Skip to main content

Market-neutral small-business guide

Cost per Job, Order or Customer: Pick the Denominator

Match the denominator to the cost pool and decision, reconcile total cost and avoid comparing unlike per-object figures.

Choose the object from the action the reader can take

A denominator is useful when it matches the thing being priced, accepted, improved or stopped. Use a job when scope and delivery resources are committed per project. Use an order when fulfilment and transaction events repeat per checkout. Use a customer when acquisition, support, returns or payment behaviour accumulate across that relationship.

Decision tree for a cost object
Decision questionPrimary objectRequired bridge before comparison
Should this scoped project be quoted or repeated?Job or projectTasks, labour, materials and job-specific overhead
Does this sales channel retain enough per transaction?OrderOrders per customer and channel-specific fulfilment/fees
Is this relationship costly to acquire and serve?Customer or cohortOrders, support events, returns and relationship period
How should one shared pool be recovered?The object that consumes the pool driverActivity count and reconciliation to the pool

Choose and validate a decision denominator

Prerequisites, sequence and checkpoints

  • Name the decision and the pool being explained. (not complete)
  • Choose job, order, customer or another object that matches the decision. (not complete)
  • Define the count, period and inclusion rules before division. (not complete)
  • Allocate shared cost explicitly and reconcile object rows to the pool. (not complete)
  • Test zero and low counts, then hand the basis to the object-specific planner. (not complete)

See why unlike denominators are not interchangeable

One 12,000 CU monthly pool restated three ways
ObjectCount and arithmeticUse
Job12 jobs; 12,000 ÷ 12 = 1,000 CU/jobQuote or job review
Order400 orders; 12,000 ÷ 400 = 30 CU/orderOrder contribution
Customer80 active customers; 12,000 ÷ 80 = 150 CU/customerCustomer economics
Keep one declared currency, period, unit and indirect-tax basis unless a row explicitly marks a boundary change.

Restate before comparing

Reproducible user scenario

The three results describe the same pool through different objects; their magnitudes cannot be ranked directly.

Illustrative inputs, arithmetic or reasoning record; not a benchmark or recommendation
StepInput or arithmeticDecision meaning
Pool reconciliation12 × 1,000 = 400 × 30 = 80 × 150 = 12,000 CUAll views reconcile
Non-comparability1,000/job versus 30/orderDifferent objects; no higher-cost conclusion
Zero-count casePool 12,000; object count 0Division is undefined; change horizon or decision basis

Restate the figures before choosing an action

The example’s 1,000 CU per job, 30 CU per order and 150 CU per customer all reconcile to the same 12,000 CU pool. None is larger or smaller in an economic sense because the objects differ. To connect them, record how many orders belong to a job or customer and how the relevant costs move through that relationship.

Check denominator drift

  • Stable numerator, falling count: investigate unused capacity or a temporary volume trough before repricing every object. (not complete)
  • Growing numerator, stable count: separate unit-rate, usage and scope changes. (not complete)
  • Growing count, falling average: confirm the total pool and service quality did not increase elsewhere. (not complete)
  • Zero or tiny count: extend the horizon, use a capacity denominator or keep the pool unallocated; do not divide by an arbitrary one. (not complete)

Limitations, evidence and next action

Use the calculation owner for the next step

Questions and boundaries

Which denominator is best?
The one that matches the named decision and can be defined consistently; there is no universal denominator.
Can per-order cost be compared with per-customer cost?
Only after both are restated through a common volume and period bridge.

Sources and scope

  • Cost drivers — OpenStax: Supports cost-pool and driver selection; allocation is not proof of causation.

Change history

  1. Initial public release of the article after pre-launch factual, editorial, source and presentation review.