Who this is for
Purchasing or e-commerce operator whose import broke or looked odd after a supplier changed its file.
A supplier file arrived with different columns, and nobody can say what changed or what the import must do about it.
The result
For one changed supplier file, a note classifies each change (rename, split, merge, type change, headers-only, truncation, encoding, date format) and suggests the matching change to your import mapping, as text for your developer. A headers-only or truncated file is flagged as not importable. Nothing is applied.
What is included
- Two versions of one supplier file, old and new, as redacted or synthetic copies
- Your current import mapping, as text: column names to fields
- A note that classifies each change and suggests the mapping change
- A smoke test on 12 drifts we plant, and a test on 6 changes you make and keep from us
- The script, run steps and stored decisions
You receive
- The change note
- The suggested mapping changes, as text
- The classification table for the 12 planted drifts and your 6 changes
- The script, run steps and stored decisions
What is not included
- Applying any change to your import or store
- Deciding whether to import the new file
- Judging whether the supplier's values are true
- A standing watch on every new file
- Files with personal data
What we need from you first
- The supplier's column names in each version, as text
- How the import maps columns today, in a sentence
- No real records, personal data or credentials in the first enquiry
Never send passwords, keys, customer records or confidential code in the first enquiry. Secure handover is agreed after scoping.
How we check it is done
- On 12 drifts that we plant (rename, split, merge, type change, headers-only, truncation, encoding change, date-format change, extra column, missing column, reordered columns, placeholder nulls), at least 10 are classified correctly with a right mapping suggestion. We plant these and know the answers, so this is a smoke test: 10 of 12 shows only that the true rate is at least 56% with 95% confidence. Then on 6 changes that you make to a copy of your own old file and do not tell us about (for example rename a column, split one, truncate the file, change a date format, add a column, swap two columns), at least 5 are classified correctly with a mapping suggestion your developer confirms. 5 of 6 is a small test: it shows only that the true rate is at least 41% with 95% confidence.
- A planted headers-only file and a planted truncated file are each flagged as not importable, never as 'no changes'.
- Comparing a file with itself reports no change.
- Your import mapping file is byte-identical before and after the job.
- Re-running the script reuses the stored model decisions and gives an identical note; the stored decisions are handed over.
You inspect the planted-drift table and the note, and sign off in writing. Payment follows sign-off. Your developer applies any change.
When we would stop or decline
- There is no written import mapping and none can be written
- The files hold personal data that cannot be removed
- You want changes applied to a live import
- You want every future file watched, which is a different service
Questions
How is this different from the file change report?
That job lists which products changed. This one says which kind of structural change happened to the file and what your import mapping should do about it.
Will it change my import?
No. It writes suggestions as text.
Can it watch every new file?
Not in this job. A standing service does that.
Price and terms
£195 · untested offer price. £195 after the agreed checks pass and you sign off. No payment before sign-off.
This is a new service with no published client results. The price is a starting point we have not yet tested with buyers. Nothing is ordered or charged by the enquiry. The full specification is on the Synthetic Industry catalogue.