This is the desk’s proposed method:
Total cost per accepted file = all attributable production, review, correction, exception, reinspection, technology and transition costs for a defined cohort ÷ accepted files from that same cohort.
“Accepted” should mean passing the agreed production standard. It should not mean received, uploaded, invoiced, deemed accepted under commercial terms, or settled.
What belongs in the numerator?
| Cost bucket | Include | Avoid |
|---|---|---|
| External production | Base charges, complexity charges, paid revisions, minimum commitments and surge charges | Treating the headline estimate fee as the complete cost |
| Internal production | Loaded labour for drafting, assembly, intake queries and supervision | Counting the same hour again as review |
| Review and correction | All first reviews, repeated reviews, corrections and acceptance administration | Recording only the first review cycle |
| Exceptions | Specialist escalation and additional handling attributable to the production route | Adding unrelated coverage work to only one option |
| Reinspection | Repeat visits attributable to missing or defective production inputs | Charging all repeat visits to the producer regardless of cause |
| Technology and governance | Incremental licences, integrations, support, rule maintenance and vendor oversight | Ignoring retained platform costs or counting sunk costs as immediately avoidable |
| Transition | Implementation, training, parallel running and exit preparation | Hiding launch costs or allocating them over unsupported future volumes |
What does a like-for-like example look like?
Synthetic example—not observed pricing. Assume four alternative routes process the same type of 1,000 routine residential files, all eventually accepted under the same standard. Amounts are invented solely to demonstrate the calculation; they are not recommendations for staffing or supplier rates.
| Cohort cost | Improved in-house | Estimate-only outsourcing | Managed production | Internal plus targeted automation |
|---|---|---|---|---|
| External production or incremental automation fees | $0 | $90,000 | $120,000 | $25,000 |
| Internal production and assembly | $70,000 | $15,000 | $3,000 | $50,000 |
| Technology, supervision and governance | $20,000 | $5,000 | $5,000 | $12,000 |
| Review and correction | $25,000 | $35,000 | $15,000 | $20,000 |
| Exceptions and attributable reinspection | $5,000 | $8,000 | $4,000 | $5,000 |
| Allocated transition costs | $10,000 | $5,000 | $8,000 | $18,000 |
| Total cohort cost | $130,000 | $158,000 | $155,000 | $130,000 |
| Cost per accepted file | $130 | $158 | $155 | $130 |
The arithmetic illustrates three points:
- The $90 outsourced draft becomes a $158 accepted file.
- The higher-fee managed route costs less than estimate-only outsourcing in this scenario—but more than the internal alternatives.
- Automation can tie with improved internal production after implementation and retained labour are counted.
Break-even review test: if one proposal costs $25 more per file and all other costs are equal, at an assumed $75 loaded hourly review cost it must remove 20 minutes of review and correction to break even: $25 ÷ $75 × 60.
How do you prevent denominator games?
Use an assignment cohort and retain every file’s disposition: accepted, returned, declined, escalated, withdrawn and still open. Report acceptance yield and backlog age alongside cost per accepted file.
An immature cohort should be marked provisional. Include rescue costs when another team completes a failed assignment; do not quietly remove the difficult files from the supplier’s denominator.
For internal teams, show both fully allocated operating cost and avoidable cash cost. Released salaried hours are capacity, not automatically payroll savings. Identify separately whether the proposal reduces overtime, contractors, planned hiring or an actual budget line.