Sample frameDay 116·Finance · Allocation

Good morning. Somewhere on your customer list is an account everybody is proud of. It costs more to keep than it brings in.

Finance.

Cost to Serve

5 min read·Apply by lunch

The question

Pick the customer you'd fight hardest to keep: could you say, within a few thousand dollars, what it costs to serve them once the calls, the returns, the bespoke reporting and the chasing of a payment that always lands late are counted? Almost nobody runs that number, and the obstacle is rarely the arithmetic — it's that the account in question has the most impressive name on the list.

The idea

In September 1988 Harvard Business Review published 'Measure Costs Right: Make the Right Decisions', by Robin Cooper and Robert Kaplan, both then at Harvard Business School. It arrived a year after Kaplan and H. Thomas Johnson had argued in Relevance Lost that management accounting had stopped telling managers anything they could act on, and it offered the repair. Stop spreading shared costs across products using a single volume-based measure — direct labour hours, machine hours, revenue — and instead find the activity that actually causes each pool of cost, then charge it out on that. The consequence is uncomfortable and remarkably consistent. Volume-based allocation over-costs the simple, high-volume work and under-costs the small, complicated, customised work, which means your standard product has been quietly subsidising your special ones for years. The method became known as activity-based costing. Its most useful application turned out not to be product costing at all, but pointing the same lens at customers — and finding that profit isn't distributed anything like revenue is.

The framework

01
Choose what you're costing before you touch a number.
Cost to serve means nothing in the abstract; it needs an object. Pick one: a customer, an order, a channel, a service tier. For most businesses the customer is the revealing choice, because that's the level where the invisible costs gather — the calls, the exceptions, the rework, the invoice that has to be chased twice. Take ten to twenty accounts rather than the whole ledger, and choose them to cover the range: your largest, your smallest, the one everybody complains about, the one you never hear about. A rough answer on twenty accounts beats a perfect system you never finish building.
02
Pool the shared costs, then name what actually drives each pool.
This is the whole method. Group the costs you currently spread evenly — support, delivery, account management, finance, returns handling, technical work — into pools, and ask of each what causes it to grow. Support scales with tickets, not revenue. Delivery scales with drops and distance. Finance time scales with invoices raised and payments chased. Account management scales with meetings. Then charge every pool out on its own driver. The moment you stop using revenue as the universal allocator the answer changes shape, because revenue never caused a single one of those costs in the first place.
03
The curve behind the average.
Sort your costed accounts by profit, best to worst, then accumulate the total as you work down the list. The line almost always climbs steeply through a small group at the top, flattens across a long middle, and then falls away as the loss-makers in the tail eat back part of what the leaders produced. The average customer, the figure you've been quoting in meetings, describes almost no account in the ledger. What you're hunting for is the point where the cumulative line turns down, and the names sitting past it.
04
Three moves work the tail, and the order matters more than the moves.
An unprofitable account is a design problem before it's a relationship problem, so take the options in sequence. Reprice: raise the price, or charge separately for what they consume — the bespoke reporting, the small urgent orders, the extra delivery run. Redesign the service: move them to a standard tier, a minimum order, a self-serve channel, a scheduled call instead of an open line. Only when both have genuinely failed does releasing them make sense, and it's far less dramatic than it sounds. Skipping the first two steps is how you fire a customer who'd happily have paid.

Members only · 4 principles + template + AI mentor

364 more frameworks are waiting.

Cost to Serve is Day 1 of 365. One framework every morning for a year — across strategy, sales, negotiation, leadership, and more.

$1 /day

Billed $365/yr · cancel any time

365 frameworks

12 topics · templates · AI mentor

Reserve your place — $365/yr →

One payment a year · the whole library · no auto-renewal.

More Finance frameworks

All 26 finance frameworks →