Video summary

How to Research Stocks like a Wall Street Analyst

Main summary

Key takeaways

Educational

Main ideas / lessons conveyed

  • The goal of stock research is to develop repeatable “common business sense”, so you don’t rely forever on rigid templates or checklists.
  • The research process is framed into four parts:
    1. Where to start (how to choose which stocks to research)
    2. What to look for (how to analyze the business)
    3. Crafting an investment thesis (how to turn understanding into an argument with numbers)
    4. How to find information (sources, transcripts, filings, and alternative methods)
  • A key mindset: “destroy” your thesis by actively trying to find ways you could be wrong.
  • Many analytic tools (e.g., SWOT/MBA-style frameworks) are less important than building understanding from first principles:
    • How the company creates value
    • How it captures value
    • How it protects captured value (moat/defensibility)
  • Valuation/thesis should be tied to quantitative returns (DCF and/or reverse DCF), not just narrative.

Methodology / step-by-step instructions (with detailed bullets)

Part 1 — Where to start (how to pick stocks to research)

  • Use idea-sourcing channels (4 main options):

    • Stock screeners
      • Example: Yahoo Finance screener
      • Mentions Joel Greenblatt “Magic Formula” style screening using earnings yield and ROIC
      • Not necessarily the most compelling, but useful for generating initial candidates.
    • 13F filings
      • Use them to see what major investors/managers hold.
      • This can be a starting point (i.e., “copy” what top managers are buying).
    • Social research (Twitter/X and Substack)
      • Read others’ investment ideas and research notes for leads.
      • Mentions a friend’s Substack: “Mostly Borrowed Ideas.”
    • Consumer-first idea generation (recommended)
      • Identify public companies behind products you use in everyday life.
      • Rationale: if you like a product, you may detect consumer surplus (you receive more value than you pay), which can signal a good business.
      • Examples of consumer-driven discovery:
        • Chipotle (interest started from liking it)
        • Apple/iPhone (interest came from believing PC/Android substitutes were inferior)

  • Kick out stocks quickly (focus on your circle of competence):

    • If you can’t understand the business, remove it from your list early.
    • You can expand beyond your circle later, but for now:
      • For biotech/pharma: you must understand the science and basic products.
      • For semiconductors: you must understand the technology/science basics (e.g., “Semiconductors 101”).
      • For most investors: prefer sectors where understanding is more accessible (e.g., retail/food service and many consumer/tech businesses).

  • Prioritize among many candidates:

    • Use intuition and signals such as:
      • A stock trading at a lower multiple
      • News/sentiment that draws your attention
    • You can also study great companies even if expensive—either:
      • for future timing, or
      • to learn what a strong company looks like.

  • Quick financial “weed-out” scan:

    • Review P&L and balance sheet basics:
      • debt levels
      • whether it’s loss-making
    • If a company is consistently loss-generating, it can be harder to analyze—so you may temporarily pass unless you’re ready for the complexity.
    • Mentions tools like Fintel / Fiscal AI:
      • Example: Fiscal AI for fast scanning of operating income trend, revenue growth, debt, cash, operating cash flow, stock-based comp, and capex.
    • Even with tools, still read the actual annual report (10-K) later.

Part 2 — What to look for (core business analysis principles)

  • Avoid a huge checklist (e.g., “52 things,” “100 things”):

    • The point is you shouldn’t rely on constantly referencing a rulebook.
    • Instead, build patterns through repetition so the key issues become obvious.

  • Use three first-principles questions (main framework):

    1. What is the business / product?
      • What business are they in?
      • Why do customers come to them?
      • What preferences/needs does the product satisfy?
      • Example style given: explain why customers choose Chipotle (healthy enough, quick, convenient, relatively cheap).
      • Extend questioning until it’s clear (example: why Moody’s vs S&P Global vs Fitch for credit ratings).
    2. How do they make money?
      • Don’t stop at “they sell ads” (e.g., Meta).
      • Identify the actual monetization mechanism(s):
        • Are advertisers bought-and-placed through generic ad inventory, or tied to return metrics (e.g., ROAS)?
        • Example: an Amazon-like business needs deeper unpacking:
          • commission varies by product category
          • different fees for logistics
          • different fees for advertising services
    3. How defensible is the business?
      • Ask what prevents competitors from copying it.
      • Reframe as:
        • Create value
        • Capture value
        • Protect value
      • Emphasizes deep reasoning from first principles vs memorizing “moat categories” (network effects, scale advantages, etc.).

  • Practice “frame problem” awareness when reading disclosures:

    • Annual reports/investor days contain many facts; your job is to decide what’s relevant.
    • Research often becomes “pulling on threads”:
      • when something seems suspicious, investigate.
    • Accept that:
      • you might chase rabbit holes
      • sometimes you won’t reach a perfect conclusion
      • you’ll need to estimate how much research is enough eventually.

  • Build confidence by attempting to disprove the investment:

    • Treat research as a process to try to destroy the investment.
    • Generate hypotheses (even “crazy” ones), then test them with evidence.
    • Example: when researching lending/credit behavior (via an illustrative credit-card/risk-standards “thread” approach).

