Job old-supplier-codes-matched-to-new-products · revised 12 October 2026
My supplier changed all its product codes: match the old ones to the new ones
Each old product code maps to one new product, or to an unsure list, with the attributes that justify the link. Look-alike traps such as pack sizes are tested on a small sample you label.
You might be seeing
- The new file has none of the codes your catalogue uses
- Prices or stock would land on the wrong product if you imported it
No passwords, keys, card details or admin invites needed to start.
What usually happened
When a supplier changes every code there is no key to compare on, so a straight join finds nothing. Matching on names alone links a 12-pack to a 6-pack. A wrong link puts price or stock on the wrong product.
Who it’s for: Purchasing or e-commerce operator whose stock, prices and images are keyed to supplier codes that no longer exist.
Usually starts when: A supplier re-numbered its range and the new file no longer lines up with your catalogue.
The result: Each old code maps to one new product, or to an unsure list, with the attributes that justify the link. Exact barcode or part-number matches are counted apart from the rest. A held-back set of pairs you label tests the false-match rate on a small sample (0 wrong links in 15 look-alikes bounds it under 19%), and pack-size traps you plant must land on unsure.
Check whether this job fits
Answer these without sending files or logins. Nothing is submitted unless you choose to contact us.
What you get
- The old-to-new mapping file
- The unsure list
- The exact-match count
- The justifying attributes for each link
- A results table on the held-back pairs
- The script and the stored decisions
Included
- One old file and one new file of product data, up to 20,000 rows each
- Exact barcode or part-number matches, counted separately
- For the rest, matching on brand, size, pack, colour and description, with a model only for pairs left unclear
- A held-back test on 30 pairs you label, plus pack-size and variant traps that you plant
- A mapping file, an unsure list, the script, the stored decisions and the steps to run it
Not included
- Applying prices, stock or images in your store or database
- Inventing a barcode or a part number
- Deciding that a product is discontinued
- Products that cannot be told apart without inspection
- Personal data of any kind
How we know it’s done
Agreed with you before work starts. Each check produces evidence you keep.
Exact barcode or part-number matches are counted on their own line and need no model.
Evidence: The exact-match counts for each file.
Of the 15 look-alike pairs in the held-back 30, none is marked linked. Of the 15 true pairs, at least 12 are linked to the right new product; a pair marked unsure counts as a miss. 12 of 15 shows the true rate is at least 56% with 95% confidence, and 0 wrong links in 15 bounds the wrong-link rate under 19% with 95% confidence. This is a small test. A mapping that links nothing fails.
Evidence: The results table beside your labels.
Six traps (a 12-pack against a 6-pack, 500 ml against 1 litre, left against right hand, and similar), which you plant and we do not see beforehand, all land on unsure.
Evidence: The planted list and the unsure list.
No product is linked to two old codes unless listed, and every identifier in the output appears in the source files.
Evidence: The many-to-one list and a script check.
Re-running the script reuses the stored model decisions and gives an identical mapping file; the stored decisions are handed over.
Evidence: Both files compared, and the stored decisions.
Sign-off. You inspect the mapping, the unsure list and the held-back results, and sign off in writing. Payment follows sign-off. You apply the mapping in your own systems.
If it fails. If the held-back test is not passed, you do not pay for this fixed scope. You keep the results, the failures and the test set, and we agree whether to stop or re-quote. No surprise work.
When it fits, and when we stop
It fits when
- Both files are product data and carry at least two identifying attributes, such as brand plus size
- You label 60 pairs: 30 for tuning and 30 held back (15 true pairs and 15 look-alikes)
- A named person decides every unsure row
- The files are redacted or synthetic copies with no personal data
- Two people on your side label the same 20 pairs first and agree on at least 17
We stop and tell you if
- Products cannot be told apart without seeing them
- Neither file carries two identifying attributes
- You want the mapping applied to a live store by us
- Two people on your side label the same 20 pairs and agree on fewer than 17
What could go wrong
The job reads copies you send and writes files. It changes nothing in your systems, so there is nothing to revert. Our copies of your files are deleted when the job ends.
Scroll the table sideways to read it all.
| Risk | How we handle it |
|---|---|
| Price or stock lands on the wrong product. | The mapping is a file you check. Nothing is applied by us, and look-alikes go to the unsure list. |
| A model invents a code that looks real. | Every identifier in the output must appear in the source files. One invented identifier fails the whole test. |
| Your files go to a model provider outside your control. | Before any file is shared we name the provider, its country and its retention terms in writing. Entry jobs run on redacted, synthetic or public copies, or on your own provider account with a key you hold. |
A separate automated reviewer checks the matching rules and the planted traps before hand-over and does not mark the held-back test. A named person on your side labels the test set and signs off. No person on our side is included unless your agreement names one, and at launch the work is largely automated.
How we deliver
We arrange the work and independent review, then show you the result against the agreed checks. You keep authority over your systems.
- Agree the files, the attributes, the pass mark and the severe errors in writing
- Count exact barcode and part-number matches; set them aside
- Two of your people label 20 pairs to check they agree
- You label 60 pairs; 30 stay held back until the test run
- Build attribute matching and run your planted traps and the held-back pairs
- Independent review of the traps, the mapping and the unsure list
- Hand over the mapping file, script and stored decisions; delete our copies
This is a one-off job, not emergency cover or a subscription. We confirm eligibility, the price, a start window and a delivery date before you accept, and work starts only after inputs, a secure handoff route and permissions are in place. Model usage up to a cap agreed in writing is included, and an enquiry creates no charge or booking.
Need to keep it working?
Need the same check each time the supplier changes its file? Ask about a quoted re-check when the supplier changes the file again.
Ongoing work is separately scoped and quoted: no monitoring, response-time guarantee or automatic subscription is included in this job.
Explore an ongoing managed lane, or mention the responsibility you need in your enquiry.
What you can check
This is a new service. We have not delivered this job for a client yet.
Other ways to get this done
- Ask the supplier for a cross-reference table of old and new codes. Many suppliers have one.
- A spreadsheet lookup on barcode handles the products that carry one; this job covers the rest.
Questions
Will it update my store?
No. You get a mapping file. You or your developer apply it.
What if two products really are the same?
Two old codes can point to one new product only if the mapping lists them. Otherwise they go to the unsure list.
Does it make up barcodes?
Never. Every identifier in the output must appear in your files.
Send an enquiry
Send us
- The two files' column headings and row counts
- Which columns might hold brand, size, pack or colour
- No real records, personal data or credentials in the first enquiry
Later, once you agree
- Redacted copies of both files by an agreed secure route
- Your labels for 60 pairs
- A named person who decides the unsure rows
- A written cap on model usage
You keep both files and every decision. We work on copies you send by an agreed secure route, return a mapping file and a script, and delete our copies when the job ends. We never write to your store or database.
Email fallback: open your mail app
If website submission is unavailable, review and send the fallback email yourself. An email fallback is not a website receipt. Or write to hello@syntheticindustry.ai with “old-supplier-codes-matched-to-new-products” as the subject.