How to Analyze a Stock Before Buying: The Complete Process
Quick answer: The sequence does not change from stock to stock. What changes is the answers. This page is the eight-step checklist filled in for one company a beginner could actually sit with for an evening — Amazon (AMZN). It is a notebook, not a recommendation.
If you want the questions and the sources, use How to Research a Stock. If you want empty boxes, use the checklist. If you have not bought the default index yet, start at How to Start Investing and ignore this page.
Beginner kit on this example: sections 1–3 and 7 are enough to decide whether AMZN is even a candidate. 4–6 and 8 are how it gets serious.
Tonight's header: AMZN · notes from one sitting · not a buy ticket.
1. Business
The question: What does it sell, and who pays?
Notes:
- Three businesses sharing a ticker: the store, AWS, and ads — plus Prime sitting on top of the store.
- Shoppers pay the store. Companies pay AWS. Advertisers pay for the traffic the store already has.
- The store is still most of the revenue. The profit story is increasingly AWS and ads.
- Paid how: take-rate and subscriptions on retail/Prime, usage bills on AWS, ads sold against shopper intent.
Done when test: I can explain that in two minutes with the ticker covered. Pass.
Still open: I cannot recite, off the page, what share of last year's profit was ads vs AWS. That is a 10-K Item 1 / segment note, not a headline.
FW 01
2. Both sides
The question: Honest case for, honest case against, same length.
Bull, as specific as I can make it tonight:
- AWS is a real software-like business sitting inside a retailer. If cloud keeps growing and they monetize AI on that base, the mix of the whole company gets better.
- Ads are attached to purchase intent, not to a random feed. That is a different business than "more traffic."
- Prime and the third-party marketplace are hard to recreate in one product launch.
Bear, forced to equal length:
- Cloud growth can slow. Azure does not have to "win" for AWS to stop being the story the price is priced for.
- The store is huge, low-margin, and already everywhere the easy customers live.
- Regulators have the store and the marketplace in their sights. A forced change there is not a fantasy risk.
- Capex on data centers only pays if customers rent the capacity. Spending is not the same thing as earning.
Done when: the bear case uses company facts. Pass enough to keep going. Not a finished debate.
FW 02
3. Worth
The question: What is this price already assuming?
Notes:
- Tonight's price is a claim that AWS stays a growth engine and that ads keep compounding on the store.
- If cloud growth cools and ads flatten, I do not have a private "cheap" number in my head. I have a walk-away rule: I will not add until I can finish the sentence "this only works if ___."
- For the arithmetic on implied growth, use the What's Priced In calculator. I did not pretend to finish a DCF in one sitting.
Done when: one sentence about what the price requires. Draft: "This price needs AWS to keep growing at a healthy clip and ads not to stall."
Still open: a written cheap / fair / expensive range I would actually trade against. That is the next sitting, or a Framework 03 run — not something I should invent to look finished.
FW 03
4. Earnings real
The question: Did the cash show up?
Notes:
- Retail profit and AWS profit are not the same quality of dollar. I have to look at them separately or the income statement will lie to me by averaging.
- One sitting is enough to open the cash-flow statement next to net income and see whether cash from operations tracks the story management is telling.
- Watch for "adjustments" around stock comp, one-time items, and capex so large it swallows the cash the earnings press release celebrated.
Still open: two full years of income vs cash, written side by side. I have not done that homework in this notebook. That is the gate before sizing up.
FW 04
5. People
The question: Five-year trust?
Notes:
- This is no longer a founder-operator story in the old sense. The question is whether the current team allocates capital like owners now that the easy growth is behind them.
- I care more about how they talk when AWS slows than how they talk when it beats.
- Form 4s: I have not pulled them tonight. Open-market buying would mean more than a prepared letter.
Still open: compensation metrics and a year of Form 4s. "Famous company" is not a management grade.
FW 12 / 08
6. Risks
The question: What actually kills the thesis — not the 40-page Item 1A list?
The two I will not live with:
- AWS growth going cold for several quarters while spending stays as if nothing changed.
- A regulatory outcome that breaks the store / marketplace economics I am counting on.
Everything else (a bad retail quarter, a noisy headline, a competitor keynote) is noise unless it feeds one of those two.
FW 13
7. Thesis and invalidator
The question: Why own it, and what proves this notebook wrong?
Thesis draft:
Own a piece of a commerce + cloud + ads bundle if — and only if — AWS remains a healthy growth engine and ads keep earning on the store's traffic. Horizon is years, not a week.
Invalidator draft:
AWS growth goes negative, or stays weak while capex does not come down, and ads slow with it. If that sentence becomes true, I do not "wait for the brand." I re-underwrite or sell.
That is written before any add. A research journal is fine if I want it on a cleaner page.
FW 05
8. Fit
The question: If this goes to zero, does the rest of the account still work?
Notes:
- If I already own a total-market or S&P 500 fund, I already own AMZN. A separate lot is a concentration decision, not a first introduction to the company.
- Max size gets written before the order. A few percent of invested assets is a beginner cap. Half the account is not conviction. It is a hole.
- Position Size calculator for the arithmetic. The number itself is a judgment.
Done when: max size is on the page. Tonight's rule: no add until sections 3, 4, and 5 are less empty than they are right now.
FW 06
There is no best first stock. AMZN is just a company most people can describe. If sections 1–3 and 7 were blank, the honest move would be to own the index and close the laptop.
After the notebook
Empty boxes: Stock Research Checklist.
The questions and 10-K map: How to Research a Stock.
Which leftover question maps to which other framework: Which Framework, When?
What a full tool run on this same ticker looks like — not tonight's notes: AMZN analysis.
FAQ
What's the difference between researching and analyzing a stock?
Research gathers the facts — filings, numbers, news. Analysis weighs them against the current price to form a view. Both belong in the same written process.
Do I need a spreadsheet to analyze a stock?
No. A notebook and the actual 10-K get you further than a spreadsheet built on assumed numbers.
How long should stock analysis take?
Long enough to state the bull case, the bear case, and what the price already assumes — often an hour or two per company, not days.