Guide · Matching
Customer part number matching: from your customer’s number to your own catalog article
A buyer does not look up your article number. The request is printed from the buyer's own system and carries the buyer's own material numbers — exact, always the same and meaningless to anyone outside that company. Customer part number matching means turning those numbers into your articles: once through a person who knows, and from then on through a cross-reference that has to be right. This guide shows where such mappings come from, which errors they produce, when a mapping may count as safe and how Quotrail handles them.
Updated October 5, 2026
Why the customer's number is the best and the most difficult evidence
Inside the customer's company the number is unambiguous: it names exactly one item in their system and appears unchanged in every request. That makes it better evidence than any description. Outside that company it says nothing. You cannot search your catalog for it, and the same string can mean a different article at another customer.
Two things follow. A mapping belongs to one customer, not to the catalog — so the customer has to be recognised before it can apply. Quotrail recognises the customer by the sender's domain and looks up learned mappings only for that customer. And a mapping is a decision somebody made at some point: it needs an origin, a date and a way to correct it.
Where the mappings come from
Scroll horizontally to see all columns.
| Source | What it is good for | Its weakness | In Quotrail |
|---|---|---|---|
| A cross-reference list you already keep — in the ERP or in a spreadsheet | Many mappings at once, for your regular customers | Nobody knows how old the individual entries are | Imported as CSV or Excel with your article number, the customer's part number and the customer number. A trial run comes first; if one row is wrong, nothing is written |
| Requests that give both numbers | The customer supplies the mapping | Only works when the request has a column for each number | When a spreadsheet has a column for the customer's part number, confirming the line stores that number for the article |
| Decisions in review | Confirmed by the person who knows the customer | Grows only request by request | A resolved line is stored in the customer profile — with who confirmed it and when |
| Earlier orders and quotes | Show what the customer actually received | Repeat old substitutions and discontinued articles | An article the customer has bought before counts as evidence of its own — weaker than a mapping confirmed several times |
Whatever the source, the number is compared without regard to upper and lower case, spaces, hyphens and dots. “AB-4711/32” and “ab 4711 32” are the same number.
The errors cross-references produce
Scroll horizontally to see all columns.
| Case | The trap | What Quotrail does | Reviewer checks |
|---|---|---|---|
| The list maps one customer part number to two of your articles | The first entry wins and nobody notices | The import rejects the row: the number is already matched to another article, and the existing mapping has to be reviewed first | Which of the two articles is right |
| In review, a different article is chosen for a known number | The old mapping keeps coming back | Moves the mapping to the chosen article, counts confirmations from one again and records the correction in the log | Nothing further |
| Zero-filled numbers from an ERP, such as 000000000000221599 | No match because of the zeros | Compares the number without the zero fill as well | Numbers with one or two leading zeros: write them in the list the way the customer writes them |
| The mapped article has been discontinued | Quoting with full confidence what you cannot deliver | Never marks a discontinued article clear, however often the mapping was confirmed | The successor with the customer |
| Two customers use the same number for different articles | One customer's mapping is applied to the other | Stores mappings per customer and uses them only for the recognised customer | That the request was assigned to the right customer |
| A mapping was learned wrongly | It returns with every request and grows stronger with every confirmation | Lets you forget it in the customer profile: the mapping is deleted and the log records who did it | The customer profile, from time to time |
| The customer's part number looks like one of your own article numbers | Two articles each have strong evidence | Marks a line clear only when the best candidate is clearly ahead of the second best; otherwise it stays in review | Which article is meant |
Worked example: one customer part number, four requests
Fictional example. The customer, the numbers and the article are invented to show how the documented rules treat a customer part number over time; they are not a measurement of match rates.
A maintenance buyer requests a shaft seal under the internal number AB-4711/32; their spreadsheet has a column of its own for it. Your catalog does not know this number. This is what happens over four requests:
Scroll horizontally to see all columns.
| Request | Evidence found | Status | What happens in review |
|---|---|---|---|
| First: AB-4711/32 with a short description | No number evidence — only description and attributes | Needs review | Pick the article from the ranked candidates; the mapping is stored |
| Second: the same number, written “ab 4711 32” | Customer mapping, confirmed once | Clear or Needs review, depending on the other evidence | Confirm; the mapping now counts as confirmed twice |
| Third: the same number again | Learned customer mapping, confirmed twice | Clear, provided no attribute contradicts and no other candidate is close | Glance at price and quantity |
| Fourth: the article has since been discontinued | The same mapping, but the article is marked as discontinued | Needs review | Agree the successor with the customer |
Each of these lines shows the evidence itself: the customer's spelling, how often the mapping was confirmed, by whom last and on which day.
When a mapping may count as safe
A customer part number is strong evidence, not proof. In Quotrail a learned mapping is among the kinds of evidence that can make a line clear — but only together with the other conditions:
- The mapping belongs to the recognised customer and has not been corrected more often than confirmed.
- No hard technical attribute of the line contradicts the article.
- The best candidate is clearly ahead of the second best.
- The article is not discontinued.
It matters how often a mapping was confirmed. From two confirmations on, it weighs almost as much as an exact match on your own article number; after a single one, clearly less. A mapping that was corrected more often than confirmed counts against the article instead of for it. And whatever the line looks like: a quote only becomes a quote when a second authorised person approves it.
Starting with the mappings you already have
- Export the cross-references you keep today. Three columns are enough: customer number, the customer's part number, your article number.
- Maintain the sender domains of your customers, so that requests are assigned to the right customer and the mappings can apply.
- Treat an imported mapping as a starting point: it counts like a mapping confirmed once; further confirmations in review make it stronger.
- Go through the customer profile from time to time and have mappings forgotten that no longer hold.
Quotrail is built for technical and industrial distributors whose inside sales teams quote from their own catalog. See the plans and prices, book a call to try this with your own customer part numbers, or read the test cases for units, packs and part number errors.