Amazon Review Mining

What is Amazon review mining?

Amazon review mining is the habit of reading the reviews of best-selling products in one category, mostly the 2 and 3 star ones, to find what buyers wanted and did not get. Those buyers already proved they will pay for a solution. What they write after paying tells you exactly where the current products fall short.

Here is why the middle stars matter most. A 1-star review is about a broken box, a late delivery or a faulty unit, and the 5-star crowd just says “love it”. The 3-star reviewer liked the product enough to keep it and is annoyed enough to explain the gap, in sentences that start with “I wish”, “the only problem” and “I ended up”. That is your raw material.

When should you use it?

• Physical products: kitchen, baby, pets, fitness, beauty, travel, home office. Any category with clear best sellers and hundreds of reviews per product.
• Digital products and info products: read the 3-star reviews of the top books in a niche. Readers say which chapter they wanted and never got, and that missing chapter can become a course, a template pack or an app.
• Services hidden in products: reviews that say “I had to pay someone to install it” point to a service business.
• Skip it for B2B software (use competitor review analysis on G2 and Capterra instead) and for categories with only a handful of reviews per product, where one angry buyer distorts everything.

How do you do it, step by step?

1. Pick one narrow category and 5 to 8 best sellers. Open Amazon’s Best Sellers page and click down to a subcategory, say “Kids’ Lunch Boxes” instead of “Kitchen”. Choose products from different brands and price levels, each with at least 200 reviews. Copy the product name, price and star average into your sheet.

2. Filter the reviews by star rating. Amazon’s own shopping guide says you can open “All Reviews”, filter by 1 to 5 stars, sort by “Most recent” or “Top reviews” and tick “Verified purchase”. Start with 3 stars, then 2 stars. Sort by Most recent so you read about the version on sale today.

3. Search inside the reviews. The same guide describes a search bar inside the reviews section. Type these words one at a time: wish, but, instead, returned, ended up, hack, for my. “For my” is the sneaky one: it reveals unexpected buyers (“for my mum with arthritis”), and an unexpected buyer is a niche.

4. Code each review with one gap type. Use 5 codes and stick to them:

   • Missing feature (“wish it had a lid that locks”)
   • Quality failure (“seal cracked after 3 washes”)
   • Wrong fit for a person or use (“too heavy for a 4-year-old”)
   • Workaround (“I put cling film under the lid”)
   • Instructions or setup (“no idea how to assemble”) Aim for 20 to 30 coded reviews per product.

5. Count across products, never within one. A gap that shows up in 1 product is that brand’s mistake. A gap that shows up in 4 of 6 best sellers is a CATEGORY gap, and that is the one worth your time. Build a simple grid: gaps down the side, products across the top, a tick where the gap appears.

6. Note who is buying and what they pay. From the text, write down the buyer (parent, carer, nurse, gym beginner) and the price of each product. If the gap persists at every price point, money alone has not fixed it.

7. Check the AI summary last. Amazon shows a short AI-written “review highlights” paragraph on many product pages. Read it after your own coding, as a cross-check. It averages everyone, so it hides the small, angry group you are looking for.

What does a worked example look like?

Say Hana, a hypothetical mother of two in Rotterdam, wants to launch a physical product for parents. She picks kids’ bento-style lunch boxes and 8 best sellers between low and premium prices.

She reads 3-star reviews first, then 2-star, sorted by Most recent, and codes 160 reviews in one weekend. Her grid shows three category gaps:

• Yogurt and sauces leak between compartments: 6 of 8 products.
• A 3-year-old can’t open the latches alone: 5 of 8 products, mostly in reviews from nursery parents.
• The lid warps in the dishwasher: 4 of 8 products.

The workaround column tells her even more. Parents use silicone muffin cups as liners and add a strip of tape to stop leaks. Two parents mention buying a second box only for wet food.

Here’s her conclusion: the pairing of “leakproof” with “openable by a toddler” appears in no product’s reviews as solved. Leak-free boxes have stiff latches; easy boxes leak. That is a sharp problem statement for interviews with nursery parents, before she touches a factory.

Which free tools help?

Amazon explains its star filter, sorting and review search in its own article on how to read Amazon customer reviews.

What mistakes should you avoid?

• Reading only 1-star reviews. You learn about couriers and faulty batches. The middle stars hold the product gaps.
• Mining one product. Every product has flaws. Only cross-product patterns signal an opening.
• Ignoring the date. A complaint from 4 years ago may describe a version the brand fixed. Sort by Most recent.
• Copying the competitor and fixing one flaw. If your only difference is “same box, better latch”, a big brand can copy you in a season. Look for a gap tied to a specific buyer.
• Letting AI do the reading. Summaries flatten the odd, specific complaint that becomes a business. Read at least 100 reviews yourself.

How do you know it worked?

You have a grid with at least 5 products and 100+ coded reviews. Our rule of thumb: a gap counts when it appears in at least half of the best sellers and comes with workarounds or returns.

Go: you can name the gap, the buyer who feels it most and the price range they already pay. No-go: gaps are scattered and every product fails differently. The category may be fine as it is; move to the next one.

What should you do next?

Map the products you studied with a competitor analysis grid and a competitor pricing map, then test your gap with real buyers in problem interviews. For apps, run the same logic with app store review research. When you want structure and feedback on the whole journey, the startup game for women by Fe/male Switch pairs you with an AI co-founder for each validation step.

FAQ

Why read 2 and 3 star reviews instead of 1 star?

Because 1-star reviews mix product gaps with shipping damage, defects and buyer error. Middle-star reviewers kept the product and still explain what is missing, which is the most precise description of a gap you will find.

Does Amazon review mining work for digital products?

Yes, through books. Read 3-star reviews of the top books in your niche to see which topics readers wanted covered. Those missing chapters point to courses, templates, communities or apps.

How many reviews should I read?

Our rule of thumb is 20 to 30 per product across 5 to 8 products, so 100 to 240 in total. Fewer than that and one memorable rant can steer your whole idea.

Updated 26 October 2026 · Written by the Fe/male Switch team

Mark this method as done