How to analyze a company in 30 minutes
A working method, run end to end on Microsoft's own filings: where each number comes from, which ones mislead, and what to make of each one.
Dan Mercer11 min readUpdated
Most of what gets written about a company is written about its share price. The price is the one number on the page that the company did not produce, and it is the one I look at last. Here is the order I actually go in, run start to finish on Microsoft, with every figure traced back to the filing it came from and a stated verdict at the end of each step.
In this post
This takes about half an hour once you know where things are. I have run it on a paid account, because three of the nine steps do not exist without one, and I say which as we reach them.
What follows is fundamental analysis in the plain sense of the phrase: reading what the company itself filed, in the order the numbers depend on each other, and stopping where the evidence stops. The whole method is a 10-K filing, a handful of Form 4s, the 13F holdings other investors filed, and one live price at the very end.
Start with what was filed, not what the price did
The first screen tells me four things and I want all four before I read a word of analysis: what it costs, what the market thinks it earns, who else owns it, and how stale each of those facts is.
That is the screen at the top of this post, and it is worth pressing to read full size. Microsoft closed at $505.41, a market capitalisation of $3.68T. The quote is marked delayed and carries the vendor that supplied it, which matters because it is the only number on this page that did not come from a filing.
Underneath it: 17 of 47 tracked fund managers held the stock as of Q2 2026, and one sentence about insider selling I come back to in step five. The quarter label is not decoration. A 13F can be filed up to 45 days after the quarter it describes, so this is a picture of who owned it in June, read in September.
Open a company pageType any US ticker. The overview, the financials and the insider history are all readable without an account.The read: Not a signal
Nothing to conclude yet, which is the point of reading this screen first.
The only job of this step is to find out how old each number is before any of them persuade you of anything. A page that shows fund holdings without showing you they are a quarter stale is letting you believe something it cannot support.
What the business earned
Now the income statement, and only two lines of it to begin with. Revenue tells me whether the business is growing. Net income tells me what was left after everything the company chose to spend.
Nineteen filed years, 2008 to 2026, drawn to scale from Microsoft's own annual reports. Fiscal 2026 revenue was $331.84B and net income $133.75B. Against fiscal 2025, at $281.72B and $101.83B, that is revenue up 17.8% and profit up 31.4%.
I do not stop here, because net income is an accounting result. It is a set of decisions about when to recognise revenue and how fast to depreciate things, all of them legal and most of them reasonable, and none of them cash. The next statement is the one that cannot be arranged.
The read: Reassuring
Reassuring, and the least informative number on the page.
Revenue up 17.8% and profit up 31.4% is a genuinely good year, and profit growing faster than revenue means margins widened rather than volume alone. It is also the figure a press release leads with, it is an accounting result rather than cash, and on its own it would have told me nothing about what the next step found.
Where the cash actually went
This is the step that changes my mind most often, and it did here. The cash flow statement starts from the same business and asks a blunter question: how much money actually arrived, and where did it go.
Operating cash flow climbing, free cash flow turning down, and the reason sitting between them. The capital expenditure chart is the one to look at. Microsoft generated $182.94B of operating cash flow, more than ever. Free cash flow, which is that number minus what was spent on property and equipment, was $66.99B. In fiscal 2025 it was $71.61B, and in fiscal 2024 it was $74.07B.
Profit rose by a third over two years while the cash left over after building things fell by nearly a tenth.
The gap is capital expenditure: $115.95B, or 63.38% of all the cash the business produced. Three years earlier that figure was $28.11B. It has gone up more than fourfold while revenue went up by about two thirds.
Read a cash flow statementDo this for something you hold. The figure to find is capital expenditure as a share of operating cash flow, and whether free cash flow is following profit or parting from it.The read: Both
Both, and the single most important thing on the page: a record year is converting into less spare cash than it did two years ago.
Spending 63% of your operating cash flow on plant is either building the thing that earns the next decade or overbuilding into a cycle, and no filing can tell you which. What the filing does settle is that the decision is being made at this scale, that the money is genuinely gone, and that anyone quoting the profit line alone is describing a different company.
What the people inside it did
Insider selling is the most misread data in public markets, and most of the misreading happens before anybody looks at the numbers. Executives are paid in stock. Stock vests on a schedule. Shares are withheld to pay the tax on it. All of that generates filings that look exactly like selling and mean nothing at all.
So the only question worth asking is narrower: of the sales genuinely decided by a person, how many were decided in advance?
The same 29 sales, split by what the Form 4 itself says about them. The right-hand column is the honest one. Since April 2023 the Form 4 has carried a box saying whether a trade ran under a pre-arranged 10b5-1 plan. Across 29 open-market sales filed between April 2025 and September 2026, one filing ticked it, covering $43.39M. Four filings explicitly did not, covering $10.11M. Nine filings, covering $119.87M, say nothing either way. In the same period there were two open-market purchases.
See the insider splitCheck it on a company you own. What you are looking for is how much of the selling sits in the unlabelled column before you read anything into it.The read: Not a signal
Not a signal in either direction, and establishing that is the whole value of the step.
