Guide

How to analyse customer reviews.

A five-step method a team can follow by hand, the mistakes to avoid, and what changes when software does it continuously.

In shortTo analyse customer reviews, gather them from every site you sell on, attach each one to the product it concerns, group them by topic, count how each topic moves over time, and decide one fix per topic. The trend of each topic tells you more than the average rating.

The five steps

  1. Gather the reviews from every place you sell. Marketplaces, retailer sites, your own store and the communities where your category is discussed. Keep the date, the rating, the source and the exact product.
  2. Attach each review to the exact product and variant. Size, shade, flavour or format. A problem often concerns one variant only, and an average hides it.
  3. Group the reviews by topic. Packaging, texture, taste or fit, scent, results, value for money, delivery, and whether the product matched its page. One review can mention several topics; note for each whether it is praise or a complaint.
  4. Count each topic over time. Month by month, look for the topics that grow. Compare before and after a change such as a new batch, new packaging or a new product page.
  5. Decide one action per topic. Reply to the shoppers, change the page, change the product or inform the supplier. Then measure the same topic again a month later.

Four mistakes to avoid

  • Looking only at the average rating. It moves late and does not say why.
  • Looking only at complaints. Praise tells you which reasons to buy belong at the top of the product page.
  • Ignoring rival products. Their reviews show what shoppers compare you with, and where they are weaker.
  • Doing it once. A review analysis is out of date as soon as a batch, a page or a rival changes.

By hand or with software

By hand, the method works for a few hundred reviews and one or two products. Beyond that, two things become hard to keep up: grouping every new review by topic, and noticing a change in the week it happens.

Where Purples fits

Purples is a consumer intelligence platform for beauty, food and beverage, and fashion brands. It runs these five steps continuously on your reviews from Amazon, Sephora, Ulta, Trustpilot and Reddit, and on the rivals you name. Each figure links to the reviews behind it, and the replies and product page fixes are drafted for your approval.

Questions, answered straight

Related questions.

How many reviews do I need before the analysis means something?

There is no fixed number. A topic mentioned in a handful of reviews is a signal to watch. Treat it as a trend when it keeps coming back over several weeks.

Should I analyse my competitors’ reviews too?

Yes. They show what shoppers compare you with, which complaints your rivals have not solved, and which claims work in your category.

How often should reviews be analysed?

Continuously if you can, and at least every month. A problem with a batch or a package shows in reviews within days.

See this on your own products.

Book a 30-minute demo. A person replies within 48 hours.

Book a demo