SellerSort
Try it free

Whose data is in your reviews? What to strip before you analyze them

A review belongs to the customer who wrote it, and a review export often carries more of them than the text: names, email addresses, order numbers, sometimes a delivery address pasted into a complaint. Before you send reviews to any analysis tool, ours included, keep only the review text, the rating and the date, and remove anything that identifies a person. The analysis does not need it, and you avoid moving personal data you have no reason to move.

What this comes down to

  • Review exports often include names, emails and order numbers alongside the text; analysis needs none of them.
  • Keep three columns for analysis: the review text, the star rating and the date.
  • Check the review text itself too, because customers sometimes paste order numbers, phone numbers or addresses into complaints.
  • Before using any tool, read where the data goes, how long it is kept, and whether it is used to train models.

SellerSort Team · Updated

What a review export usually contains

Platforms export more than the words. Depending on the platform, a review file can include:

  • The reviewer’s name or display name
  • An email address, especially from store platforms and review apps
  • An order number or customer ID
  • Photo links, which can show faces or homes
  • The text, the rating and the date

Only the last line is needed to understand what customers are saying. Everything above it is personal data that belongs to the customer, not to the analysis.

The analysis needs what was said, not who said it.

Strip it before you analyse

  1. Open the export in a spreadsheet.
  2. Delete every column except the review text, the rating and the date.
  3. Search the text column for @, for long runs of digits, and for words like order, street and phone. Customers often paste these into complaints.
  4. Replace what you find with a placeholder such as [order number].
  5. Save the result as a new file and analyse that one.

If you need to link a finding back to a customer later, for a reply or a refund, keep a row number in the stripped file and the original export somewhere private. The link back stays with you.

Four questions to ask any analysis tool

  1. Where is the text sent to be analysed, and in which country is it processed?
  2. How long are uploaded files and results kept, and can you delete them sooner?
  3. Is your data used to train models?
  4. Is anything pasted, rather than uploaded, stored at all?

SellerSort’s answers

We would rather you check than trust us, so here they are, briefly:

  • Review text is sent to TypeSafe, the company whose model does the classification. TypeSafe says its services are hosted in the United States, and that it will not train or fine tune models on your input.
  • Uploaded files and generated output are deleted automatically after 7 days.
  • Reviews pasted into the analyzer are not stored by SellerSort.
  • Results of file jobs stay with your account, so your history remains available.

The privacy policy has the full detail, including every third party involved. Even with those terms, stripping personal data first is the better habit, for any tool.

Questions

Do customer reviews contain personal data?

Often. Review exports can include names, email addresses, order numbers and photo links, and the review text itself sometimes contains order numbers, phone numbers or addresses.

What should I remove from reviews before analysing them?

Keep only the review text, the rating and the date. Delete name, email, order and customer ID columns, and search the text for pasted order numbers, emails, phone numbers and addresses.

Does SellerSort use my reviews to train AI models?

The text is classified by TypeSafe, which says it will not train or fine tune models on your input. Uploaded files are deleted after 7 days, and pasted reviews are not stored by SellerSort.

Read next