- 01
CoverThe question the issue asks, in a headline.
- 02
What it isThe company in plain words, and the problem it is for.
- 03
ProductHow the product works, in concrete terms.
- 04
CustomersWho uses it, and how that base has grown.
- 05
OriginHow it started, and the first move that worked.
- 06
CompetitionWhere it stands among rivals, and where the pressure is.
- 07
MoatsWhich advantages are real, and which are not.
- 08
Business modelHow it makes money, and what the price is paying for.
- 09
RisksWhat could break the bet.
- 10
VerdictThe call, and the time window it is for.
- 11
SourcesWhat this issue knows, and what it doesn’t.
Startup Research
10 questions, one magazine.
How I read a company: understand the tech, analyze the business.
The analysis framework
10 questionsEach question gets its own file and its own method. Open the 8 moats checklist under Competition.
Raw material
Podcast transcriptsVia podcast-dig: founder and investor interviews, exact quotes with timestamps.
Public researchProduct, pricing, funding, press.
Hands-on useTrying the product myself.
Also: AI Term Explainer films for the tech
How I take a startup apart10 questions · one file each
- 1People & product
10Team
Did the founders live the pain?
Founder-market fit, from podcast transcripts.
20Product
Mission-critical or nice-to-have?
PMF signals; moat claimed vs. judged.
- 2Market & moat
30Market
Why now, and how big, bottom-up?
Own TAM, not analyst reports.
40Competition
What is the moat, really?
7 Powers, cross-checked with Gokul Rajaram's 8 moats.
- Data · Proprietary data that compounds with use.
- Workflow · Wired into how people work every day.
- Regulatory · Licenses or compliance rivals can't easily get.
- Distribution · A channel to customers others can't buy.
- Ecosystem · Others build on it, so leaving costs more.
- Network effects · Each new user makes it better for the rest.
- Physical infrastructure · Hard assets that take years to copy.
- Scale · Unit costs fall as it grows.
Rubric: Gokul Rajaram on 20VC, Mar 2026
- 3Evidence
50Data
What do the filings say?
SEC Form D, Companies House; every ARR sourced.
60Expert views
What do named investors and the founders say?
Founder claims from podcast transcripts, timestamped.
65Positioning
How has the story changed over time?
Podcast transcripts lined up by date.
- 4Durability
70Sustainability
Does it survive the next model?
Stress test against model progress.
80Risks
What kills it?
Five risk categories, one kill shot, with odds.
90Tech
How is it built?
Six layers: bought or built, and where's the moat.
- Then: the magazine ↓
The writing framework
one issueThe research runs long. What I publish is a business-magazine issue, built on this spine.
Rules
- One issue, a shared spine. Cover to sources. A company may add pages of its own.
- Headings are claims. Each section title states its conclusion, not its topic.
- Lead with the point. Every paragraph opens with what it means; evidence follows.
- Quotes with timestamps. Each founder quote links to the second it was said, and gets one line on what it implies.
- Estimates say so. Anything undisclosed is a range marked as my estimate.
- End on what would change it. Odds on the biggest risk, a time window, three signals to watch.
The same research can also be published as a video — one more channel for this work, not a separate method. The Clay analysis is also on YouTube ↗.
What the skill learned from
the libraryThe skill reads a living library of public essays, talks, and frameworks when it researches a company. Titles link to the source. Founder interviews are pulled first through podcast-dig, before it searches the open web.
The YC circle
What We Look for in FoundersPaul GrahamUsed to judge early founders when traction is still thin, and whether this is a real startup.
Startup PlaybookSam AltmanUsed to read a seed company end to end: the product, the users, and what the team should prove next.
How to Evaluate Startup IdeasKevin HaleThe rubric for scoring an idea before there is traction.
Tarpit ideasDalton CaldwellUsed to stress-test ideas that everyone likes, and whether the founders have a right to win.
Earnestness and misfit foundersGarry TanUsed to judge whether a founding team is the real thing, especially unpolished or contrarian founders.
If your product is Great, it doesn’t need to be GoodPaul BuchheitUsed to judge whether a first product is truly focused, and to separate founder quality from the idea.
Big-firm GPs
The Only Thing That MattersMarc Andreessen / a16zUsed when the open question is demand: is the market pulling the product, or is the team pushing it?
Revenue Quality & Marketplace PhysicsBill Gurley / BenchmarkUsed to test whether growth deserves the valuation — unit economics and market quality, not the story.
Zero to One: The Seven QuestionsPeter Thiel / Founders FundUsed on category-creating bets: what the company believes that others do not, and whether it can become a monopoly.
Founder-problem fit, pre-mortems, and pre-paradesRoelof Botha / SequoiaUsed at seed and Series A: is the founder fused to the problem, and does the timing survive a pre-mortem?
Gene Pool EngineeringVinod KhoslaUsed on technically heavy bets: does the team actually cover the risks that could kill the company?
Spikes, barrels, and ammunitionKeith Rabois / Founders FundUsed to judge early teams: is there a genuine spike, and can the founder ship?
Angels and the AI era
How to Angel InvestNaval RavikantUsed at the angel stage, when there is almost no data and the decision is a read on the founders.
AI Market ClarityElad GilUsed to judge whether an AI market is investable now — timing and structure, not just the team.
Pattern BreakersMike Maples Jr. / FloodgateUsed on ideas that look too early: a real non-consensus bet, or only contrarian and wrong.
Software 3.0Sarah Guo / ConvictionUsed on AI-native startups: a product in use, or only a capabilities demo.
The Unicorn ClubAileen Lee / Cowboy VenturesUsed to check pattern-match claims about teams against actual base rates.
The New Business of AIMartin Casado / a16zUsed on AI companies to test whether they scale like software or like a services firm.
Moats and how the report is built
7 PowersHamilton HelmerThe competition scorecard: the company and its top two rivals, scored cell by cell.
The 8 Moats of Enduring Software CompaniesGokul RajaramThe moat lens used in the Clay issue: eight advantages scored so the real ones stand apart from the rest.
Playbook, Power, Bull/BearAcquired (Ben Gilbert & David Rosenthal)Shapes the playbook read and a bull/bear case in which the bear has a real voice.
海外独角兽拾象Shapes the opening bet and the market section: thesis first, then why now and the kinds of players.
Company reportsContrary ResearchThe 11-section skeleton, so a company report never misses a part, and the prose style of the write-up.
Date and source on every numberSacraEvery number in the report must carry a date and a source.
Venture capital due diligenceWall Street PrepThe due-diligence checklists the research agents work through: team, product, business model, and risk.