You find product opportunities in app store reviews by reading a fixed sample (100 is enough), tagging each review with a small set of themes, counting the tags, saving a verbatim quote for each theme, and then converting the biggest themes into opportunity statements. The method is simple, but it only works if your tags and counting rules are fixed before you start reading.
This guide gives you a review-coding sheet, a walk-through of 100 reviews for one app, and a template for turning themes into opportunity statements. It also fits into a 14-day schedule, so you can finish a portfolio-ready piece of analysis in two weeks.
Why review analysis is a core product management skill
Reviews are unsolicited, specific and time-stamped. People write them when something has annoyed or delighted them, so they show where the product’s promise and the real experience diverge. A product manager uses them to spot patterns early, to check whether a release made things better, and to find gaps competitors have left open.
Reviews also have limits. They over-represent angry and delighted users, and they say little about people who quietly stopped using the app. They describe symptoms (“it keeps logging me out”) rather than causes. Your job is not to treat the star rating as truth but to treat the text as evidence that needs interpreting.
Choose one app and build a defensible sample of 100 reviews
Pick one app you actually understand as a user. A food delivery, UPI payments, learning, ride-hailing or fitness app works well because reviews are plentiful. Avoid apps with fewer than a few hundred reviews, since you will not be able to sample properly.
Sampling rules to fix before reading
- Source: use one store (for example Google Play) and note the date you collected the reviews.
- Recency: take reviews from roughly the last three to six months so you are not analysing bugs that were fixed long ago.
- Mix of ratings: take 20 reviews at each star level (1 to 5), so the sample is balanced rather than dominated by whatever is most common.
- Language: keep English and clearly readable Hinglish or Telugu-in-Roman script only if you can code them confidently; otherwise skip and replace them.
- Exclusions: skip empty reviews (“good”, “nice app”), duplicates and reviews that only complain about something unrelated to the product, such as a courier’s behaviour. Record how many you skipped.
Write these rules at the top of your sheet. A reader should be able to repeat your sampling and get a similar result.
Free 2-Week CareerSwami Internship
Online, 14 days, with a free offer letter and a free completion certificate once you finish the work. Check with your college that it fits your requirement before you enrol.
The review-coding sheet: columns, tags and counting rules
Use a spreadsheet with one row per review. Keep the tag list short (8 to 12 tags). A long tag list produces thin counts and inconsistent coding.
| Column | What to enter | Rule |
|---|---|---|
| ID | R001 to R100 | Number in the order collected |
| Date | Review date | Needed to spot post-release spikes |
| Stars | 1 to 5 | Copy exactly |
| App version | If shown | Leave blank if not shown |
| Primary tag | One tag from the list | The main complaint or praise |
| Secondary tag | Optional second tag | Only if clearly a separate point |
| Sentiment | Negative, Mixed, Positive | Judge the point, not the star count |
| Request? | Yes or No | Yes only if the user asks for a specific feature or fix |
| Quote | Verbatim text, trimmed with ellipses | Never rewrite; keep spelling as written |
| Job or context | What the user was trying to do | Write “unclear” rather than guess |
A starter tag list (adapt to your app)
- Bugs/crashes: app freezes, errors, failed actions.
- Performance: slow loading, battery or data use.
- Login/account: OTP not arriving, forced logouts, verification.
- Payments/refunds: failed payment, money deducted, refund delays.
- Pricing/fees: hidden charges, discounts that do not apply.
- Usability: hard to find things, confusing flow.
- Missing feature: something the user expected and cannot do.
- Support: unhelpful or unreachable help.
- Content/catalogue: wrong, stale or limited listings.
- Trust/privacy: spam, permissions, data worries.
- Praise: what is working, tagged with the specific thing praised.
- Other: use sparingly; if it exceeds 10 reviews, split it.
Counting rules
- Count the primary tag only for your headline ranking. Show secondary tags separately.
- Compute each tag’s share as count divided by the number of coded reviews, not 100 if you skipped some.
- Also compute the average star rating for each tag. A common tag with a 1.5 average hurts more than a common one with a 3.5 average.
- Re-read your first 20 rows after finishing. If you would tag any differently, fix them and note the change.
Walking through 100 reviews: an illustrative run on a food delivery app
The numbers below are illustrative, made up to show the method. They are not real data from any app, and no tool was run to produce them. Your own counts will differ.