The largest block of selling by value, $119.87M across nine filings, carries no statement about whether it was pre-arranged. Unknown is the honest reading, and folding those into either of the other two columns would manufacture a signal out of a missing checkbox. Two open-market purchases in seventeen months is likewise too thin to mean anything.
Who else owns it, and how much it matters to them
Fund managers with over $100M in US equities file a 13F every quarter listing what they hold. It is backward-looking and excludes anything sold before the quarter ended, so the only useful thing to read from it is proportion, not presence. This is the filing behind every institutional ownership percentage quoted for a stock, and what it can and cannot support is a post of its own: how to read a 13F filing, on Berkshire Hathaway’s.
All seventeen managers, and underneath, how many added against how many trimmed each quarter. The complete list is a paid view; a free account sees the largest few. Dodge & Cox holds $4.79B, which is 2.51% of their disclosed book. Pershing Square holds $2.32B, which is 11.89% of theirs. The dollar figures say Dodge & Cox owns twice as much. The percentages say it matters roughly five times as much to Pershing Square.
The strip underneath is the part I find hardest to get anywhere else: managers who added against managers who trimmed, quarter by quarter. Q1 2026 was 8 added against 14 trimmed; Q2 2026 turned round to 10 against 8.
The read: Both
Widely held and lightly held, which are not the same thing and are usually reported as if they were.
Seventeen of forty-seven tracked managers own it and most added over the quarter, which reads as consensus. The third column says the consensus is thin: 2.51% of Dodge & Cox's book against 11.89% of Pershing Square's. A position that size is a view; one at 2% is a place to keep money. Both facts are a quarter old by law.
What the annual report started warning about
Item 1A of a 10-K is where a company writes down what could go wrong. Almost nobody reads it, for a good reason: it is long, defensive and largely identical to last year. Which is exactly what makes the difference between this year and last year worth reading, because that part is not boilerplate.
The July 2026 10-K against the July 2025 one, matched passage by passage rather than by word count. Thirty-three passages are new, seventeen are gone and thirty-six were redrafted, out of ninety-six. A plain word count would have reported nothing useful, because every 10-K rewrites this section cosmetically; what is counted is passages with no match in the previous year.
The read: Worth watching
A prompt rather than an alarm, and unusually specific about where to look.
A third of Item 1A being new language is not by itself evidence of anything; every 10-K redrafts this section. What makes it worth a minute is what the new passages are about, which is competing in AI against hyperscalers, open-source models and the company's own partners. That is the same subject as the capital expenditure in step three, described from the risk side.
What the price is already assuming
Everything so far described what happened. This step is the only one that touches what is expected, and it is why the tool needs an account: it starts from the filed revenue and margin and needs a live price, which is the one figure here that costs money every time it is fetched.
The model run forwards: pick a growth rate, a margin and an exit multiple, and see the price they produce. Every input starts from a filed figure. Paid feature. Run forwards on 10% growth and the current margin, Microsoft earns $215.4B in 2031, and what that is worth depends entirely on the multiple: $591.34 at 20 times, $798.31 at 27, $916.58 at 31. A spread that wide is the honest answer to a five-year question, and any single number quoted without it is hiding the same spread.
Then turn it round and solve for what today’s price needs. A reverse DCF is the same discounted cash flow model with the unknown moved: instead of choosing a growth rate and reading off a price, you fix the price at what the market is charging and read off the growth it requires.
The same model, backwards. Each answer holds every other assumption fixed, so the two do not combine. Holding the margin and the multiple where they are, $505.41 implies revenue growth of 0.4% a year. Holding growth at 10% instead, it implies the net margin falling from 40.3% to 25.5%. Either way, the price is not asking for a continuation of what just happened.
See what an account costsThe free plan does not expire and does not want a card. Steps one to five run on it, and most of them run with no account at all.The read: Both
The surprise of the whole exercise: at this price the market is not asking the company for much growth at all.
Hold the margin at 40.3% and the exit multiple at 27 times, and $505.41 is justified by revenue growing 0.4% a year. The company just grew 17.8%. That gap is the finding, and it is not a recommendation: the answer is extremely sensitive to the 27 times, which is an assumption rather than a fact, and moving it moves everything. What the tool is for is making that assumption visible instead of leaving it implied.
The same thing, written down for you
One thing worth knowing before you spend thirty minutes doing this by hand: the product will write most of it for you. The Company Brief is generated once per company from its own filings and reused for every reader until the company files again.
The memo covers the same ground as the eight steps above it, citing the filings rather than summarising the internet. Paid feature. It is worth reading against the walkthrough rather than instead of it, because it states the five-year shape in one line: revenue compounding at 13.7% a year since fiscal 2022, with the operating margin widening from 42.1% to 46.8%.
Here is what I actually have after all of this.
- Microsoft earned more than ever in fiscal 2026 and kept less of it in cash than in either of the two years before, because it is spending at a scale it has not spent at before.
