the headline price at all.