  • Example “layer-by-layer” narrative building (Meta, Jan 2023):

    • Identify issues:
      • TikTok competition
      • Apple tracking transparency limiting data
      • FTC concerns about divestiture
      • EU issues
      • Reality Labs heavy spending
      • Transition to Reels affecting monetization timing
    • For each issue, form an opinion:
      • “Not worried about FTC outcome because…”
      • “Even if divestiture occurs, value might be greater split…”
    • The thesis becomes confidence from working through each concern.

  • Develop “common business sense”:

    • Over time, you should instinctively react to what sounds obviously wrong (bad business plan examples).
    • Example caution: rapid expansion into lending for people without credit history.
    • Use history and analogs:
      • even if the situation is novel (e.g., SpaceX), you can contextualize it with past patterns.

Part 3 — Crafting an investment thesis (how to turn research into a thesis)

  • Thesis often emerges naturally from research:

    • It’s generally not forced; it becomes articulable once understanding is built.

  • Avoid overly cute, overly specific thesis narratives:

    • Critique example: overly narrow predictions (e.g., an AI memory bottleneck story).
    • References David Deutsch’s idea: the more specific you are, the more likely you are to be wrong.

  • Thesis can be simple and “plain-vanilla”:

    • Example thesis style for Meta:
      • valuation seems attractive
      • earnings growth continues
      • cash flows aren’t wasted
      • AI capex could generate good ad-return

  • Tie thesis to numbers (valuation + returns):

    • Use a DCF approach:
      • forecast revenues/earnings based on your assumptions
      • decide whether to apply multiples or do a DCF explicitly
    • Reverse DCF / implied discount rate method (detailed):
      • Build a DCF with assumptions producing discounted cash flows.
      • In Excel, use What-If analysis so:
        • the sum of discounted cash flows equals the current market cap.
      • Solve for the discount rate that makes the DCF match market cap.
      • Interpret this discount rate as the return currently priced in.
      • Compare attractiveness:
        • high implied return (with reasonable assumptions) can be attractive
        • low implied return or unrealistic assumptions can be unattractive.

  • Relate to multiples as shorthand:

    • Estimate future earnings and compute:
      • PE multiple ≈ market cap / earnings estimate
    • Multiples map back to DCF logic (framed as shorthand).
    • Rule-of-thumb timing:
      • don’t look out more than ~3 years when assessing how many years it takes for a multiple to normalize below current levels.

Part 4 — How to find information (where to get evidence and how to handle missing data)

  • Start by checking whether the info exists:

    • For newer investors: begin with the company’s annual report / 10-K.

  • Use a tiered reading approach:

    • Skim the annual report first to understand disclosures and segmentation.
    • Go deeper.
    • If available, read investor day materials:
      • transcript + presentation
    • Then read:
      • earnings transcripts
      • (He sometimes reads long histories; gives Copart as an example over ~20 years.)

  • Why “AI might miss things”:

    • You can’t rely on AI to guess which obscure disclosures exist (frame problem).
    • Example: Copart disclosed market share only twice in the entire history; he used triangulation rather than expecting standard data to appear.

  • Typical sources and workflow:

    • SEC EDGAR (regulatory documents)
    • Investor relations websites (cleaner versions)
    • Check for inconsistencies:
      • earnings release / earnings presentation may differ from the 10-K—verify across documents.

  • If key data isn’t directly disclosed (example: Adobe enterprise revenue share):

    • Search within 10-K segments.
    • If not found:
      • use earnings transcripts and presentations to infer indirectly (segment changes, carve-outs).

  • Alternative methods when disclosures are missing:

    • Talk to users/customers:
      • B2C: easier via reviews and social channels
      • B2B: harder to access directly
    • Mentions expert call networks:
      • paid expert interviews (e.g., talking to Adobe users/enterprise decision-makers)
    • Creative workarounds if you can’t pay:
      • build a network (students may get responses more easily)
      • use LinkedIn, Twitter/X, Reddit, YouTube
      • search in the company’s local language
      • possibly commission surveys (example: researching “Coupon” in Korean; ~150 Korean consumers)

  • Decision rule when information remains uncertain:

    • Ask:
      • Does missing info kill the thesis?
      • Or can you judge based on:
        • what you’re comfortable assuming
        • management credibility (e.g., belief they won’t pursue destructive short-term growth)
    • If confidence is low:
      • drop the stock
      • return to earlier steps and pick another candidate.

  • Repetition and continual learning:

    • Improve research notes/writing over time (older notes may look “laughably bad”).
    • Learn from both successful and failed businesses to understand what to avoid.

Speakers / sources featured

  • Speaker (primary / narrator): Drew (referred to as “Drew” in subtitles; examples address “Drew” directly)
  • Companies mentioned as examples/case studies:
    • Goldman Sachs
    • Apple
    • Chipotle
    • Meta (Facebook)
    • Copart
    • Amazon
    • Google
    • Nvidia
    • Moody’s
    • S&P Global
    • Fitch
    • eBay
    • Marcato Libre
    • SpaceX
    • Adobe
    • Micron
    • Reality Labs (in the Meta example)
    • TikTok
  • Tools / services mentioned:
    • Yahoo Finance stock screener
    • Joel Greenblatt “Magic Formula”
    • Fiscal AI (show sponsor)
    • SEC EDGAR
  • Referenced author / concept:
    • David Deutsch (on increased specificity raising the chance of being wrong)
  • Substack mentioned:
    • “Mostly Borrowed Ideas”

Original video