Skip to content

Catalogue data audit as a small service

Pilot routeE-commerce operationsReviewed 2026-11-04
Check an online shop’s product file for duplicates, gaps and format errors — and hand back a clear fix list.

Problem and buyer

Small online shops add products over years. Their catalogue files collect duplicates, missing fields, mixed units and broken identifiers that hurt search, feeds and operations.

Who pays: Owner or operator of a small online shop with roughly 100–2,000 products who edits the catalogue in a spreadsheet export.

The offer

Audit of an authorised catalogue file — up to 200 rows and 10 fields in the pilot — with a defect report, proposed fixes and questions for the client. IDs and prices are never changed.

  • Defect report with row, field, rule and proposed action
  • Corrected copy with only safe fixes applied
  • Change log and open questions for the client
  • Delivery checklist and acceptance criteria

How the model works

Productized, fixed-scope data work with an objective acceptance check.

Today’s alternatives: The owner fixes it by hand, a general chatbot plus spreadsheet formulas, the shop platform’s own import warnings, or a feed-management tool.

Fit for your market

Keeps
Objective, countable output — errors are easy to show.
Changes
Platform export formats differ by shop system and country.
Not included
No direct write-back to Shopify or other systems in v1.

First test

Two synthetic practice files with a hidden answer key, then 3–5 real shops that share a file with written permission.

Continue if: ≥ 27 of 30 seeded defects caught with no changed IDs or prices, and at least 2 shops ask for a paid audit.

Why it might fail

  • A general AI + spreadsheet workflow may be equally good — the test compares them directly
  • Shops may not see enough value to pay
  • Client files can contain personal data — permission and deletion rules are part of delivery

Claims in this dossier

  • Rule-based checks catch formatting defects reliably on synthetic files.Fact
  • Shop owners will pay for a 200-row audit.Hypothesis
  • 1.5–3 hours per audit including human review.Estimate

Why this is the pilot route

Catalogue data audit is the first route Proofoak supports end to end, for three reasons. The deliverable is objectively checkable — a defect is either found or missed, an ID is either unchanged or changed — so quality can be measured before anyone sells it. It needs skills many people already have: careful spreadsheet work and clear written English. And it can be tested cheaply: synthetic practice files cost nothing, and a free 20-row check is an easy first step for a real shop.

It is a route to test, not proven demand. The pilot exists to find out whether shops pay for it and whether the workflow beats a plain chatbot plus spreadsheet.

What the work actually involves

A small online shop exports its product catalogue — usually a CSV from its shop platform — and shares it with permission. You check it against a fixed set of rules and return a report. The core checks are:

CheckWhat you look forCorrect action
Exact duplicatesIdentical rowsPropose removal; don't delete
Near duplicatesSame product under a different ID or reworded titleFlag for the client to merge
Missing required fieldsEmpty title, brand or categoryFlag and ask — never invent
Missing optional factsEmpty colour or weightMark "unknown"
Inconsistent units"KG", "grams", "0.75kg" in the weight fieldNormalise when safe; flag otherwise
Malformed identifiersIDs that break the shop's formatFlag only — IDs are protected
Invalid barcodesGTIN length or checksum errorsFlag only
Price formatting"12,99", "€12.99"Flag only — prices are protected
Category problemsCategories outside the list, or that don't fit the productPropose the closest valid category

Two rules never bend: IDs and prices are never changed, and missing facts are never guessed. They protect the client and they are what make the audit trustworthy.

What the client receives

  • A defect report listing every issue by row, field and rule, with the proposed action.
  • A corrected copy with only safe fixes applied, such as normalised units and "unknown" for missing colours.
  • A change log showing what changed and why.
  • A list of open questions for anything that needs the client's knowledge.

The readiness check

Before you use the route commercially, Proofoak asks you to practise on a synthetic 200-row file with seeded defects and a hidden answer key. The check scores defects found, false alarms, protected fields unchanged and the reasoning behind each fix. Passing is valid for this route and rubric — it is not a professional certificate — and if you don't pass yet, you get feedback and another try.

A realistic first month

WeekWorkOutput
1Profile, practice file, readiness checkPassed sample
2Offer and labelled sampleOne-page offer
3Ten sourced shops, approved messagesTest running
4Replies, calls, first free checksEvidence and a decision

Pricing the service

The price you charge a shop is set from your buyer conversations and is separate from your Proofoak subscription. As a starting hypothesis, a 200-row audit that takes 1.5–3 hours including review is often priced as a fixed fee. Use the service price and margin calculator to check what a job really earns after your time, tools and fees.

Data protection basics

Catalogue files can contain personal data — supplier contacts, staff names, sometimes customer details. Before you accept a file, agree with the client what's in it, how you'll store it and when you'll delete it, or ask for a version without personal data. The client onboarding checklist includes these steps.

When this route isn't right for you

  • You don't enjoy detailed, careful work.
  • You can't work in English with clients.
  • Your buyers' platforms already fix these issues automatically — test this early in conversations.

If any of these apply, Proofoak's fit result will say so and suggest another direction or a preparation step.

Other dossiers

Your first real test can start this week.

Free to start, no card. See the models that fit you, open a ready-to-run kit and plan your first 10-day test.

  • We never send messages, publish or spend money for you.
  • We don’t invent experience, clients or results.
  • Every claim is labelled: fact, estimate or hypothesis.