Order Errors in FMCG Distribution: Causes and Fixes
Most order errors are not typos. They are matching failures, stale rules and ambiguity resolved silently — each with a different fix and a different cost.
When an order goes wrong, the post-mortem usually stops at "someone made a mistake". That is true and useless. The interesting question is which kind of mistake, because there are four, and they have nothing in common except the credit note at the end.
One: the line was matched to the wrong product
The customer wrote "orange squash 1L". You stock it in a six-pack and a twelve-pack. The clerk picked one. It was the wrong one.
This is the most common error and the least likely to be recorded as an error. It gets found in the warehouse, or by the customer, and gets fixed as a delivery problem rather than an order problem. The order desk never hears about it, so nothing changes.
The fix is not more care. It is making the account's own order history part of the decision — this customer has taken the twelve-pack eleven times and the six-pack never — and flagging the line when that history is thin.
Two: the rule was not applied
The order went in under the account's minimum order value. Or on the wrong price list. Or past the cutoff, and got promised for Thursday anyway.
Nobody decided to skip the rule. It is 7am, there are ninety emails, and checking the minimum on every order is the thing that gets dropped when the queue is long. The rule exists. It is just not enforced at the moment of entry, which is the only moment it matters.
This is the cheapest class of error to fix, because the rule is already written down. It just needs to run every time rather than when someone remembers.
Three: the ambiguity was resolved silently
The order said "the usual plus 10 crackers". The clerk knew what the usual was. Probably.
This one is dangerous precisely because it usually works. Experienced people resolve ambiguity correctly most of the time, so the practice never gets questioned — until the person who knew is on leave, or the usual changed in March and nobody updated the note.
The fix is to make the assumption visible instead of banning it. An order that says "interpreted as: last order of 4 June, plus 10 crackers" can be corrected in two seconds. An order that silently assumed the same thing cannot be corrected at all, because nobody knows an assumption was made.
Four: the actual typo
120 instead of 12. This is the error everyone pictures, and it is the rarest of the four.
It is also the easiest to catch, because it is the only one that is statistically obvious. An account that has never ordered more than fifteen cartons of anything just ordered a hundred and twenty. That does not need intelligence to flag. It needs someone to have bothered to compare the line against the account's own history.
Why the costs are so uneven
The four classes cost wildly different amounts, which is why a single error-rate percentage tells you nothing.
A typo on a dry good is a re-pick. A matching error on a chilled line is a return that cannot be resold, a van slot, a credit note and an unhappy kitchen. A rule that was not applied does not cost anything visible at all — it just quietly gives away margin on every order it touches, forever, which over a year is likely the most expensive item on this list.
What to measure instead
Counting errors is not very useful. Classifying them is. Pull a quarter of credit notes and amendments, put each one in one of the four buckets above, and the shape of your problem will be obvious within an hour.
Most distributors who do this are surprised. They have been running training on carefulness to solve a problem that turns out to be product matching and unenforced minimums — neither of which is fixed by being careful.
When an order goes wrong, the post-mortem usually stops at "someone made a mistake". That is true and useless. The interesting question is which kind of mistake, because there are four, and they have nothing in common except the credit note at the end.
One: the line was matched to the wrong product
The customer wrote "orange squash 1L". You stock it in a six-pack and a twelve-pack. The clerk picked one. It was the wrong one.
This is the most common error and the least likely to be recorded as an error. It gets found in the warehouse, or by the customer, and gets fixed as a delivery problem rather than an order problem. The order desk never hears about it, so nothing changes.
The fix is not more care. It is making the account's own order history part of the decision — this customer has taken the twelve-pack eleven times and the six-pack never — and flagging the line when that history is thin.
Two: the rule was not applied
The order went in under the account's minimum order value. Or on the wrong price list. Or past the cutoff, and got promised for Thursday anyway.
Nobody decided to skip the rule. It is 7am, there are ninety emails, and checking the minimum on every order is the thing that gets dropped when the queue is long. The rule exists. It is just not enforced at the moment of entry, which is the only moment it matters.
This is the cheapest class of error to fix, because the rule is already written down. It just needs to run every time rather than when someone remembers.
Three: the ambiguity was resolved silently
The order said "the usual plus 10 crackers". The clerk knew what the usual was. Probably.
This one is dangerous precisely because it usually works. Experienced people resolve ambiguity correctly most of the time, so the practice never gets questioned — until the person who knew is on leave, or the usual changed in March and nobody updated the note.
The fix is to make the assumption visible instead of banning it. An order that says "interpreted as: last order of 4 June, plus 10 crackers" can be corrected in two seconds. An order that silently assumed the same thing cannot be corrected at all, because nobody knows an assumption was made.
Four: the actual typo
120 instead of 12. This is the error everyone pictures, and it is the rarest of the four.
It is also the easiest to catch, because it is the only one that is statistically obvious. An account that has never ordered more than fifteen cartons of anything just ordered a hundred and twenty. That does not need intelligence to flag. It needs someone to have bothered to compare the line against the account's own history.
Why the costs are so uneven
The four classes cost wildly different amounts, which is why a single error-rate percentage tells you nothing.
A typo on a dry good is a re-pick. A matching error on a chilled line is a return that cannot be resold, a van slot, a credit note and an unhappy kitchen. A rule that was not applied does not cost anything visible at all — it just quietly gives away margin on every order it touches, forever, which over a year is likely the most expensive item on this list.
What to measure instead
Counting errors is not very useful. Classifying them is. Pull a quarter of credit notes and amendments, put each one in one of the four buckets above, and the shape of your problem will be obvious within an hour.
Most distributors who do this are surprised. They have been running training on carefulness to solve a problem that turns out to be product matching and unenforced minimums — neither of which is fixed by being careful.