Cost of order mismatches and payback of the fix
Estimates the yearly cost of mismatches in order data (people’s time, returns, discounts) and how fast fixing the process pays back.
At a glance
- What question does it answer?
- What data mismatches in orders cost a year and how fast fixing them pays back.
- What data is needed?
- Orders a month and the share of orders with a mismatch.
- How to read the result?
- Yearly loss shows the scale of the problem and payback whether it is worth solving now. Check the share of mismatches on a sample of one to two months of orders. Example on the default values: 1,200 orders a month and 6% discrepancies make 72 cases, losses of RUB 183,600 a month or RUB 2,203,200 a year. With a 70% fall in discrepancies and RUB 60,000 of support the gain is RUB 68,520 a month; the RUB 1.5m investment pays back in 21.9 months and the return on investment over 12 months is −45.2%: on that horizon the fix does not pay off. If the payback is longer than the horizon, first narrow the scope of the fix or check the estimated fall in discrepancies. Next: “Stock record accuracy and the cost of mismatches” and “End-to-end cash gap”.
Why and how to reason
A discrepancy in an order (wrong size, price, quantity) looks trivial but repeats thousands of times. So we calculate the price of error as a product: how many orders x what share has a discrepancy x cost of one. The cost of one has two parts: people’s time to sort it out (hours x rate) and a direct loss (return, discount, rush delivery).
The payback of a fix is counted as for any project: what we spend once and what we receive each month. Monthly benefit = losses x expected reduction - support. A fix will not eliminate errors completely, so we take a realistic reduction.
Example. 1,200 orders x 6% = 72 discrepancies. One costs 1.5 hours x 900 + 1,200 = RUB 2,550. Losses 72 x 2,550 = RUB 183,600 a month, RUB 2.2 m a year. The fix reduces by 70%: 128,520 - 60,000 support = RUB 68,520 a month. A cost of RUB 1.5 m pays back in 21.9 months.
When to use it
It is useful to justify investment in automation, integration or training: “is it worth it”. Suitable for wholesale, online orders, receiving. It does not suit cases where discrepancies are few and the shares unmeasured: first collect a month of data. Assumption: the discrepancy rate and the cost of one are constant. If part of the loss is reputational, add it as a separate estimate. Especially useful if the decision must be defended before management.
How to read the result
Look at the payback period and the return on investment over the chosen horizon. In the example payback is 21.9 months, so within 12 months the project is still negative (-45%). That is not always a “no”: if the effect grows or the horizon is really longer, the decision changes. Check whether a cheaper path exists: partial automation, training. Decision: if payback is longer than the horizon, cut costs or re-check the loss-reduction estimate; if shorter, include the project in the plan. Build sensitivity: what if the reduction in discrepancies is only 40%, not 70%, and how much the project can bear. If payback under a pessimistic reduction is still acceptable, the project is solid.
Common mistakes
The usual mistakes: overstating the reduction of discrepancies to 100%; not including support after launch; using the hourly cost from salary without taxes; counting only hours and forgetting direct losses. Also: not allowing for discrepancies shifting to another stage. Also: not measuring the discrepancy rate before the project, so there is nothing to prove the effect.
What data is needed?
- Orders a month and the share of orders with a mismatch.
- Hours to resolve, cost of an hour, direct loss per case.
- Cost of the fix, support and the expected reduction.
- Where to take it: orders and the share with discrepancies from order records (ERP, WMS, claims log); hours per case from timing or a team survey; the hourly cost from payroll with contributions; direct losses per case (penalty, delivery, return) from claims. Fix and support costs from the contractor’s quote or an IT estimate.
Inputs
| Field | Unit | Value in the example | Meaning |
|---|---|---|---|
| Orders per month | orders | 1,200 | How many orders are processed per month. |
| Share of orders with a mismatch | % | 6 | Orders where purchasing, warehouse and finance figures did not match. |
| Hours to resolve one mismatch | hours | 1.5 | Average staff time per mismatch: search, calls, correction. |
| Cost of an employee hour | ₽ | 900 | Full cost of an employee hour including taxes and contributions. |
| Direct loss per mismatch | ₽ | 1,200 | A customer discount, a return, unshipped goods, a penalty: the average per case. |
| One-off cost of the fix | ₽ | 1,500,000 | One-off costs of fixing the process: setup, integration, training. |
| Monthly support | ₽ | 60,000 | Monthly support of the new process: licenses, maintenance. |
| Expected reduction in mismatches | % | 70 | The share of mismatches that disappears after the fix. Estimate cautiously: 100% never happens. |
| Evaluation horizon | months | 12 | Over how many months the return on the investment is assessed. |
How it is calculated
| Step | Formula |
|---|---|
| Loss | Orders × share × (hours × rate + direct loss) |
| Benefit | Loss × reduction − support |
| Payback | One-off cost ÷ monthly benefit |
Worked numbers
Default values from the app; in the interactive version you enter your own.
- Mismatches per month
- 72
- Monthly loss
- 183,600 ₽People’s time + direct losses.
- Yearly loss
- 2.2 m ₽
- Monthly benefit after the fix
- 68,520 ₽Reduced losses − support.
- Payback
- 21.9 months
- Return on investment over 12 months
- −45.2 %
Mismatches cost 2.2 m ₽ a year. The fix pays back in 21.9 months.
Nuances and limits
- The expected reduction is your estimate; make it conservative.
- Hidden losses (a lost customer, a platform fine) are better left out of direct losses without documents: the estimate would become a guess.
- Support costs after launch are forgotten: without them the gain is overstated.