Data Mining
In Brief
Data Mining is a family of generative research methods that pull customer segments, pains, and demand signals out of data that already exists — your support inbox, your website, public search trends, app-store reviews, transaction logs — instead of generating new data through interviews or surveys. This page is the orientation page for the family. The actual work happens on the sub-method pages: Customer Support Analysis, Search Trend Analysis, and Web Traffic Analysis. Pick one based on which data source you can reach for the question you are asking.
Common Use Case
You suspect there is signal hiding in data you already have access to — your own product analytics, support inbox, web traffic, or public sources like search trends and app-store reviews — but you have not systematically mined it. Before commissioning new interviews or surveys, you want to extract whatever segments, pains, and demand signals are already sitting in existing data, then pick a sub-method based on the data source you can reach.
Helps Answer
- Which existing data source is most likely to answer my current question?
- Are there customer segments or pain patterns I can see without running new research?
- Is there enough demand signal in the market to justify building anything?
- What patterns recur across customer behavior, complaints, or searches?
- Where should I focus deeper qualitative research?
Description
Data Mining is a family of generative research methods that extract insights from data sources that already exist — support tickets, web analytics, search-engine query volume, app-store reviews, transaction logs — instead of generating new data through customer interaction. The work is detective work, not interview work: you hunt for patterns, segments, and demand signals in records that were created for some other purpose. Because the data is already there, these methods are usually faster and cheaper than primary research, and they often surface candidate segments or pains that you can then validate in interviews.
Choosing the Right Data Mining Method
The three methods below differ in which data source they read. Pick the one matching the data you have access to (or can acquire most cheaply). Founders often run two or three in sequence — for example, mining support tickets to find a recurring pain, then checking search-trend data to confirm the pain has visible market demand.
AI Prompt
I want to mine existing data for [PROBLEM HYPOTHESIS — ONE SENTENCE].
Data I already have: [SUPPORT TICKETS / WEB ANALYTICS / TRANSACTION LOGS / NONE]
Data I could get cheaply: [GOOGLE TRENDS / APP STORE REVIEWS / COMPETITOR REVIEWS / ETC.]
My traffic / volume: [VISITS PER MONTH / TICKETS PER WEEK / NA]
What I want to learn: [PAIN / SEGMENT / DEMAND SIZE / FUNNEL DROP-OFF]
[CONSTRAINTS]: [B2B vs B2C, regulated industry, no product yet, etc.]
Recommend 1-2 Data Mining methods from this family that fit my situation. For each:
1. Name the method and why it fits the data I have.
2. Name the riskiest assumption it would NOT validate.
3. Suggest the next data source to check if this one passes.- Customer Support Analysis Mines support tickets, chat transcripts, and app-store reviews to surface common pains and the segments suffering from them. Best used when you already have a product in market with meaningful inbound support volume to cluster.
- Search Trend Analysis Counts query volume on terms related to the problem, using search-engine trend tools and AI-question platforms. Best used when you do not yet have customers and need to size demand for a problem before building.
- Web Traffic Analysis Mines visitors to an existing website for referral sources, on-site search, conversion funnels, and drop-off patterns. Best used when your site already has meaningful traffic; on low-traffic sites the patterns will be noise.