Supplier integration guideData operations
Start here
Before you build
Specifications
After you submit
Data operations
What happens to your data after it arrives.
These operations exist so you do not have to normalise, translate or restructure anything before submitting.
Send your data in its native shape and let the pipeline do the conversion. That is not politeness — a hand-cleaned export is genuinely harder to map than a messy one, because cleaning it throws away the original values the mapper needs to see.
Which operations are available on your account is confirmed with your integration, and several of them depend on the level your data reaches.
- Document
- PREP-OPS-021
- Revised
- 8 September 2026
- Runs
- Every cycle
The ingest pipeline
| Stage | Step | What happens |
|---|---|---|
| 01 | Ingest | The feed is retrieved or received, its schema is detected, and structural validation is applied. A parse failure stops here with E-900 and nothing is applied in part. |
| 02 | Map | Your attribute names and values are mapped to the platform taxonomy. Manual review is available where the automatic mapping is ambiguous. |
| 03 | Enrich | Localisation is applied per supported language, and regional pricing logic and currency rules per market. |
| 04 | Publish | Distributor webshops and sourcing tools receive the updated catalogue. Stock continues to sync on its own cadence. |
Stages 02 and 03 are idempotent
A resubmitted feed produces the same mapped output, so a correction is applied by resubmitting rather than by patching anything. That is why the submission process asks you to resubmit a sample rather than to send a diff.
What is done to your data
| Ref | Operation | What it does | Level |
|---|---|---|---|
| OP-01 | Translation | Product names, descriptions and attribute values are localised into the 29 languages the platform's webshops run in. Incremental — a string already translated is not reprocessed. | L1 |
| OP-02 | Regional pricing | Multiple price lists per country and per customer group, with the currency conversion rules maintained on our side. | L1 |
| OP-03 | Taxonomy normalisation | Category, material and attribute values are aligned to the platform taxonomy. Certification tagging — REACH, FSC — is applied where your source data supports it. | L1 |
| OP-04 | Print-data digitisation | Print price lists are converted into queryable pricing logic, including quantity breaks, setup charges and colour surcharges. | L2 |
| OP-05 | Asset preparation | Image clean-up, overlay templating and coordinate calibration for logo sampling. | L3 |
| OP-06 | Artwork pipeline | Vectorisation, colour management and production-file generation for Web2Print output. | L4 |
Do not pre-clean your data
Normalising attribute values before submission usually costs you effort and loses information. The pipeline maps inconsistent source values better than a hand-cleaned export does, because it can see the original alongside the target. Submit the raw shape.
Which operations are available on your account is confirmed with your integration, and several of them depend on the level your data reaches.
Result ordering is not a data operation
How prominently your range is surfaced in webshops and sourcing tools is a distribution setting. Nothing in your feed controls it, and no field in any specification in this set affects it.
| Question | Where the answer is |
|---|---|
| Can a field in my feed raise my products in results? | No. There is no such field, at any level. |
| Does reaching a higher integration level change it? | No. A level decides which capabilities your data supports and nothing else. |
| Where is it decided, then? | Promo Campaigns, which your membership makes you eligible to buy. It is a commercial arrangement rather than a technical one, and it is described on its own page. |
What no operation can do for you
The four things the pipeline cannot fix, so nothing above reads as a promise that data will be repaired however it arrives.
| Not done | Why, and what it means for you |
|---|---|
| Assigning identifiers | Codes have to come from your systems and stay stable. We can issue generated codes as a fallback, but they break the reconciliation on your side. |
| Filling in missing required fields | No operation can infer a value that was never submitted. A row missing a required field is rejected, not completed. |
| Making a nightly export fresher | Cadence is a property of your export, not of the ingest. The pipeline cannot read a file more often than you produce it. |
| Interpreting pricing intent | A rule that is not expressible as columns is discarded rather than guessed at — §3.0 of the Level 2 specification. |
Send what you already keep
The shape it is in now is the shape to send. Run the checklist for the levels you are implementing, send twenty variants, and the report will tell you what — if anything — is genuinely missing rather than merely untidy.
