Desk research and review mining

Research you can do this afternoon

If the word "interview" makes you want to skip Module 2, this lesson is for you. It requires talking to nobody. Thousands of your potential customers have already written down their pains, hopes, objections and exact vocabulary — in reviews of your competitors, in forum threads, in search boxes. It's public, free, and sitting there.

This is desk research: mining what already exists. It's faster than primary research and biased in a known direction — the vocal minority writes reviews, the silent majority doesn't — so you'll treat it as one evidence source among several, not the verdict. But an hour of it, done systematically, produces the raw material for every headline, ad and email you'll write in Modules 3–9. In the customers' words instead of yours.

Review mining: their reviews are your research department

Read reviews the way a researcher does, not the way a nervous owner does.

Competitors' 1–3★ reviews are your opportunity list. Every specific complaint about a rival is an unmet need in your market, described precisely, by someone who paid money. Their 5★ reviews are your copy bank — the exact language of delight, which is what your own messaging should sound like.

Your own reviews are the cheapest evidence you own. Tomas has 23 Google reviews averaging 4.6 that he has never once read as data — what do patients praise unprompted? What words do they use for their pain? He's been guessing at questions his patients already answered in public.

The discipline is one word: verbatim. Copy and paste. The moment you paraphrase — turning "took us three months to set up and we're a five-person firm" into "setup is slow" — you re-insert your own assumptions, and the data becomes you again. The exact words are the data.

Where to mine — the EU source map

Pick two or three. Depth beats coverage.

45 minutes at Storkflow

Lena spends 45 minutes on G2 and Capterra reading reviews of two bigger onboarding tools. Her tagged harvest, four entries of it:

Pain — "took us three months to set up and we're a five-person firm" Objection — "support only in English" Desired outcome — "clients sign everything before the kickoff call" Their language — accountants say "client intake", not "onboarding"

Look hard at the last one. Storkflow's homepage, its Google Ads term and its category vocabulary all say "onboarding" — the industry's word. Its buyers apparently say "client intake". That single line changes the homepage H1 and the Module 6 keyword plan, and it cost €0 and three quarters of an hour. Note also that the pain and the objection are opportunities Storkflow can claim only if they're true of Storkflow — mined evidence tells you what to say, not permission to say it.

Search boxes: what people ask when nobody's watching

Type your category into Google slowly and read the autocomplete. Read the "People also ask" boxes. Try a free keyword tool if you like, but the suggestions alone are the point: they are questions real people typed, ranked roughly by how often, in their own vocabulary.

For Tomas this is close to an answer to his stated obstacle — "I don't know whether they search at all before choosing a clinic." The suggestions that appear under "physiotherapie graz…" are what people in his city actually type at the moment of looking. Whatever they turn out to be, they're evidence of the words and worries in play — which is all this module needs from them. (Module 6 turns the same data into a keyword plan; here it's purely voice of customer.)

Communities, and the competitor teardown

Forums and groups. Search Reddit and the specialist forums of your trade for your category and your competitors' names. You're looking for the threads where someone asks "what do you all use for…?" — the alternatives people name, the warnings they give each other, and the workarounds they admit to are all evidence of the jobs and forces you drafted in Lesson 2.2.

Competitor messaging teardown. Open three competitors' homepages and record, for each: the headline promise, the three benefits they push hardest, the proof they show, and what they don't say. Then hold the teardown against their 1–3★ reviews. The gap between what a rival promises and what its unhappy customers complain about is a map of unclaimed positions — arguments nobody in your market is currently winning.

Count things: quantitative signals without a survey

Everything so far collects words. Now add numbers — not a survey, just honest counting of what's publicly there.

An impression says: "most competitor reviews complain about delivery." Evidence says: "11 of the 30 most recent reviews mention delivery — counted 14 August 2026." A count with a denominator and a date can be checked, compared with a recount in six months, and weighed against other themes. "Most" can only be argued with.

Countable signals, all free:

• Theme frequency — of the last 25–30 reviews of a competitor, how many mention price? Speed? Setup? Your Lesson 2.2 anxieties?
• Your own tags — once your ten quotes are collected, count them: how many pains versus objections? The distribution is itself a finding.
• Presence counts — how many reviews does each local rival have, and how recent? How many search suggestions in your category are questions about price versus questions about how it works?

Two honesty rules travel with every count. Small numbers are directional, not proof — 11 of 30 reviews is a strong signal about that platform's reviewers, not a fact about your whole market; write the denominator down and say so. And the vocal-minority bias never goes away — you are counting people angry or delighted enough to write. Interviews (next lesson) exist to reach everyone else.

Record evidence, not impressions

A finding that isn't written down with its source becomes an impression within a week, and impressions are what Lesson 2.1 sent you here to escape. Every entry in your swipe file carries the verbatim quote, one tag — Pain / Desired outcome / Objection / Their language, four tags only, more taxonomy than that kills the habit — and its source. Every count carries a denominator and a date.

One legal paragraph, because it matters. Reading public posts is fine. Copying quotes into your research notes, anonymised — no names, no usernames, no profile links — is fine. Harvesting identifiable people into a contact list or CRM to market at them is not; under GDPR that's processing personal data without a lawful basis. You are building research notes, never prospect lists.

Set a timer for 45 minutes and start with your competitors' worst reviews.

---