中文 EasyClaw
Research in 4 steps:① Market check② Find a niche③ Mine reviews (you are here)④ Margin math

How to Research Tumblers? I Scraped 5 Competitors' Reviews with Amazon Review Scraper and Found 3 Differentiation Opportunities

📅 Updated 2026-05-29 📂 Product Research · Step 3 ⏱ ~9 min 🛠️ Uses 1 EasyClaw skill
K
Operator K
3 years selling on Amazon, focused on the kitchenware category. This site documents my real process running the full tumbler workflow with EasyClaw.

By step three of product research, you've judged the market and outlined your niche. The next unavoidable question: why would my tumbler sell better than the ones already on the BSR charts?

You can't answer this from gut — you have to mine the answer from one place: competitor reviews. Reviews are the real complaints customers write after spending money — the weakest spots of existing products, and your opening to differentiate.

What problem this step actually solves

"Mining reviews" sounds mystical, but it really just answers 3 questions:

  • ① For tumblers that sell well now, what do customers complain about most?
  • ② Of those complaints, which can be solved by product design, and which are expectation problems?
  • ③ Can my tumbler improve on 1-2 of those pain points, so the listing can write a hook like "we solved XX"?
Reviews aren't negative information — they're a product requirements doc the market gives you for free. Reading them is far more accurate than sitting around imagining "what customers might want."

Why I stopped reading reviews manually

Three years ago I did it by hand too. Open the competitor page, sort by "recent" or "critical first," scroll one by one, paste a row to Excel when someone mentions leaking, another row for poor insulation…

❌ Manually reading reviews across 5 ASINs

· 50-100 reviews per competitor
· 5 competitors = a whole evening gone
· Pasted into Excel as plain text — can't see proportions
· Next day, tons mis-tagged and missed
· No confidence, can't make the call

✅ Using EasyClaw's Amazon Review Scraper skill

· 1 natural-language command
· 5 ASINs scraped in parallel
· 10 sample reviews each for a quick read
· Structured output: rating/title/body/date/type
· Total 2-3 minutes (depending on anti-scrape delay)

Why I don't use a pure scraper — I use EasyClaw

There are plenty of "review scraper" tools out there (Helium 10 / Jungle Scout both have similar features). But after a year of use, a scraper only solves half the problem — the other half is the key:

🛠️ Pure scraper tools

Scrape the reviews → hand you a JSON / CSV
the "make sense of the data" part is left to you

The biggest beginner pain: you've got a pile of reviews — how do you categorize them? High-rated competitors barely have any critical reviews, so where's the opportunity? How do you mine hidden pain points?
→ You're back to a whole evening of manual analysis — it didn't really solve the problem.

🤖 EasyClaw = "skill scrapes + LLM analyzes"

The Amazon Review Scraper skill pulls the raw JSON
→ EasyClaw's main LLM takes over that JSON and analyzes it
→ It tells you directly: which critical review matters most, what hidden pain points hide in the 4-5 star reviews, which pain points span multiple competitors…

This is what truly sets EasyClaw apart from a scraper — it doesn't just give you data, it tells you what the data means.

The division of labor across these two layers is the key — the skill scrapes the reviews completely, and the LLM reads, categorizes, and mines hidden pain points from the praise. You'll see what the LLM's categorized report looks like below.

Here's how I had EasyClaw do this

Two steps: install the skill, then send the command.

Step 1: install the skill (one-time)
EasyClaw packages every capability as an installable skill. Today's:

📦 Amazon Review Scraper
Actual capability: calls EasyClaw's local engine, opens a browser and paginates to collect Amazon product reviews, outputting structured data. Each review returns 6 fields — rating/reviewer/title/date-location/body/type, with optional image URL/video URL. Built-in 3-8 second random anti-scraping delay.

Installing is simple: open EasyClaw's skill marketplace, search "Amazon Review Scraper," and click Add — no commands needed. After installing, confirm "engine service running" in the bottom right.
The Amazon Review Scraper skill installed in the EasyClaw skill store, showing Use
📷 Search "Amazon Review Scraper" in the EasyClaw skill store; after clicking Add it shows "Use" — installed, no commands the whole way.
Step 2: send the command
I listed the 5 benchmark tumbler ASINs I'd selected and told EasyClaw:
/Amazon Review Scraper B0GJS619G5, B0F6C5GQ4T, B0FF9MMXM6, B0G4CDM3QN, B0GKQYGY2J

Use Amazon Review Scraper to pull 10 reviews from each of these 5 tumbler ASINs, focusing on 1-3 star critical reviews.

EasyClaw automatically converts the ASINs into review-page URLs (with the reviewerType=all_reviews parameter) and scrapes them in parallel via the browser engine.

🎬 Real demo: after sending the command above in EasyClaw, the Amazon Review Scraper skill scrapes the 5 ASINs in parallel and finally summarizes them into a comparison table with ratings, total review counts, and 1-3 star critical counts.

What the skill brings back

After the run, Amazon Review Scraper returned structured review data for each ASIN, each review carrying 6 fields — rating, reviewer, title, date-location, body, review type (with optional image/video URLs). The skill only "scrapes data" and is strictly forbidden from fabricating — if it can't fetch, it errors out.

