One real stock, the whole toolkit
This volume applies everything in the series to one real research page: Microsoft (MSFT) as it looked in a brokerage app in early July 2026. The stock was at $390, down about 20% in a year, after its worst month in decades.
Two warnings before we start. First, this is a snapshot: every number here has changed by the time you read this, which is itself a lesson. Second, this is a worked example of how to think, not a recommendation to buy or sell anything.
Run the five questions
Volume 4 gave us five questions. Here they are, applied.
Is growth real? Yes.
Revenue up 18%, earnings per share up 23% in a year. The cloud division grew about 40% and the AI business more than doubled. This is not a shrinking company.
Are margins holding? Yes.
A 38% profit margin is elite. Gross margin slipped a little under the weight of new data centers, worth watching, not alarming yet.
Share count stable? Yes.
EPS grew at the same pace as total profit, which means almost no dilution. Buybacks roughly offset new shares.
Beat and raise? Beating, yes.
Every quarter on the chart, the dark bar (actual) topped the light bar (estimate). The classy rhythm from Volume 4’s earnings call section.
Cash confirming profit? ⚠ Here it is.
The one question a broker’s summary screens almost never answer, and exactly where this story lives. Treat it as a mandatory check on every stock: open the cash flow statement yourself. Step Three shows why.
Four green lights and one flashing amber. On business quality alone, this is a strong company. So why is the stock down 20%?
The stock fell while earnings grew
Here is the puzzle in one line: earnings rose 23% while the stock fell 20%. Both are true. The bridge between them is the multiple: how many dollars the crowd pays per dollar of earnings.
Price is always P/E times earnings. When the P/E shrinks, the stock can fall even as the business grows. That shrinking is called multiple compression or a de-rating, and it is what happened here: the market went from paying about 34 times earnings to about 23 times.
Price = earnings × multiple. Move both and watch them fight.
At the new price, the P/E of 23 sits almost exactly on the EPS growth of 23%: a PEG ratio of about 1.0, which the Volume 4 napkin rule calls roughly fair. The same company was expensive a year ago and is ordinary-priced now, and the business barely changed. Prices are opinions; earnings are facts.
Where the bear case actually lives
A typical broker research tab shows income, valuation, dividends, and ownership, everything except the cash flow statement. In this case that missing statement was precisely where the market’s worry sat. The rule to take away: whatever a summary page omits, you must pull up yourself, in the app’s financials tab or in the company’s filings, before judging the stock.
Microsoft reported about $32B of quarterly profit but only about $16B of free cash flow, because it is pouring roughly $190B a year into AI data centers (capex). Every dollar of that leaves as cash today and only returns as revenue over years. Volume 4 called “profit without cash” a red flag; here it is not hidden or fraudulent, it is a deliberate, publicly announced construction project. The question the whole market is arguing about: does that construction pay off?
The cash machine vs the construction site (quarterly, $ billions)
One more fact that only appears if you dig past the summary numbers, another mandatory check: a large share of the company’s contracted future revenue was tied to a single AI partner. That is concentration risk: when one customer matters that much, their problems become your problems.
Undervalued by 43% and Bearish 1.2, on the same screen
This stock’s research page carried a valuation model calling it 43% undervalued, directly above a quant rating of Bearish 1.2 out of 10, two professional opinions, opposite verdicts, one screen. You will meet this exact collision on almost any stock you research. Meanwhile the average human analyst target sat far above the price, with dozens of buy ratings and zero sells. Who is right?
Valuation models
Estimate what the business is worth by adding up its parts and projecting cash flows (a sum-of-the-parts approach). They answer: what should this be worth eventually?
Quant ratings
Scores like StarMine lean heavily on price momentum and whether analysts are cutting estimates. After a terrible month, these are almost always bearish. They answer: how does this look right now?
Analyst price targets
Human forecasts, usually 12 months out, usually optimistic, and usually revised after the move happens. Useful as a mood gauge, not a promise.
The reconciliation
They are not contradicting each other; they are answering different questions on different clocks. Long-clock measures said cheap. Short-clock measures said falling. Both were accurate.
The biases this exact case triggers
Anchoring
“It was $533, now $390, so it is cheap.” The old high is an anchor, not a valuation. Cheap is about earnings and cash, never about the price it used to be.
The falling knife
Down 11% in a month with heavy volume means forced sellers may not be done. Levels and stabilization matter before any short-term entry.
Confirmation shopping
Bulls will read only the $558 estimate; bears only the 1.2 rating. The discipline is writing both cases before deciding, in full sentences.
Event risk on the calendar
The next earnings report was about three weeks away. Known dates gap prices; position sizing and timing must respect them.
Not a verdict: the two honest cases
A disciplined analyst ends not with a feeling but with two written paragraphs and a watchlist. Here are all three for this case.
What a bull must believe
The $190B of data centers converts into profitable AI revenue over 3 to 5 years, cloud growth holds near 40%, and today’s compressed multiple re-expands once free cash flow turns back up. Then a strong business was bought at its cheapest multiple in years.
What a bear must believe
The spending outruns the payoff, margins keep grinding down, the single-partner dependence turns costly, and cheap AI tools erode the office-software cash cow. Then the de-rating was the market correctly repricing a riskier decade.
What to watch, concretely
Cloud growth versus the guided ~40%. The quarter free cash flow stops shrinking. Capex guidance direction. The partner concentration easing. These are checkable facts, not vibes, and each earnings report grades them.
Concepts the earlier volumes had not needed yet
- Multiple compression / de-rating
- The P/E shrinking, so a stock falls even while earnings grow. Its opposite, expansion, is a re-rating.
- PEG ratio
- P/E divided by growth rate. Near 1 is the classic “price matches growth” napkin mark.
- Trailing vs forward P/E
- Trailing uses the last 12 months of real earnings (TTM); forward uses next year’s estimates. Forward looks cheaper when growth is expected.
- Free cash flow (FCF)
- Operating cash minus capex. The cash a business truly throws off after maintaining and building itself.
- Capex
- Capital expenditures: money spent on factories, servers, data centers. Cash out today, revenue over years.
- Price / book
- Price versus the company’s accounting net worth. More meaningful for banks and industrials than for software.
- Institutional ownership
- The share held by funds and institutions (71% here). High means the pros are present, and that their selling moves the price.
- Analyst price target / consensus
- The average of published 12-month forecasts. A mood gauge that follows prices as much as it leads them.
- Quant rating
- A machine score (like StarMine) built mostly from momentum and estimate revisions. Reads the recent tape, not the decade.
- Social sentiment score
- A measure of chatter mood on social media. Neutral here, meaning the crowd’s noise carried no signal.
- Sum of the parts
- Valuing each division separately and adding them, the method behind that $558 estimate and its division breakdown.
- Concentration risk
- Too much revenue, backlog, or exposure tied to one customer, product, or partner.
- Delayed vs real-time quotes
- Free app quotes often lag by minutes. Enable real-time before trading, or you are aiming at where the price was.
- Drawdown from peak
- The distance below the all-time or 52-week high. Descriptive of pain, silent about value.
Do this with any stock
The whole method: run the five questions, translate the price into a multiple, find the statement the summary hides (usually cash flow), name the disagreement between sources, write both cases, list the checkable facts, size accordingly. Twenty minutes, once you have done it a few times.