Step 1: Code the reviews
Suppose you coded 94 usable reviews after skipping 6 empty or duplicate ones. After the first pass, the tally might look like this.
| Primary tag | Count | Share of 94 | Avg. stars | Requests |
|---|---|---|---|---|
| Payments/refunds | 21 | 22% | 1.6 | 9 |
| Support | 16 | 17% | 1.4 | 5 |
| Pricing/fees | 13 | 14% | 2.1 | 2 |
| Bugs/crashes | 11 | 12% | 2.0 | 3 |
| Usability | 9 | 10% | 3.0 | 4 |
| Missing feature | 7 | 7% | 3.4 | 7 |
| Praise | 12 | 13% | 4.8 | 0 |
| Other | 5 | 5% | 3.2 | 1 |
Step 2: Read the table, not just the top row
Payments/refunds leads on volume and has a low average rating, so it is a strong candidate. Support is smaller but has the lowest average, which suggests intense frustration. Missing feature is small but nearly every review in it contains a request, so it is a source of concrete ideas rather than complaints.
Step 3: Capture quotes
For each major theme, save two or three verbatim quotes, and mark which ones show the context. Illustrative examples of how entries might look:
R017 | 1 star | Payments/refunds | "money deducted but order failed... refund still not received after 6 days"
R052 | 1 star | Support | "chat bot keeps repeating same answer, no way to talk to a person"
R078 | 3 star | Missing feature | "please add option to split bill with friends"
Step 4: Look for sub-patterns
Inside the 21 payments reviews, re-tag by cause: failed payment with money deducted (say 12), refund delay (6), coupon not applied (3). This second-level split is where a product opportunity becomes specific. “Fix payments” is vague; “show refund status and expected date inside the order screen” is something a team can discuss.
Step 5: Check for time clusters
Sort by date. If most payment complaints fall within one fortnight and mention the same app version, the issue may be a regression rather than a long-standing gap. That changes the recommendation from “redesign” to “investigate a release”.
Turning counts into priorities: a simple scoring table
Counts alone do not rank opportunities. Score each theme on four factors from 1 (low) to 3 (high), then discuss the result rather than treating it as a formula.
| Factor | Question to ask | High score means |
|---|---|---|
| Frequency | How many of the coded reviews mention it? | Roughly 15% or more |
| Severity | Does it cost users money, time or trust? | Average rating under 2 |
| Clarity | Can you state the user’s goal and the point of failure? | Quotes agree on the cause |
| Fixability | Could a product team plausibly act on it? | Within the product’s control |
Be honest about fixability. A complaint about delivery-partner behaviour may be frequent, but it may sit outside the app. Mark it and explain the reasoning instead of dropping it silently.
Template: converting a review theme into an opportunity statement
An opportunity statement describes a user problem and its evidence without prescribing a solution. Use this structure for each of your top three themes.
Opportunity: [short name]
Who: [type of user, as far as the reviews show]
Job they are trying to do: [goal in plain words]
Where it breaks: [step, screen or moment]
Evidence: [count] of [coded reviews] ([share]%) mention this;
average rating [x]; sample quotes: "[quote 1]" (R###), "[quote 2]" (R###)
Impact if unsolved: [money, time, trust, churn risk, stated carefully]
Not yet known: [what reviews cannot tell you]
Possible directions: [2-3 ideas, clearly marked as hypotheses]
How to test: [interview question, survey, prototype, metric to check]
Filled-in illustration (hypothetical numbers)
Opportunity: Visible refund status after a failed payment
Who: Customers who paid by UPI or card and saw the order fail
Job they are trying to do: Know whether their money is safe and when it returns
Where it breaks: After payment failure, no status is visible in the order screen
Evidence: 12 of 94 coded reviews (13%) describe money deducted with no
clear status; average rating 1.5; "money deducted but order failed..." (R017)
Impact if unsolved: Users may stop trusting the payment step
Not yet known: How often failed payments happen overall
Possible directions: Refund tracker in the order screen; automatic message with expected date
How to test: Ask five recent users how they checked their refund; compare support ticket volume if available
Notice that “Not yet known” is mandatory. It stops you overclaiming and shows reviewers you understand the data’s limits.
Biases that distort review analysis, and how to reduce them
- Negativity skew: unhappy users write more. Report shares within the sample, not as shares of all users.
- Recency spikes: one bad release can flood the recent reviews. Check dates before generalising.
- Vague praise: “great app” tells you nothing. Code it only if it names a feature.
- Your own bias: if you already dislike the app, you will read ambiguous reviews as complaints. Code sentiment on the written point, and have a classmate independently code 20 reviews to compare.
- Developer replies: these can reveal known issues, but do not treat them as confirmed facts about causes.