But raw data is inert. The real value is the next step: having EasyClaw read, categorize, and mine hidden pain points from the praise.

After EasyClaw takes over the data, it does the categorization itself

This step is the key — the skill only "scrapes data," while the categorization is done by EasyClaw's main LLM after reading all the reviews. That's what sets it apart from a pure scraper.

I continued to EasyClaw:

"Read the reviews you just scraped, categorize by problem type, tell me the typical pain points and which ASINs they involve, focus on the 1-3 star critical reviews, and also check whether the 4-5 star reviews hide any complaints."

EasyClaw's main LLM (no extra skill needed) did a second-pass analysis on this data and output a categorized report. The result was unexpected — all 5 competitors rate 4.4-4.7, with almost no critical reviews:

EasyClaw output · 5 tumbler competitors' scraped reviews (40 total)real data
ASINProduct★RatingTotal reviews1-3 star
B0GJS619G5Konokyo FLOWPLAY Push-Button 18oz4.7350
B0F6C5GQ4TKEWIXY Cherry Water Bottle 18oz4.51530
B0FF9MMXM6Halloween Tumbler 40oz4.41821 (3★)
B0G4CDM3QNDog Affirmation Tumbler 40oz4.65260
B0GKQYGY2JBrüMate Era Flip 30oz4.55300
Seeing this you might think: "So few critical reviews — no opportunity here?" Quite the opposite. Beginners only look at 1-3 star reviews; pros mine the "hidden complaints" inside the 4-5 star praise. That single 3-star review and the gripes tucked into the praise are the real gold mine.

The only critical review (B0FF9MMXM6 · 3 stars)

"The straw keeps falling off the lid into the cup, very disgusting. Keeps drinks cold fine, but the straw issue is too annoying."

Even more valuable — EasyClaw also mined a batch of "praise-while-griping" hidden pain points from the 4-5 star reviews, categorized them by proportion, and gave entry suggestions:

EasyClaw's hidden pain-point categorization report for 5 tumbler competitors: problem-classification chart + three layers of key insight
📊 The "problem classification summary + three-layer insight" report EasyClaw's main LLM auto-generated after reading all the reviews (real screenshot)

Below is the text version of this report (for easy citation and search-engine indexing). First, the proportion distribution across 8 problem types:

EasyClaw LLM output · problem classification summary (share of all hidden complaints)LLM analysis
Problem typeShareSeverity
Lid / structural part durability 17.5%High
Straw design defect 15.0%High
Cleaning / hygiene difficulty 12.5%Medium
Insulation / cold retention below expectation 10.0%Medium
Spout odor 10.0%Medium
Color / texture deviation 10.0%Low
Size / capacity mismatch 7.5%Low
Accessory / attachment issues 5.0%Low

Now the three layers of key insight EasyClaw gave — this is what sets it apart from a pure scraper:

🔴 Critical-review core (the only 3-star)

  • The straw falling off is the only thing that drove an obviously critical review
  • Strong wording from the user: "Very disgusting"
  • Even with good insulation (keeps water cold), a single straw issue was enough to drop the rating to 3 stars

🟡 Silent killer (frequent hidden complaints in 4-5 star reviews)

  • Lid plastic parts cracking is the most-mentioned issue, but users generally still gave 4 stars and only mention it "in passing"
  • This means poor lid durability may be systematically underestimated — many people don't bother updating their review
  • 30% of reviews contain a hidden complaint, indicating high category satisfaction but plenty of room to improve experience details

🟢 Competitor openings (entry suggestions)

Problem typeEntry potentialSuggested direction
Straw falling offVery highThreaded-lock straw interface to eliminate loosening
Lid durabilityHighReinforced hinge / spring structure, certified for 2000+ open-close cycles

Here's the key part: how to read "opportunity" from this table

Most beginners stop here — "oh, so few critical reviews, no opportunity" — and give up. The key is to find 3 read signals from the hidden pain points.
1

High-rated niche — opportunity hides in the "praise-while-complaining"

These 5 competitors rate 4.4-4.7 with almost zero critical reviews — a beginner would judge "no opportunity." But "straw not leak-proof / leaks" was mentioned 3 times in the 4-5 star praise, across 2 ASINs. Users willing to give 5 stars still can't help griping — that signals a "grin-and-bear-it" real pain point, exactly the entry point.

2

See how many competitors a pain point spans

"Straw leaking" and "lid plastic cracking" each span 2+ ASINs — meaning it's not a single product's fluke but a category-wide common problem. A category-level common problem is far more valuable than a single-product defect: solve it and you differentiate against the whole niche, not just beat one competitor.

3

Read the user's exact words to reverse-engineer the hook

The only critical review's exact words were "the straw keeps falling off the lid into the cup" — users can precisely describe the "straw falling off" action, meaning they're aware of the straw structure. My future listing hook is set: "Secure Lock Straw — Never Falls In". A selling point reverse-engineered from real user language is 10× stronger than a vague term like "high-quality tumbler."

