Research Methodology and Data Sources

Suppose you want to investigate reporting software. Searching feature-gap labels alone finds 5,100+ product records. Requiring a reporting match in both the feature-gap and pain-point categories reduces that set to 2,000+. This narrower set gives you a place to inspect requests alongside complaints. It does not establish that these products currently lack reporting, or that their users would buy your solution.

The screens match category labels containing reporting, integration or customiz, after trimming and lowercasing. Labels can match more than one screen. These are three illustrative themes, not an exhaustive taxonomy or a ranking of the best SaaS markets.

A product is counted once per theme and analysis type, even if it has several matching records. The two matches need not come from the same review. A missing match can reflect the way a model labelled the complaint, rather than absence of the underlying problem. Read the source text before rejecting it.

What to investigate next

1. Choose one buyer and task, such as agencies assembling client reports. This is an illustrative hypothesis, not a result established by the aggregate table.
2. Inspect reviews for that task. Record the obstacle, consequence and current workaround. Exclude generic complaints about a different workflow.
3. Check current product documentation and talk to users about the last time the problem happened. This feedback predates the query date.
4. Test a specific offer with those buyers. A repeated complaint helps choose the interview; it does not establish willingness to pay.

Sample and limitations

Population

40,900+ processed feature gaps and 39,900+ processed pain points, joined by their common product/company records. Two feature-gap-only products and four gap records are excluded.

Historical dates

Derived records were created December 4–9, 2025. The company-level latest-review dates range from April 25, 2011 to December 2, 2025. This range is not the date coverage of every underlying review.

Missingness

42 feature-gap labels are blank. Company-level latest-review dates are unavailable for 382 records. They remain in the population; a blank label matches no screen.

Interpretation

No current vendor capability, customer count, market size, market-wide complaint rate or purchase intent can be inferred. AI-derived labels and targeted review collection limit the result.

The percentage is calculated as 100 × products with both matching labels / products with a matching feature-gap label, using unrounded counts.

Start with the customer task. Use another source to check whether the same task and consequence appear there, rather than treating keyword matches as proof.

Keep the sources separate until you have checked dates, units and duplicate evidence. Do not add review counts, extracted pain points and product counts into one demand number.

Connect the BigIdeasDB MCP server to a compatible assistant, then adapt the prompt below to your buyer and task. Tool names and inputs are documented in the MCP tools reference. The prompt retrieves a bounded sample to investigate; it does not reproduce the full-corpus aggregate study above.

I am investigating client reporting for small agencies, not deciding to build yet.
Use the connected BigIdeasDB tools to retrieve evidence:
1. search_capterra_feature_gaps: query "reporting", limit 25.
2. search_pain_points: query "reporting", source "capterra", limit 25.
3. search_capterra_reviews: query "reporting", max_rating 3, limit 25.
4. search_reddit: query "client reporting spreadsheet", time_filter "year", limit 25.
For each useful result, show the source, source URL if returned, observation date if known, product, customer role, task, complaint and workaround. Mark unavailable fields as unknown. Separate original source text from AI-derived summaries. Keep only evidence relevant to agencies and explain each exclusion.
Group matching tasks, not just matching keywords. Separate repeated product records from independent reviewers. Do not calculate market-wide percentages from these bounded search results.
Return three hypotheses with supporting evidence, contradictions, current alternatives to check, and one interview question each. If a result does not support a hypothesis, say so. Do not invent quotes, URLs, customer counts, current features or willingness to pay.

Use web search to check current alternatives and product documentation. Use connected research tools to retrieve the relevant feedback. If a tool returns no source URL or observation date, keep that field unknown and verify it separately.

For the decision after research, see how to test willingness to pay with a paid pilot.