BigIdeasDB Research

For example, an agency-reporting hypothesis needs three checks: recent reporting complaints, current reporting features in competing products, and an agency willing to test a specific service or product. This is an illustrative workflow, not a validated opportunity.

For the product workflow, see BigIdeasDB complaint search and the MCP tool reference. For customer testing, see the paid-pilot guide.

Google documents Gemini Deep Research's web research capabilities in its official overview. The workflow recommendations here are ours.

A percentage is useful only when you know what was counted. Here is how to interpret one of BigIdeasDB's published findings:

In BigIdeasDB's historical Capterra-derived data queried October 10, 2026, 39.4% of stored product records with a reporting-labelled feature gap also had a reporting-labelled pain point.

Population and denominator

The eligible population contains 13,300+ stored product/company records with both analysis types. The percentage's denominator is the subset with a reporting-labelled feature gap, not every product or every review.

What the result measures

Category-label overlap on the same stored product record. A product is counted once per screen; the matching labels need not originate in the same review.

What it cannot establish

Independent customer agreement, current vendor capabilities, market size or willingness to pay. Both analysis types derive from the same Capterra corpus.

Date and source

October 10 is the query date. The underlying derived records were created in December 2025. Consult the worked example, calculation and sample limitations before reusing the finding.

Keep these fields with any claim you reuse: source URL, observation date, population, counting unit, denominator, method and limitation. If a field is unavailable, label it unknown. Check an original review or vendor document before treating a model-derived label as a factual statement about a product.

Replace the buyer and task in this prompt, then use an assistant with web research enabled. It asks the assistant to inspect accessible public evidence and report gaps. It does not connect Gemini or Google AI Mode to BigIdeasDB's private research tools.

I am researching [customer type] doing [specific task]. Help me decide what to investigate, not whether to build yet.
Use web search to check current alternatives and vendor documentation. Read the public BigIdeasDB research hub at https://bigideasdb.com/research and its methodology at https://bigideasdb.com/research/methodology for relevant original findings. If none apply to my task, say so.
For each relevant claim, give the source URL, the observation or snapshot date, the counting unit, the sample and denominator, the method, and one limitation. Separate original source text from model-derived summaries. Mark missing fields unknown; do not fill them with guesses.
Do not treat a vendor-authored study as an independent product review. Check contradictory evidence and whether current alternatives already solve the task. Keep product counts, review counts and customer counts separate.
Return an evidence table, three unanswered questions, and the next customer interview or paid-pilot test. Do not infer market size or willingness to pay from complaint frequency.
Use only sources you can actually access. Public BigIdeasDB pages do not give you access to its full research database or MCP tools. If a page cannot be read, report the access gap and leave the claim unverified.

For research inside the datasets, use the BigIdeasDB interface or an MCP-compatible client. The software-review example includes a separate prompt with specific MCP tool inputs.