Stacking the three signals, I reach this round's differentiation conclusion:

🎯 Entry point = solve the two category-level common problems "straw falling off / leaking + flimsy lid plastic."
Hook = anti-falloff lock straw design + reinforced lid clasp + leak-proof seal structure.

Same data, two seller types decide completely differently

At this point, "mining the differentiation opportunity" is done. But how you use it diverges completely between premium FBA and dropship.

🟠 Premium FBA · reverse customization

Turn hidden pain points into a product redesign spec

Take this "straw leaking 3× / lid cracking 2×" data to a 1688 factory and tell them clearly: "I want an anti-falloff lock straw, a reinforced lid clasp, and it must pass both drop and seal tests." Sample 3-5 versions and test, then make a branded private mold once finalized.

Next action: take the differentiated selling point → find a 1688 factory to sample → test against competitors → finalize

🔵 Dropship · risk filtering

Treat hidden pain points as a "negative list" for selection

You can't change the product, but you can pick styles that won't blow up. When sourcing on 1688, use "straw falls off easily," "thin lid plastic," "leaks" as exclusion terms, and only pick styles rated 4.6+ where the buyer photos show a straw lock structure and a thick lid.

Next action: use hidden pain points in reverse → filter when sourcing on 1688 → list reliably-reviewed in-stock styles

Operator K's pitfall notes

I've fallen into these 4 traps myself — don't repeat them

  • Only looking at 1-3 star reviews: this time the 5 competitors had almost 0 critical reviews — anyone looking only at critical reviews gives up immediately. The real gold mine is the "praise-while-griping" in 4-5 star reviews — have EasyClaw scrape the praise too and analyze hidden pain points.
  • Treating a single-product defect as a niche opportunity: an issue appearing once on one ASIN may be a fluke. Check whether the pain point spans multiple competitors — only those across 2+ ASINs are category-level common problems worth differentiating on.
  • Not reading the user's exact words: a category label ("leaking") is abstract; the user's exact words ("the straw keeps falling off the lid into the cup") carry product-design direction. Extract listing hooks from the exact words.
  • Treating every gripe as an opportunity: "color differs from the photo" or "slow shipping" are expectation-management problems, not product defects — redesigning won't fix them. Only pick pain points you can improve at the design/craft level.

Differentiation opportunity found — next, run the numbers

💰

Next: can this differentiated tumbler actually make money (margin math)

The differentiated selling point is set (removable silicone seal + 360° leak-proof lid), but adding these improvements raises cost — can it still make money? The last step of product research is the math: break down the 1688 cost + inbound + full FBA cost, and use dual-mode standards to decide whether this product is worth committing to. It's the finale of the 4-step research method, and the final "go or no-go" decision.

FAQ about Amazon Review Scraper

Q: Can the Amazon Review Scraper skill really pull raw reviews?
Yes. The skill calls EasyClaw's local engine service (HTTP 127.0.0.1:10027), opens a browser and paginates to collect, returning 6-8 fields per review (rating, reviewer, title, date-location, body, review type, with optional image/video URLs). The skill is strictly forbidden from fabricating data — if it can't fetch, it errors out, as explicitly stated in the skill docs.
Q: How does the scraped JSON become a pain-point report?
This is the key — the skill only scrapes data; the categorization is done by EasyClaw's main LLM. After the skill outputs JSON, you simply tell EasyClaw "categorize by problem type, compute proportions, find keywords," and it reads all the reviews and analyzes them itself. This "skill scrapes + LLM analyzes" combo is the core of what sets EasyClaw apart from a pure scraper.
Q: How many reviews should I scrape per ASIN?
Depends on the goal. Quick read (judging whether a niche has many critical reviews): 10-20 per ASIN is enough — like this time, 10 each across 5 competitors quickly revealed "high-rated niche, opportunity in hidden pain points." Deep analysis (computing pain-point proportions): 100-300 per competitor recommended. The skill's commentNumber parameter maxes at 999 — adjust as needed.
Q: Will Amazon ban EasyClaw review scraping?
Amazon Review Scraper has a built-in 3-8 second random delay for anti-scraping. We recommend logging into a buyer account in the EasyClaw-launched browser before scraping, and not scraping 50 ASINs in one night from a single IP. With normal competitor-research usage (a few ASINs a day), I've had no issues in a year.
Q: A review says "leaks" but the product page claims leak-proof — how do I judge?
Look at the date_location field in the JSON. If the leaking reviews are concentrated in the last 3 months — usually a new-batch QC issue (that's an opportunity). If spread across years — it's a design defect (also an opportunity). Have EasyClaw aggregate by month to see it.
Q: Is review analysis still necessary for dropship?
Absolutely. Dropship can't change the product, but it can pick styles that won't blow up. I've personally tested filtering out styles with 5%+ critical-review rates, and my listed products' return rate dropped from 8% to 3%, stabilizing the store rating. The core competitiveness of dropship is "accurate selection," and review analysis is the selection tool.

🤖 Run your full Amazon tumbler workflow with EasyClaw

Product research → sourcing → listing → promotion → operations — every stage has a matching skill.
Install once, ask across the whole chain.

Try EasyClaw free →