How to Find Product Opportunities in Your Competitors' Customer Reviews

Choose rivals from your own data, not a review site's category list:

• Closed-lost reasons and competitor fields in your CRM.
• "What were you using before?" answers from onboarding and sales calls.
• Churn notes that say where customers went.

Past five, you are doing category research, not competitive research. If your real rival is a spreadsheet or an in-house tool, there are no reviews to read, so use churn and lost-deal notes for that one.

Stick to the last 12 months; older reviews describe a product that may have changed. As a rule of thumb, 30 to 50 reviews per competitor, weighted toward 2 and 3 stars, is enough to see its top five problems.

Don't skip the idea board. A request with hundreds of votes that has sat at "under review" for a year is real pain the competitor has chosen not to fix.

How Do You Collect Competitor Reviews Ethically?

• Read in the browser and take notes, or use data the site licenses. G2's terms, last updated July 9, 2026, bar collecting content "through automated, programmatic, or mechanical means" without consent, and bar using it to train or evaluate AI models (G2 Terms of Use).
• Treat Reddit the same way. Since May 2024, Reddit has said businesses need a contract to use its data commercially (TechCrunch).
• Respect reviewers. Quote sparingly, link to the original, and never use reviews to contact their authors.

Rewrite each complaint as a problem you can test. "Exports fail on reports over 10,000 rows" is testable; "wants better exports" is not.

• Log the basics: competitor, source, date, stars, reviewer company size or role, and type (weakness, unmet need, or strength).
• Group into 8 to 12 clusters across all competitors, so problems that span rivals stand out.
• Count distinct reviewers, not mentions. Someone who posts the same complaint on G2 and Reddit counts once.
• Set a bar for a pattern: as a rule of thumb, three independent reviewers, or the same problem at two competitors.
• Keep segments apart. Enterprise admins and solo users complaining about "permissions" usually have two different problems.

AI can take the first pass on reviews you collected by hand. Use a business plan or the API so your inputs are not used for training, and do the counting yourself in a spreadsheet.

Below are customer reviews of [Competitor], copied from public review pages.
Each has an ID, date, star rating, reviewer company size, and text.

1. List every distinct problem each reviewer describes, as a one-line problem
   statement in neutral language: what goes wrong, for whom, and when.
   Do not propose a solution.
2. Tag each problem: weakness (a complaint), unmet need (something the product
   does not do), or strength (praise).
3. Copy the reviewer's exact words as the quote. If there are none, skip it.
4. Group the problems into 8 to 12 clusters and list the review IDs in each.

Do not count or rank the clusters. I will count unique reviewers myself.

Reviews (ID | date | stars | company size | text):
[paste]

This is the step most competitor research skips, and it decides which box each cluster lands in. Search your tickets, call transcripts, interviews, churn notes, and closed-lost reasons for the same problem.

Read the quotes before you call it a match: "export" complaints about speed and about file format are different problems.

Complaints tell you what annoys people; switching reasons tell you what makes them pay the cost of leaving. Switching is the stronger signal for a product bet.

• G2 asks reviewers "What did you switch from?" and their reason for switching, and its vendor analytics group those answers into switching themes (G2 docs).
• Capterra reviews often name the product the reviewer switched from and why they chose the new one, under "Switched from" and "Reason for choosing."
• Reddit threads asking for alternatives usually state the trigger in the first post.

Tag the trigger when you see one: a price or packaging change, a scale ceiling such as seat limits or slow reports, a missing integration with a tool the buyer just adopted, or a redesign that broke a daily workflow. Watch the dates, too: complaints that start the month after a price change are an opening that closes once the competitor adjusts.

Name each opportunity after the customer problem, never the rival. "Beat Rival A on reporting" dies the day Rival A ships a fix; "Reports time out for accounts over 50,000 contacts" stays true and testable.

Opportunity: [Customer problem, in plain words, no competitor name]
Type: Weakness to exploit / Unmet need / Strength to match
Competitor evidence: [N distinct reviewers, N competitors, last 12 months]
  "[quote]" (G2, [Competitor], [Month Year], [company size])
  "[quote]" (Reddit, [Month Year])
Your evidence: [N accounts or prospects] or "none found"
  "[quote]" ([source], [segment])
Switching signal: [Appears in switching reasons? Where?]
Segment: [Who has this problem]
Next step: [Discovery interviews / Spec / Positioning only]
Re-check: [Date of next sweep]

Then rank it against the rest of your backlog on the same terms: affected accounts, revenue, strategic fit, and effort. Competitor evidence adds weight; it does not get its own lane. For the ranking method, see how to decide what to build next.

Findings in the "their weakness only" box go to marketing as a short positioning brief, quotes included.

A gap list goes stale within a couple of quarters, because competitors ship fixes and break things.

• Quarterly: a full sweep of every competitor, timed before planning.
• Monthly: a 15-minute skim of new reviews for your top two rivals.
• Right away: after a competitor's major launch, price change, acquisition, or outage.

Reuse last quarter's clusters so you can see which problems are growing, and drop any gap a competitor's changelog shows they have closed.