How to Analyze a Stock Like Berkshire — the 2026 Field Guide

Contents

1. Why the order matters more than the tools
2. Step 1 — the Berkshire-style quality checklist
3. Step 2 — the valuation ceiling
4. Step 3 — writing thesis breaks that actually work
5. Step 4 — timing signals, last
6. Worked example: Apple (AAPL), snapshot 2026-08-28
7. Seven mistakes that break the workflow
8. How this scales to 753 tickers
9. FAQ

1. Why the order matters more than the tools

Ask most retail investors how they analyze a stock and the answer is some version of "I read about it, checked the chart, and it felt right." The missing piece is not information — 2026 has infinite information — it's sequence. Analysis done out of order quietly lets the last step contaminate the first: you see a beautiful chart, then go hunting in the fundamentals for reasons to justify it. Psychologists call it confirmation bias; in a portfolio it just calls itself conviction.

The Berkshire tradition — and our entire product — is built on a fixed sequence:

1. Business quality. Would this company still be excellent in ten years if the market closed tomorrow?
2. Valuation ceiling. What price pays too much for that excellence?
3. Thesis breaks. What measurable events should change my mind?
4. Timing. Only now: is there a structural edge to entering today?

Each step is a filter. A stock that fails step one never reaches step two. By the time you reach a timing signal, you've already decided — on paper, in cold blood — what kind of business this is, what you'll pay, and what would make you leave. The signal just schedules the decision you already made.

2. Step 1 — the Berkshire-style quality checklist

The checklist is deliberately mechanical. Where a human reads "consistent earnings" and sees what they want to see, a rule-based score asks the same questions of every company, in every market, in the same order:

What it measures

• Durable profitability — are earnings positive and stable across the scanned window, or lumpy and dependent on one-offs?
• Balance-sheet strength — is debt funded by real cash generation, or by optimism?
• Cash conversion — do reported profits actually arrive as cash?
• Circuit-breakers — any pattern historically associated with value traps (serial dilution, shrinking margins with rising revenue, receivables outrunning sales)?

What the score is — and is not

The output is a 0-100 grade. On our latest full scan cycle (snapshot dates stamped on every page — the AAPL snapshot, for example, is 2026-08-28 via yahoo), 546 US names carry scores, and the system scans across Technology (86 names), Industrials (81), Financials (78), Healthcare (59), Consumer Discretionary (54), and Consumer Staples (34) among others.

What the score is not: a verdict on price. A 95-quality compounder can be a terrible buy at the wrong price, and a 45-quality cyclical can be a fine one at a deep discount. That separation is the whole point of running quality before valuation — never in the same breath.

3. Step 2 — the valuation ceiling

Working value analysis doesn't ask "will Apple grow?" It asks "what does today's price already assume?" The ceiling converts your assumptions into one defensible number:

The two-stage residual-income model, in plain language

1. Stage one projects the spread between return on equity and the cost of equity for a fixed horizon. If the business earns more than its equity costs, value is being added; the model accumulates that spread.
2. Stage two prices everything after the horizon with a conservative fade — the assumption that competitive advantage erodes eventually, because it usually does.

The output is a ceiling: the highest price your thesis can pay without underpaying for risk. Below the ceiling you're paying a discount. Above it, you're paying for hope. Neither is automatically wrong — the discipline is knowing which one you're doing, deliberately, every time.

You can run the model yourself, free, at the Berkshire intrinsic value calculator. Every input has a documented rationale, and the point is not the tool's answer — it's forcing your assumptions out of your head and onto the table where you can argue with them.

4. Step 3 — writing thesis breaks that actually work

A thesis break is a measurable event that, if it happens, should change your mind. Not a feeling ("the story seems weaker"), not a price level alone ("if it drops 15%"), but an event in the business:

• Revenue from the top three customers contracts for two consecutive quarters
• Gross margin compresses past X% for Y quarters with no offsetting volume gain
• Buybacks funded with debt instead of free cash flow
• Key executive departures alongside a guidance cut — together, not alone

Sector context matters, which is why every decision brief on this site includes a sector-specific break template — four candidate conditions tailored to the industry, that you edit into your own words. The point is not our list. The point is that conditions written while you own nothing are always colder than conditions written while you're down 20%. Write them first. Date them. Re-read them monthly.

5. Step 4 — timing signals, last

Only after steps 1-3 pass does timing enter. Our screener emits three states: LONG, SHORT, and WAIT — and WAIT is a feature, not a failure. In the 2026-08-28 cycle, of the scored US names the engine was in LONG on most and WAIT on the rest; the honest state for a market with no edge is no opinion.

Three fields accompany every signal, and each has a job:

• Age (bars) — how long the signal has run. Old signals owe you a fresh look: the reason they fired may have expired.
• uPnL — the unrealized result of the open signal, from its entry price. It measures the engine's history, not your future.
• Total return — the same series, cumulative. Useful for auditing us later, which is exactly why we publish it dated.

Nothing here is a promise. The audit trail exists so you can check, month by month, whether the discipline held up — including when it didn't.

6. Worked example: Apple (AAPL), snapshot 2026-08-28

Walk it through the workflow:

1. Quality: 100/100 clears the checklist. Fine — but this tells us nothing about price.
2. Ceiling: at $314.58, run the calculator with your own growth and fade assumptions. Whatever ceiling you compute, write down the number and the date.
3. Breaks: for Technology, candidates include top-customer revenue contracting two quarters, sustained gross-margin compression, platform commoditization, and debt-funded buybacks. Pick two, in your words.
4. Timing: LONG, 291 bars old — old enough that the honest move is a fresh brief before relying on the signal's original reason.

Total elapsed time for a first pass: 15-20 minutes. The full AAPL walkthrough lives at the AAPL decision brief (English) and the Thai brief.

7. Seven mistakes that break the workflow

1. Starting at the chart. The chart is step four. Starting there means every later "analysis" is a justification.
2. Treating the quality score as a buy rating. It grades the business, not the deal.
3. Arguing with your own ceiling. If the price is above the ceiling you computed, the honest responses are: raise the price you'd pay knowingly, or walk. Not quietly lower the fade rate.
4. Writing break-conditions after entry. Drawdown-brain writes worse conditions than cold-brain. Always.
5. Confusing uPnL with expected return. The engine's history is evidence about the engine, not about your trade.
6. Ignoring signal age. A 291-bar-old LONG is closer to a rumor than a signal.
7. Skipping the date check. Every number on this site carries a snapshot date. If you can't find one, the page wasn't built by us.

8. How this scales to 753 tickers

Doing this by hand across five markets is impossible, which is why it used to be an institution-only workflow. The full book on Balance Labs covers 753 tickers across US, Hong Kong, China A-shares, Korea, and India, each with a dated decision brief, and 27,000+ pairwise comparisons generated from the same snapshots. The checklist, the ceiling, and the break templates are identical everywhere — so your notes on one stock are comparable with your notes on another, and with ours.

Start with three names you already own. Run the four steps. If the workflow changes your mind about one of them, it has already paid for the afternoon.

FAQ