Choose for variety, not only convenience
A sample made entirely of straightforward products may hide the questions that will dominate later work. Include a normal item, a family with variations, an incomplete measurement, and a product with an unresolved media reference. The sample should reflect meaningful differences in the intended range.
The information owner can explain which cases are ordinary and which are unusual. That context helps keep a small project focused without designing an entire system around a rare exception.
Limit the output as well as the input
Select one named destination format and the minimum information needed to prepare it. A limited scope becomes hard to review if every possible marketplace or sales document is included from the beginning. Other destinations can be recorded as future requirements.
State which parts are manual in the first slice. Manual review can be appropriate while the team learns which rules are stable enough to implement repeatedly.
Agree the decisions the slice should answer
The review might establish where variant-specific attributes belong, how a missing unit is handled, or whether the buyer’s categories can be mapped reliably. These are concrete questions against which the proposed work can be assessed.
Identify who can resolve each question. A software reviewer may confirm the behavior of a screen, while a product-information owner confirms the meaning of the content. Both may be needed for a useful result.
Use the result to scope the next step
A first slice can produce a model note, a small prototype, sample output, and a list of unresolved dependencies. It should make the next decision clearer, whether that is a broader build, a configured existing platform, or further source-data work.
A proposal defines the specific outputs, review responsibilities, and exclusions. It does not assume a full catalog migration or a continuing support service. Contact the company with a short description of the source and destination to begin that discussion.