Part numbers are not unique on their own.
SKF 6204-2RS and FAG 6204-2RS are different products. Deduplicating on the part number alone silently merges them and destroys stock and pricing. The key is brand plus part number.
Product Data Operations
PIM implementation · Replatforming · Supplier onboarding · Catalogue consolidation
Implementing a PIM? Your supplier data has to get there first.
Send us the supplier spreadsheets, PDFs and legacy exports. You get back an import-ready catalogue for your agreed schema, a validation report, and a queue of everything the source could not answer. Fixed price, agreed in writing, no meetings.
- No meetings required
- Written scope
- Validation reports included
Source files
- XLSXsupplier_catalog.xlsx
- PDFproduct_specs.pdf
- CSVtarget_schema.csv
- Profile
- Map
- Normalize
- Validate
Output
- catalog_ready.csv
- validation_report.csv
- exceptions.csv
- transformation_log.csv
source records
- ready
- need review
- exact duplicates excluded
A value is either derived from your source, or it is an exception. Nothing is invented to make a row pass validation.
Images are not attributed on a guess.
6000-2RS1.jpg is not the image for part 6000. A filename carrying no brand cannot be assigned to one of six manufacturers sharing that part number, so it is reported for review. In the demonstration run that is why 181 of 300 candidate images were auto-assigned and the rest were flagged.
A missing category stays missing.
168 rows arrived with an empty group column. They are held with the raw value and a suggested action, not filed under a plausible-looking guess.
Pricing and turnaround.
The unit of work is not the SKU. It is the decision about an attribute value.
A catalogue with 30 attributes costs several times what the same catalogue costs with 6. Any quote given without seeing the target schema is a guess. That is exactly why the audit exists.
Data Audit
$200
Send a representative sample of up to 500 SKU and your target schema. You get a written scope: which source fields map cleanly, which need agreed rules, what cannot be resolved from the source at all, and a fixed price for the full batch. Credited in full against the project.
Turnaround2 business days.
Single Supplier Batch
from $400
One supplier set of up to 1,000 SKU prepared to your agreed schema. Import-ready file, validation report, exception queue, transformation log. One round of revisions.
Price depends on the number of attributes in your target schema, not on the number of products.
Turnaround5 business days.
Multi-Source Migration
from $1,800
Several suppliers consolidated into one catalogue: cross-supplier deduplication, variant grouping, attribute extraction from spec-sheet PDFs, image attribution. Final price fixed in the audit.
TurnaroundQuoted in the audit.
Final price is fixed in the audit, before any production work begins. No hourly billing, no open-ended scope.
These are Variorum Data prices, in US dollars.
From supplier files to a consistent product dataset.
See what changes, what stays unchanged and what gets flagged instead of guessed.
Supplier source supplier_catalog.xlsx
Prepared output target schema
Select a tag in the prepared output to see the rule behind it.
Demonstration data. Transformation rules are agreed before production.
A full pipeline run: five supplier files, one catalogue.
Worked demonstration on synthetic data. Not client data.- rows read from 5 supplier files
- 2,928
- exact duplicate rows removed
- 26
- cross-supplier duplicate part numbers resolved
- 51
- records after deduplication
- 2,851
- import-ready (93.2%)
- 2,656
- held for review, none dropped
- 195
- exception line items, each with a reason and a suggested action
- 485
- attributes recovered from supplier PDFs
- 327
- images attributed one to one, zero double assignments
- 181
- CSVImperial units and USD.
- XLSXHeader buried on row 5 under a title block, prices stored as text.
- CSVGerman, with semicolons and comma decimals.
- XLSXNo category column, variants encoded only in the product name.
- CSVMissing references, duplicate rows and two attributes that exist only inside a PDF.
AI assists the work.
Clear rules control the output.
We use AI where it helps with extraction and mapping. Agreed rules check structure and consistency, and every step below is handled in writing.
Missing data stays visible.
Four files come back.
Synthetic sample, not client data.ZIP with the four CSV files, the 100 synthetic source rows and a README listing every demo rule.
A validation pass checks the agreed rules. It does not certify the accuracy of the source material or guarantee acceptance by every platform.
The data work behind a cleaner launch.
Platform implementation, API integration and publishing to production are not part of the default scope.
Your implementation team. Our data-production support.
Keep ownership of the platform, client relationship and business decisions. We prepare product data to the specification you approve, with delivery and review handled in writing.
White-label delivery follows the permissions and terms agreed for each project, including what your client contract requires you to disclose.
Request a Data Audit
We reply in writing to agree the sample and target schema for your audit. Production work begins only after the audit fixes the price.