Unit economics depends on what one unit represents and which costs and value belong to it. Define the service unit, cohort, time horizon and allocation rules before dividing totals. An order, active account and completed engagement can each support a useful but different measure.
Choose the unit that matches the operating decision
Start with the decision rather than the familiar label. Staffing a delivery process may require cost per completed engagement, while evaluating account support may require cost per active account-month. Define what qualifies a unit and how cancellations, partial completion and repeated use are handled. A unit must be countable consistently across the numerator's period and population.
Check that the denominator does not reward the wrong behaviour. Cost per closed ticket can fall if a team splits work into more tickets or closes unresolved cases. Cost per account can change when inactive accounts are included. Preserve the underlying counts and status rules, and pair the measure with an outcome or quality safeguard. A lower ratio is not automatically a better service.
State the value and cost boundary
Identify which amounts are observed and which are allocated. Direct delivery labour, payment fees, refunds, acquisition and shared overhead can belong to different questions. State inclusions and exclusions instead of calling every residual profit. Align currencies and periods, and avoid mixing current costs with lifetime value estimates without a clearly justified model and horizon.
Allocate shared costs using a rule tied to the purpose, such as workload or an agreed activity basis. A convenient equal split may be useful as a planning convention but should remain visible. Keep unallocated amounts separately where no defensible assignment exists. Sensitivity analysis can show whether reasonable allocations change the conclusion, without pretending that one precise allocation reveals an objectively unique economic truth.
Use per-unit results with scale and capacity context
Distinguish a historical average from the additional cost of serving one more unit. Fixed commitments, capacity steps and specialist bottlenecks can make them differ. A service operating below capacity may absorb extra work differently from one requiring another shift. Do not multiply a current average by a large growth target and present the result as a guaranteed operating forecast.
Compare like cohorts with similar maturity and service mix. Include unresolved work where it belongs in the cost boundary, rather than assigning all cost to completed easy units. Record assumptions and review triggers. The result should support a specific decision, such as investigating rework or testing a delivery model, while leaving broader accounting and investment judgements to the relevant controlled process and qualified owners.
A fictional engagement-cost calculation
A fictional scenario to explain the method, not a customer result or a market benchmark.
A fictional advisory service records 30,000 currency units of revenue and 18,000 of defined direct delivery cost for 60 completed engagements in a quarter. Revenue per completed engagement is 500, direct cost per engagement is 300 and the defined contribution is 200 per engagement. A further 6,000 of shared operating cost is excluded from that contribution boundary and shown separately.
The service also has 100 active accounts, some with no completed engagement. Dividing the same direct cost by accounts gives 180 per active account, a different measure rather than a correction to the 300 engagement cost. If shared cost is allocated equally across completed engagements, it adds 100 per engagement, leaving 100 after those defined costs. The analyst labels the allocation and checks whether unfinished engagements consumed labour in the numerator. These fictional ratios do not establish net profit or predict the marginal cost of serving an additional customer. If unfinished work is material, the reviewer either expands the analysis to an appropriate engagement cohort or explicitly labels the completed-work boundary and its limitations. The choice is documented before comparing teams, because different completion patterns could otherwise make the same cost allocation appear more or less efficient.
TAKE THIS INTO YOUR NEXT CONVERSATION
A practical next-step checklist.
- Define the service unit and eligibility rule.
- Align cost, value, population and horizon.
- Disclose shared-cost allocations and exclusions.
- Distinguish average from marginal cost.
Common questions.
Can cost per account and cost per engagement both be valid?
Yes. They answer different questions when their populations and cost boundaries are explicit.
Is contribution per unit the same as net profit?
No. It excludes whatever costs lie outside its defined boundary, which should be stated clearly.
Further reading & source context.
These primary references provide background. They are not evidence for the fictional example above and do not endorse Accuralytix.
Found an error or a missing qualification? Send an editorial correction.
Check my data readiness.
Use a practical tool to turn this idea into a clearer starting point.
Check my data readiness