How to Build an ICP From Customer Reviews
Let the People Who Stayed, and the People Who Left, Draw the Profile
Short answer
Build an ICP from reviews by tagging each one for fit, then comparing happy and unhappy reviewers. Tag who wrote it, the trigger, the previous tool and the outcome. Traits that cluster among happy reviewers form the profile; traits that cluster among unhappy ones become disqualifiers. Validate the result against your billing and churn data.
1. Why build an ICP from reviews?
Reviews are unprompted descriptions of fit, written by buyers. The happy ones show who gets value and in which words; the unhappy ones show who should have been turned away. An ICP needs both halves, and most internal workshops only produce the first.
- Ideal customer profile (ICP)
- A description of the type of customer who gets the most value from a product and is most valuable to the business, including the traits that should disqualify a prospect.
- Also known as: ICP, Ideal client profile
Internal ICP work tends to describe the customer the company wants. Reviews describe the customer the company actually serves well, which is often narrower and less glamorous. That gap is the useful part: it tells marketing where to stop spending.
Reviews also carry the vocabulary the profile needs. When a customer explains why they stayed, they name the trigger and the job in their own words, which is the raw material for targeting and copy. The guide on jobs to be done for marketers covers how to read those triggers.
2. Which reviews should I use?
Start with your own reviews on every platform, then add competitor reviews in two directions: their unhappy customers who match your strengths, and their happy customers who show a segment you should not chase.
| Source | What it tells you | Watch out for |
|---|---|---|
| Your own positive reviews | Who gets value, what triggered the purchase, the words they use | Solicited reviews skew toward your happiest customers |
| Your own negative reviews | Who should have been disqualified before buying | One loud review is not a segment |
| Competitor negative reviews | Buyers the rival serves badly, who may fit you | Only counts if you genuinely handle that complaint better |
| Competitor positive reviews | Segments the rival serves well, often worth avoiding | Their fit is not automatically your misfit |
| Support tickets and sales notes | The same signals, without the public filter | Internal wording, so translate to buyer language |
Platform matters for what you can see. Trustpilot reviews must be based on a genuine experience, generally within the last 12 months, but they rarely carry company details, so you infer them from the text [s1]. Software review sites such as G2 and Capterra often show a reviewer's role, company size or industry next to the review, which makes firmographic tagging much faster. The guide on competitor review monitoring covers choosing platforms by market.
3. What should I tag in each review?
Seven fields per review: who wrote it, the trigger, the alternative they used before, the job they hired the product for, the outcome or complaint, the words they used, and a fit verdict. One row per review, in a plain spreadsheet.
The example is for your company, an invented business selling booking software to independent physical therapy clinics. The review summaries are illustrative paraphrases written for this guide, not real reviews.
| Who (clues) | Trigger | Before | Job | Outcome or complaint | Their words | Fit |
|---|---|---|---|---|---|---|
| Owner, single room, treats all day | Receptionist left | Paper book and phone | Stop the phone interrupting treatment | Patients book themselves overnight | "no more phone tag" | Strong |
| Owner, three rooms, two therapists | Too many Monday no-shows | Norvane | Refill cancelled slots | Waitlist fills gaps | "empty Mondays" | Strong |
| Practice manager, multi-site group | Head office mandate | Sondera | Central reporting across sites | Missing multi-site reports | "not built for groups" | Misfit |
| Owner, mobile home visits | Wanted route planning | Spreadsheet | Plan travel between patients | No travel features | "cannot plan my day" | Misfit |
5. What does the finished ICP look like?
A short profile with observable traits, the triggers that make the customer act, their current alternative, the words they use, and a list of disqualifiers. It should fit on one page and be usable in ad targeting and sales qualification.
| Field | Your ICP |
|---|---|
| Who | Owner-managed physical therapy clinics with one to five treatment rooms, owner still treating patients |
| Triggers | Losing or not having a receptionist; a run of no-shows; opening a second room |
| Current alternative | Paper appointment book and text messages, or a generic booking app without a waitlist |
| Job | Keep the week full without spending evenings on the phone |
| Their words | "No-shows", "empty Mondays", "phone tag", "booking system" |
| Success looks like | Patients book outside opening hours, cancelled slots get refilled |
| Disqualifiers | Multi-site groups needing central reporting; mobile practices needing route planning; hospital departments |
Buyer personas then describe the people inside that profile, and the guide on buyer persona vs. ICP covers where one stops and the other starts.
6. What are the limits of review-based ICPs?
Reviews over-represent the very happy and the very angry, rarely include company details on consumer platforms, and describe the past. Use them to form the profile, then confirm it against retention data and recent sales conversations.
- Selection bias: companies that ask for reviews mostly ask happy customers, so your own positive reviews overstate fit.
- Missing details: consumer review platforms rarely show company size or role, so some tags are inferences. Mark inferred tags so you know which ones to verify.
- Recency: reviews describe why people bought months ago. If the market moved, recent sales calls are a better guide to current triggers.
- Competitor reviews show their failures, not your wins. A rival's unhappy customer only fits your profile if you genuinely solve their complaint.
Oppira collects Trustpilot reviews for the competitors you track on the Basic plan and above, so the competitor half of the sheet refreshes itself. The Playbook keeps Ideal Customer Profiles as their own layer, and when review patterns shift, the agent can propose an edit with the reviews that prompted it.
Key Takeaways
Reviews show fit and misfit
Happy reviews describe who gets value; unhappy ones describe who should have been turned away. An ICP needs both.
Use competitor reviews in both directions
Their unhappy customers may fit you. Their happy customers often mark a segment you should not chase.
Tag seven fields per review
Who, trigger, previous alternative, job, outcome or complaint, their words and a fit verdict.
Traits must be checkable before the sale
Size, decision maker, current tool and recent events can be verified. Attitudes cannot.
Disqualifiers are the most used row
They tell marketing which audiences, keywords and partnerships to stop paying for.
Validate against retention
If customers matching the profile do not retain better, the reviews misled you, usually through selection bias.
Frequently Asked Questions
Sources
- Guidelines for reviewers Trustpilot, June 2026.Reviews must reflect a genuine experience, generally within the last 12 months.
Explore More
Related analyses, benchmarks, and industry insights
Related Guides
Glossary Terms
Turn reading into a reaction
Oppira watches your competitors, keeps your playbook current, and drafts the response. See your market clearly by tomorrow.