A 14-day plan for completing this analysis
If you are doing this alongside the free two-week online internship from CareerSwami, a schedule like this keeps you on track. It is a suggested pacing, not an official programme requirement.
| Days | Task | Output |
|---|---|---|
| 1-2 | Choose the app, set sampling rules, build the sheet | Blank coding sheet with tag definitions |
| 3-4 | Collect 100 reviews and clean the sample | Sample log with skipped counts |
| 5-7 | Code every review and capture quotes | Completed sheet |
| 8-9 | Tally, compute shares and averages, add sub-tags | Summary table |
| 10 | Second-coder check on 20 reviews | Agreement note |
| 11-12 | Score themes and write three opportunity statements | Opportunity page |
| 13 | Build a one-page summary with charts and quotes | Draft deliverable |
| 14 | Review, trim, and prepare to present | Final PDF or slides |
Satavahana University students: using review analysis alongside your degree and career support
Satavahana University’s University Profile page says it offers programmes through University College in Karimnagar and University PG College in Godavarikhani, and that it has over 160 affiliated colleges. Its constituent colleges offer M.Sc. (including Computer Science), M.B.A. and M.C.A. Product analysis suits students from any of these routes, since it needs reading, structured thinking and a spreadsheet more than coding.
For curriculum, the Courses & Syllabus page links to a B.Sc. Data Science syllabus for 2025-26, an MBA IV-semester revised syllabus and a B.Sc. Computer Science syllabus. Those PDFs have not been checked here for project or internship rules, so do not assume that a self-run exercise like this one counts for anything in your programme. The Placement Cell page names a Placement Officer and says the cell was established in 2010, but its listed activities date from 2013 to 2015, so check current support directly. Because affiliated colleges may follow different practices, ask your own college as well.
Verification steps for Satavahana students
- Ask your department office, in writing, whether an online internship or self-directed project can be recorded in your programme, and which document sets the rule.
- Open the syllabus PDF for your programme and year on the official Courses & Syllabus page and look for project work or internship sections.
- If you study at an affiliated college, ask the principal’s office whether the college or the university sets internship rules.
- Contact the Placement Cell using the details on its page and ask what career support is currently offered and whether a certificate from an online internship is acceptable for any process.
- Check the almanac for your programme and semester on the official Almanacs page so you schedule the 14 days around exams and breaks.
Presenting your findings so a reviewer can trust them
A strong one-page summary has five parts: the sampling rules, the tag table, three themes with quotes, three opportunity statements, and a short limitations note. Put the evidence before the idea. If you are sharing it in a portfolio or interview, keep your raw sheet available so someone can check a number.
Avoid claiming that your findings represent all users, that a fix would raise ratings, or that you know why users behaved as they did. Say what the reviews show and how you would test further.
Frequently Asked Questions
Is 100 reviews enough for app store review analysis?
It is enough to practise the method and spot the biggest themes, not to prove precise percentages. Treat shares as indicative. If a theme appears in 20% of your sample, it deserves attention, but a gap between 8% and 11% means little.
Should I code one tag per review or allow several?
Use one primary tag for ranking and an optional secondary tag for extra detail. That keeps counts from adding up to more than your sample and still captures reviews that raise two issues.
How do I handle reviews that are in Hindi, Telugu or mixed languages?
Code them only if you understand them well. If you translate a quote, label it as your translation and keep the original beside it. Otherwise exclude and record the exclusion.
Can I use a tool or AI to tag the reviews?
You can use one to suggest tags, but read and check each review yourself. Automated tagging often merges distinct problems, such as failed payment and refund delay, which hides the specifics you need.
What is the difference between a complaint and a product opportunity?
A complaint is what a user said. An opportunity is a pattern you have evidenced, tied to a user goal and a point of failure, with a clear test. The statement template above is designed to force that step.
How should I handle fake or incentivised reviews?
Watch for near-identical wording, bursts on one date and praise with no specifics. Exclude suspicious entries, record how many, and mention it in your limitations.
Can I compare two competing apps with this method?
Yes. Use identical tags and sampling rules for both, then compare shares and average ratings. Where a competitor’s complaints are low and the first app’s are high, you have a candidate gap, but confirm with user interviews.
What should I put on my resume or portfolio from this exercise?
Describe the work factually: the app, the sample size, the tags, and the three opportunity statements. Do not claim business results, since you had no access to the company’s data.
Related Product Management guides
- Khaja Bandanawaz University Students: Writing a Product Brief for an Ed-Tech Feature
- Manav Bharti University Students: Using Product Management Internship Learning in Your Final-Year Project
- Woxsen University Students: Getting Ready for a Product Case Round With a Practice Scorecard
- All Product Management internship guides
Ready to put this into practice?
Join the free two-week CareerSwami online internship in Product Management. The internship, the offer letter and the completion certificate are free.














