What should you reserve against a recall?

A recall estimate built from single numbers lands on one figure and hides the two inputs that decide the tail. Six uncertain quantities, one run, and a cost distribution you can put a reserve against.

Operations Advanced Monte Carlo Pro engine

Install free How Monte Carlo simulation works

After you install, this is the model to open.

What Would a Recall Cost?

In your spreadsheet, click the Sortia icon in the strip of icons down the right-hand edge. No strip? Click the arrow at the bottom-right to open it. You can also use Extensions, then Sortia, then Open Sortia.
Click Start from a template and put that name in the search box.
Pick the card with that name and click Load this template. It arrives on a new tab with real numbers already in it.

This one runs on a Pro engine, and every free install includes five full-quality runs on your own numbers, shared across all five Pro engines rather than five for each. After that, Pro is $199/year.

The answer

Recall cost, mean: $4.60M the single-point sheet says $3.62M
P95: $7.23M the reserve conversation
Worst of 20,000 trials: $10.82M
Committed on day one: $900,000 the fixed programme cost of any recall

Six uncertain quantities and one certainty. The certainty is the $900,000 fixed program cost, which you pay the moment you decide to recall anything at all. Everything else is a range: how much of the build is really defective, how many of those units come back, whether a penalty lands and how big it is, and how much margin the line loses while it is stopped.

Click Run on the 20,000 trials the template loads with. The mean is $4,603,127 against the $3,623,700 the single-point sheet shows, so the estimate on the board paper is roughly a million dollars light before anything goes wrong. The median is $4,367,912, the P90 is $6,443,267, the P95 is $7,233,318 and the P99 is $8,567,127. The worst trial in twenty thousand reached $10,824,885.

Now read the tornado, and expect it to contradict the room. The two quantities a recall meeting spends its time on, how much of the build is defective and how many units come back, rank third and fourth. The two that decide the answer are the margin lost while the line is stopped and whether a penalty is imposed at all. The arithmetic behind that is on the sheet: at the typed numbers the entire return campaign, every unit repaired and freighted both ways, is $623,700 of a $3,623,700 total.

A recall is a downtime event and a regulator event that happens to involve some logistics, and it is usually discussed the other way round. Second run, and it is the one that ought to change a budget. Tighten the defective share to a PERT of 0.05, 0.06 and 0.07, which is what a complete batch trace buys you, and rerun on the same seed. The mean falls from $4,603,127 to $4,466,294 and the P95 falls from $7,233,318 to $7,050,671, which is 3% off the average and 2.5% off the tail.

Serialisation is worth having for half a dozen reasons, and this is not the number that pays for it. Third run, on the input the tornado actually named. Put the lost margin while the line is stopped on a PERT of $600,000, $1,200,000 and $2,400,000, which is a quarantine and a screened restart rather than a full line stop, and rerun: the mean falls to $3,486,319 and the P95 to $5,933,805.

A quarter of the expected cost of a recall, and nearly a fifth of its tail, sits in how fast you can get the line running again. The penalty behaves differently from everything else and it is worth seeing why. It is off entirely in 75 of every 100 trials, so it barely touches the middle and it dominates the top of the range: drop its chance from 25% to 10% and the median moves 5% while the P95 moves 13%.

That is the difference between a cost that scales and a cost that switches on, and it is the reason a reserve set from the average is the wrong reserve. What this cannot tell you is whether you have to recall at all. This sheet prices the campaign; the decision to run it is a different question, and the Decisions tools are where that one belongs.

To adapt it, put your own units in the field and your own per-unit repair and logistics costs in, set the return rate from a published campaign in your category rather than from optimism, and set the line-stop margin from what your plant manager says a screened restart really takes.

The model

It arrives on a tab called Template: What Would a Recall Cost:

Units of the affected build in the field	420000
Share of the build actually defective	0.06
Units in scope of the recall	25,200
Share of recalled units actually returned	0.55
Units returned	13,860
Replacement or repair cost per returned unit ($)	34
Return logistics cost per returned unit ($)	11
Fixed program cost: notification, admin, legal ($)	900000
A regulatory penalty is imposed (1 = yes)	0
Penalty if imposed ($)	1500000
Lost margin while the line is stopped ($)	2100000
Total cost of the recall ($)	3,623,700

Once it is in your sheet

The model arrives with real numbers in it and runs as it stands, so you can press the button first and understand it second.
Change the numbers to yours. The sheet marks which cells are inputs and which hold formulas, and most labels carry a note explaining the row.
Press the run button at the bottom of the panel. It is labeled for the tool you are in, and the result lands on its own tab, with a written reading of it beside the figures.

Never used Google Sheets? Start here goes the whole way, in seven steps, and assumes nothing.

Next question

Every model like this one, and the method behind them: Monte Carlo simulation.

Install free All 379 templates