Cost to serve: finding out which orders lose money

Cost follows behaviour, not revenue

The drivers are order frequency, order size, line count, drop density, delivery window, packaging and labelling requirements, returns, and the amount of manual coordination each account generates. A customer placing frequent small orders with tight windows and bespoke labelling can consume several times the cost of one buying the same annual volume in planned bulk deliveries. Averaging that cost across all revenue subsidises the demanding account with the margin of the efficient one.

Allocate by activity, and only where it changes the answer

The method is to identify the activities that consume resource, establish what drives each, and assign cost accordingly. Precision beyond the point of decision usefulness is wasted effort: the aim is to separate the clearly profitable from the clearly unprofitable, not to compute a figure to the last unit. Spreading overhead evenly is what the exercise exists to avoid, so any allocation that reverts to a percentage of revenue defeats the purpose.

Reading the distribution

Plotting cumulative profit against customers ranked by contribution typically produces a curve that rises, flattens and then falls, with the decline caused by accounts whose service cost exceeds their margin. The shape matters more than any individual figure, because it shows how much profit is being consumed by the tail and therefore how much is available from fixing it.

Turning the finding into action

Loss-making accounts are rarely resolved by exit. The usual levers are changing the ordering pattern through minimum order values or fixed delivery days, charging for services that were previously absorbed, adjusting price to reflect actual service consumption, or moving the account to a lower-cost fulfilment route. The analysis is only useful if the commercial team is part of it from the start, since they are the ones who will have the conversation.

Frequently asked questions

How detailed does the model need to be?

Detailed enough to distinguish materially different service behaviours and no more. A model that takes months to build and cannot be refreshed will be obsolete before it is used, whereas a simpler one maintained quarterly keeps influencing decisions.

Should unprofitable customers be dropped?

Usually not immediately. Many become profitable once the ordering pattern or the service level is changed, and some absorb fixed cost that would remain after they left. Exit is the last option, after the cheaper commercial adjustments have been tried.

How does cost to serve relate to landed cost?

They cover opposite ends of the flow. Landed cost accumulates everything spent getting goods into your possession at a usable location; cost to serve accumulates what is spent moving them from there to a specific customer under that customer's particular requirements.