- The dividend is growing and is taking a rising share of a shrinking free cash flow.
- Insider selling is unremarkable in size and mostly unlabelled, so it supports no reading in either direction.
- The company rewrote a third of its own risk section, mostly about competing in AI.
- At $505.41 the market is pricing in almost none of the growth the company just delivered.
What I do not have, and what nothing above could give me, is a view on whether $505.41 is a good price. That needs a judgement about what the spending earns and what multiple the business deserves in 2031, and neither is a filing. Anyone who tells you the number without telling you their assumption has simply hidden the interesting half of the argument.
The read: Not a signal
A description of a business, not a verdict on it, and the difference is the whole discipline.
Half an hour of this leaves a paragraph rather than an answer, and that is the correct output. A description survives being wrong: if the spending turns out to be the right call, nobody has been made a fool of, because nobody claimed it would not be.
Questions
- What is fundamental analysis?
- Fundamental analysis is working out what a business is worth from what it reports rather than from what its share price is doing: revenue and profit from the income statement, cash from the cash flow statement, what management warns about in the risk factors, and who owns it. The inputs are the company's own filings, which for a US company means the 10-K, the 10-Q and the Form 4s its officers file.
- Where do these numbers come from?
- The statements, insider transactions and ownership figures are extracted from filings Microsoft and its investors submitted to the SEC: the 10-K, Form 4 and 13F respectively. The share price is a delayed quote from a market data vendor, and it is the only figure here that is not from a filing.
- Do I need an account to follow along?
- For most of it, no. Steps one to five use pages readable without signing in, including the full financial statements and the insider history. The complete holder list, the generated Company Brief and the valuation tool need an account, because the last of those needs a live price.
- Why start with cash flow instead of earnings?
- Net income depends on decisions about when revenue is recognised and how quickly assets are depreciated. Cash either arrived or it did not. Microsoft is a clean example of why it matters: profit rose 31.4% in fiscal 2026 while free cash flow fell for the second year running.
- Does insider selling mean something is wrong?
- Usually not. Executives are paid in stock, that stock vests on a schedule, and shares are withheld to pay tax on it, all of which produces filings that look like selling. The question worth asking is whether a sale was pre-arranged under a 10b5-1 plan, and the Form 4 has said so since April 2023 when the filer ticks the box.
- What does a reverse DCF actually tell you?
- It runs the model backwards: instead of guessing growth and getting a price, it takes today's price and solves for the growth that would justify it. That makes the market's implied assumption explicit rather than leaving you to argue with a number you cannot see. It is also highly sensitive to the exit multiple you hold fixed, so it frames the question rather than answering it.
- Is any of this investment advice?
- No. Everything here is a description of what a company filed, and the post deliberately stops short of a view on whether the share price is right.
Sources
Every figure above was read off the product, and comes from the documents below. Where a figure is not in a filing, the post says so rather than estimating one.
Nothing here is investment advice, and the post makes no forecast about any company or fund it names.
Written by
Dan Mercer
Founder and writer, Investailor
- Builds Investailor, a research desk that reads SEC filings, 13F holdings reports, Form 4 insider transactions and ETF holdings files, and traces every figure it shows back to the document it came from.
- Writes each post by working through the real filing in public: every number in a walkthrough is read off a page the reader can open, and the primary documents are linked at the foot of the post.
- Publishes the method rather than the conclusion. These posts say what a filing records, what it leaves out, and where the evidence stops, and they carry no price targets, ratings or forecasts.
Do this for a company you actually hold
Be told the next time it files
Put any company on a watchlist and a Form 4 worth a row, a new 10-K or 10-Q, or a tracked manager's changed position lands in the bell and a daily email, with vesting and automatic tax sales already struck out. Free.
See who bought, not who was handed shares
The same split this post ran on Microsoft's insiders, on any company: every Form 4 sorted by whether the filing itself says the sale was planned months ago or decided that week.
Every filing, not the last handfulMeasure
Ask what growth the price is already betting on
The reverse DCF from step eight, on whatever you're looking at: hold today's price fixed and solve for the growth it requires, instead of guessing a growth rate and hoping the price agrees.
The reverse DCFMeasure
Watch the same stakes the professionals are watching
Every tracked position priced at the quarter it was filed and diffed against the one before it: how big a bet it is for them, and by how much it moved, on all seventeen managers rather than the top few.
Every position, sizedMeasure
Free, no card, and it does not expire.
Read next
How much do your ETFs overlap? VOO, VTI and QQQ
Seven pairs of funds people actually hold, measured against each other from the holdings each issuer publishes, with the two numbers that answer different questions.
9 min read
PayPal: 42 insider sales, one purchase, one checkbox
One company, three sets of filings, read against each other: what PayPal's Form 4s, its statements and its 13F holders each say, and why they do not agree.
7 min read
How to read a 13F filing: what Buffett and the funds own
What a 13F is, what it leaves out and what it can never tell you, read line by line on Berkshire Hathaway's filing for Q2 2026.
8 min read