Most route businesses know their gross margin by product and almost nothing about their profitability by customer. That gap matters, because in a route business the cost of serving a customer varies enormously and gross margin is a poor proxy for contribution.
Why customer cost varies so much
Two customers buying identical volumes at identical prices can differ in delivery cost by a factor of several, because of:
- Drop size. The cost of a stop is largely independent of how much is delivered.
- Visit frequency. Twice weekly costs roughly twice as much to serve as weekly.
- Location and density. A customer 20 minutes from the nearest other stop carries the whole detour.
- Service time. Cellar drops, restricted access, waiting and merchandising all consume minutes.
- Time window constraints. Narrow windows reduce route efficiency for everyone on the route.
- Returns and credits. Handling and administrative cost.
- Payment behaviour. Cash collection, late payment and credit control cost real money.
- Order channel. Phone orders taken by a person cost more than electronic ones.
In most route businesses the least profitable customers are not marginally unprofitable — they are substantially so, because the fixed cost of a visit swamps the margin on a small order.
Building the analysis
1. Establish the cost of a route hour. Driver cost including on-costs, vehicle fixed and variable cost, fuel, and an allocation of depot and administrative overhead.
2. Allocate time to customers. From telematics and delivery records:
`` Customer time = service time at the stop + travel time attributable to that stop ``
Attributing travel is the analytically awkward part. Two defensible approaches:
- Marginal: the additional travel caused by including this customer in the route — the theoretically correct measure, and computationally heavy.
- Proportional: total route travel allocated across stops, weighted by distance from the route centroid or by inter-stop distance. Simpler, adequate for identifying the tail.
Start proportional. The extreme cases — the customer requiring a 25-minute detour — are visible either way.
3. Add non-delivery costs. Order taking, credit control, returns handling, customer-specific promotions, merchandising visits.
4. Compare with gross margin per customer over the same period.
5. Plot it. Cumulative contribution against customers ranked by profitability. The resulting curve — often described as a whale curve — is usually startling: a portion of customers generate more than 100% of profit, a middle band is roughly neutral, and a tail actively destroys it.
What to do about the tail
Removing customers is the last option, not the first. In order:
1. Increase drop size. Minimum order values, incentives for consolidated ordering, or moving to less frequent but larger deliveries. This is the highest-value lever because it improves the economics without losing revenue.
2. Reduce frequency. Many small customers are visited more often than their volume justifies, usually for historical reasons.
3. Change the service model. Move small customers to telesales, a wholesaler, a collection point or a third-party carrier.
4. Reprice. Delivery charges below a threshold, or a service charge that reflects the cost. Price the service, not just the product.
5. Improve efficiency. Better routing, clustering by delivery day, and reducing service time through access improvements.
6. Exit. Only where the others fail and the relationship has no strategic value. Do it deliberately, with notice, and consider who else you might lose with them.
Cautions
Strategic customers. A small unprofitable outlet may belong to a chain whose larger sites are highly profitable. Analyse at the right level — the account, not just the delivery point.
Growth customers. A new customer is unprofitable by definition for a period. Segment new accounts separately.
Contribution, not full absorption. Removing a customer does not remove your fixed overhead; it reallocates it across fewer customers. Judge marginal decisions on contribution above avoidable cost, not on fully absorbed profit.
Route density effects. Removing customers from a route can worsen density for the remainder, raising their cost. Model the route, not the customer in isolation.
Making it operational
Cost-to-serve is most valuable as a routine report rather than a one-off study:
- Quarterly refresh, ranked by contribution
- New customers flagged separately with a defined ramp period
- Alerts when an existing customer moves into the loss-making band
- Cost-to-serve visible to sales at the point of negotiating terms
- Minimum order and frequency policies enforced in the order system
That last point matters most. An analysis that never reaches the person agreeing terms with the customer changes nothing.
Frequently asked questions
How precise does the cost allocation need to be?
Precise enough to rank customers reliably. Arguments about whether a customer costs a slightly different amount are a distraction; the analysis is designed to identify the customers whose cost is multiples of the average, and those are visible under any reasonable method.
What proportion of customers are typically unprofitable?
It varies by sector, but route businesses running this analysis for the first time commonly find a substantial tail of customers contributing negatively. The specific figure matters less than the shape of your own curve.
Should we tell customers they are unprofitable?
Not in those terms. Frame it commercially: a minimum order value, a delivery charge below a threshold, or a change in delivery frequency. Most small customers accept a minimum order or a reduced frequency rather than losing supply entirely.
How do we handle customers who are unprofitable but strategic?
Decide explicitly and record the decision. A customer retained for strategic reasons is a legitimate choice; one retained because nobody noticed is not. Review the strategic justification annually.
Does this analysis need special software?
It needs your route data, your sales data and a spreadsheet or BI tool. Some route accounting systems include profitability analysis natively, which makes it routine rather than a project — but the first analysis can be done with data you already have.
