How to mine reviews, Reddit, and app stores for the words buyers use, and match each quote to an awareness stage so the right hook meets the right audience.
It is 11pm and you are on hook variation fourteen. "Track your macros effortlessly." "The smarter way to eat." Every line is grammatical, on-brand, and dead. Meanwhile, somewhere on Amazon, a one-star review of your biggest competitor reads "I lasted nine days before this app made me feel like a failure," and that sentence would beat everything in your doc.
That is voice of customer research for ads in one scene. The copy you need already exists, written by the people you want to reach, sitting in public reviews and forum threads. The job is collection and sorting. No copywriting genius required.
The takeaways
Five sources cover most categories: Reddit, Amazon reviews, the Apple App Store, Google Play, and the niche forums where your buyers complain to each other. Search them for your competitors' names and for the product category itself, and you will have more raw language in an hour than a focus group produces in a month.
Where to dig within those: 1-star to 3-star reviews of competitors (the unmet need, stated with feeling), 5-star reviews of anything adjacent (the outcome people brag about), and Reddit threads that start with "does anyone else". App store reviews are underrated for this because the rating forces a verdict; the text underneath explains it in the reviewer's own vocabulary, usually within two sentences.
The phrases themselves, word for word. From every usable quote, pull 2 to 4 short phrase markers, the fragments a buyer would recognize as their own thought. Then attach two labels: the emotion behind it (frustration, relief, embarrassment) and the register (clinical, slangy, resigned). Those labels matter later, when you decide which audience sees which line.
Here is the gap paraphrasing destroys. A positioning doc says "low retention." The buyer says "another app that gives up on me after a week." Run the second one as a headline and a scrolling thumb stops, because the reader has had that exact thought in those words. Smooth the phrase into marketing English and the recognition is gone. Keep a spreadsheet of raw quotes; resist the urge to tidy them.
Eugene Schwartz laid this out in Breakthrough Advertising in 1966: every prospect sits at one of 5 awareness stages, from Unaware through Problem-Aware, Solution-Aware, and Product-Aware up to Most-Aware. Tag each mined quote with the stage its author was in when they wrote it. Then match the stage to the temperature of the campaign.
A glowing "this fixed my tracking in one afternoon" is most-aware language. Show it to cold traffic and it bounces, because that audience does not yet believe it has a tracking problem. Cold prospecting wants problem-aware lines ("I gave up logging meals by week two"). Retargeting can carry product-aware and most-aware quotes, since those viewers already know who you are.
This sorting step is what most review-mining guides skip, and it is where mined copy usually dies. A strong quote at the wrong stage loses to a mediocre line at the right one.
Frequency is seductive. Forty quotes complain about price, so price feels like the angle. But buyers mention price with a shrug; it shows up in every thread about everything. The six quotes charged with embarrassment ("made me feel like a failure") carry the purchase decision, because that is the feeling someone will spend money to end.
So rank pains by intensity, not frequency. Read for charged language: capital letters, profanity, time-wasted arithmetic ("three Sundays rebuilding this report"), anything close to shame. One high-intensity pain with six quotes behind it outranks a lukewarm one with forty.
Pulling hooks from real customer language raises your hit rate; it does not exempt you from noise. When the first mined hook posts a wild ROAS in three days, the same skepticism applies as with any creative. I wrote a separate piece on telling a winner from a lucky coin before you scale it.
I did this manually for years, spreadsheet and all, and eventually built it into Adscalr's audience research because the sorting was eating my Sundays. It searches the same five sources (Reddit, Amazon reviews, App Store, Google Play, plus forums you point it at, seeded with your and your competitors' URLs), keeps 2 to 4 phrase markers per quote with emotion and register, maps every quote to one of Schwartz's five stages, and folds them into 4 to 6 personas, each backed by at least 3 real quotes. Demographics on those personas are labelled as AI estimates, because that is what they are. If you would rather skip the spreadsheet stage, that is what the audience intelligence does.
This is the thinking behind Adscalr.
See the product →