Product Research: Generate a Publishable 4-Page PDF with Talos in 15 Minutes

Product research is a high-frequency activity for product managers, investors, and operations roles. To gauge how far a product has come, you typically need to check official pricing, revenue databases, founder retrospectives, and competitor reviews simultaneously. Doing a presentable research report manually usually takes 1 to 2 days. This time, using Talos, a 4-page product research PDF was generated in 15 minutes.

The case study chosen is Buffer. It is modest in scale and has publicly available data, making it suitable for verifying the completeness and credibility of the output.

The difficulty in product research centers on three points: data is scattered across seven or eight places, vendor claims are mixed with third-party estimates, and the final write-up easily becomes a mere piling of materials. Even harder is credibility judgment; the same number can differ threefold across official sites, reviews, and revenue databases. Not citing sources is as good as not doing the work.

The goal this time was to produce a report ready for immediate distribution, with every key figure traceable to its origin.

Step 1: Write the Research Requirements as a Structured Prompt

Research quality depends on the framework. By laying out the questions to answer, word count per section, and data handling rules clearly from the start, the output structure remains stable. The complete Prompt used this time:

Research a product and output a Markdown file ready for direct publishing, following these eight sections:

1. Product Name: Chinese and English names, domain, launch year
2. Product Introduction: Target audience, problem solved, features offered, business model and monetization
3. Current Status: User scale, annual revenue, growth rate
4. Company: Company name, founding date, location, headcount, funding status
5. User Personas & Scenarios: Who uses it, provide 2-3 specific scenarios
6. Official Site & Trial Experience: URL, what the free version offers, credit card required or not
7. Feature List: Grouped by Publishing, Creation, Engagement, Analytics, Collaboration
8. Pricing & User Reviews: Tiered pricing table, plus one authentic user review

Requirements:
- 1-3 sentences per section, the entire document should not exceed two A4 pages
- Six specific metrics must be found: Users, Revenue, Growth Rate, Headcount, Pricing, Launch Date
- Write "Undisclosed" if not found, do not substitute adjectives
- Estimated figures must be marked as "Estimated"
- Label the source for every key figure, provide a credibility table at the end
- Conclude with one verdict: What is its real profit engine, is its moat genuine, and what market gap does it leave open?

This Prompt contains four key elements:

• Fixed Structure: The eight-section sequence is locked, allowing direct horizontal comparison across every output.
• Numbers First: The six hard metrics must land on specific figures, avoiding purely qualitative descriptions throughout.
• Source Constraints: Mark as undisclosed if not found, label estimates, eliminating fabrication.
• Conclusive Wrap-up: The final sentence must answer what this means for us.

Step 2: Talos Executes the Research Automatically

Upon receiving the request, Talos proceeds through five stages:

1. Locating Official Information: Opens the official homepage, pricing page, and about page to confirm product positioning, tiered pricing, and company entity.
2. Checking Revenue & Scale: Searches revenue databases and public metric pages to obtain ARR, paying customer count, and team size.
3. Finding First-Hand Retrospectives: Searches for founder retrospectives and media reports to fill in key milestones on the growth curve.
4. Cross-Referencing: Lays out numbers for the same metric from different sources side-by-side, explicitly noting conflicts.
5. Writing According to Framework: Composes the eight sections segment by segment, controlling word count per section, and appends the credibility table.

The entire process takes about 15 minutes.

Step 3: Output Content Structure

Using Buffer as an example, the output is a 4-page PDF with eight body sections and two closing parts.

Part	Content
Sections 1-2	Product positioning & business model: Per-channel subscription, starting at $5 per social media account.
Section 3	Scale: 57.5k paying customers, $26.8M ARR, +13% YoY.
Section 4	Company: Founded in 2010, 74 employees, only $3.5M Series A raised.
Sections 5-6	Audience & Experience: Creators, small teams, agencies; free version covers 3 channels.
Sections 7-8	Features & Pricing: Three tiers, 20% saving with annual billing.
Closing	A one-sentence verdict, plus a 6-row credibility table.

Three figures among these are worth examining individually: the $26.8M ARR comes from the official public metrics page; 57.5k paying customers comes from a third-party revenue database; the +13% growth rate corresponds to a re-acceleration after it stagnated around the $20M mark for eight years. Only by viewing these three numbers together can one grasp the true return velocity of the distribution layer business.

Metric	Manual Creation	Talos
Total Time	1-2 days (estimated)	15 minutes
Information Sources	Manually open 5-8 pages	Automated search & cross-reference
Data Credibility	Judged by memory	Source labeled per figure, summarized at end
Output Format	Handwritten in a document	Structured Markdown plus 4-page PDF
Reusability	Framework rebuilt each time	Prompt saved as template, swap product and run again

15 minutes, a 4-page PDF, all eight sections supported by data, with 6 key figures sourced and credibility rated.

• Single Research Task: Compressed from 1-2 days to 15 minutes, provided a stable framework is in place first.
• Repeat Research Tasks: Once the Prompt is saved, swapping a product name generates another report, with near-zero marginal cost. This difference is magnified tenfold when creating a competitive matrix.
• Information Quality: Every figure comes with a source, conflicts are explicitly marked. Team reviews no longer ask "Where did this number come from?".
• Format Consistency: With the eight-section structure fixed, multiple reports can be directly assembled into a single comparison table.

1. Lock the Framework First: The eight-section structure, once established, enables horizontal comparison across multiple reports.
2. Require Source Labeling: Every key figure needs an origin for credibility to have a foundation.
3. Set Rules for Missing Values: Write "Undisclosed" when not found, label estimations, avoid filling gaps with adjectives.
4. The Final Sentence Must Conclude: Answer what this means for us, rather than ending with a mere list.

Q: What if revenue data cannot be found?

A: Mark it as "Undisclosed." Do not substitute with phrases like "significant scale." For public companies, check financial reports; for private ones, check revenue databases and public founder retrospectives. If neither source has it, leave it blank and note this in the credibility table.

Q: Sources disagree on the same number. Which one takes precedence?

A: Prioritize the official source, followed by third-party reviews with verification dates. If a significant discrepancy exists, retain both numbers and note the conflict in the text. Do not take a median yourself.

Q: Can the generated PDF be edited?

A: Yes. The Markdown is the source file. Edit it and re-export. The PDF is for external distribution; the Markdown is for ongoing maintenance.

Q: Can this method be reused for other types of research?

A: Yes. Substitute the eight-section framework with a structure for industry research, competitive matrix, or investment memo, while keeping the rest of the rules unchanged: prioritize numbers, ensure source traceability, and end with a conclusive statement.