How to Evaluate a Stock: 25 Questions With a Threshold

How to evaluate a stock comes down to 25 questions, and each one is worth exactly as much as the number and the threshold standing behind it. "Does the company have pricing power" is a question. "Did the last 5% price increase go through without a drop in volume, and which line of the filing shows it" is a criterion. The first can be talked around. The second cannot.

Without thresholds, 25 checks are 25 opinions arranged in a column.

Why a low P/E is not enough#

The P/E ratio is the fastest filter anyone has invented, and that is not a criticism. It tells you what you pay for a dollar of last year's profit, it takes three seconds, and it sets aside ninety companies in a hundred before you open a document. Starting there saves you a week of reading.

The trouble is that the P/E describes the price, not the denominator.

Last year's profit is not a promise about the profit three years out. A company can trade at six times earnings because its largest customer just gave notice, and the market already knows. The same six at a company that renewed that contract for five years means the opposite. Same ratio, different business, and the difference never shows up in a screen, only in the note on major customers.

The 25 questions below are an audit of the denominator. Not to arrive at a company's "true value", because no list arrives at that. To know whether the profit you are paying six times for comes from reasons that survive three more years, or from reasons that are running out now.

What turns a question into a criterion#

A bank credit committee does not vote on whether a business looks solid. It has a sheet, the sheet has a criterion, and the criterion has a line: net debt to EBITDA under three, interest cover above two. That line is arbitrary and everyone around the table knows it. It still does what no discussion does: it writes the decision down before anyone knows how it turned out.

Each of the 25 checks below gets three parts. First a location: the segment note, the cash flow statement, management's commentary, the buyback table. Then a threshold, the value at which the answer flips to no. Then the trap, which is how a number manages to be technically true and misleading at once.

Say a company reports revenue up 12% year on year. Without a split between price and volume, that says nothing about pricing power: 12% from price increases against shrinking volume and 12% from new customers at an unchanged price list are two businesses wearing one number. I put the threshold on volume. If volume falls more than a few percent after a price rise, that is not pricing power, it is a customer loss on a delay.

Where does the threshold itself come from? From the company's own history, from the spread across its industry, and from the level at which somebody already lost money. None of the three is hard, and none needs to be. Writing it down before you read stops you moving the line two points after the fact so the company you have grown fond of squeaks through. That is how a checklist usually stops working, and it happens to everyone.

If you want the order of operations rather than the list of criteria, the same process from the other side is in the piece on how to analyze a company before you invest.

The business and its capital, ten items#

The first ten ask how the company physically makes money and what funds it. All of it comes out of documents alone, without talking to anybody, which is why it goes first. If you are still assembling the toolkit, the basics of investing in stocks sit one step before this list.

1. Stage of development. In the cash flow statement, count the last eight quarters with positive operating cash flow. Fewer than six means someone other than the customer funds the growth.
2. Intangibles. Ask what is left if the twenty best people walk out. A patent, a licence or a database stays. "Culture" leaves with them.
3. Geographic spread. In the segment note, find the largest country's share. Above 60% is regulatory risk, not currency risk.
4. Product spread. Find the revenue of the best-selling line. Above half and you own one product with accessories, whatever the deck says.
5. Research spending. Set R&D against revenue over five years. A falling share on rising sales is harvesting, not thrift.
6. Brand. Compare gross margin with the nearest competitor selling the same thing. A gap under five percentage points means the brand lives in the deck, not the price list.
7. Pricing power. Look in management's commentary for the split of growth into price and volume. The absence of that split is itself an answer.
8. Direction of the industry. Compare the growth of the company's market with the growth of the economy. Slower than the economy and the rest of this list matters less.
9. Customer concentration. In the segment note, count customers above 10% of revenue. One is a question, two is a risk, three is the business model.
10. Capital structure. Divide net debt by a year of EBITDA. Above three in a cyclical business means the bank, not management, decides when the next recession starts here.

Items nine and ten can kill a company in a quarter. The other eight take years.

The product and the moat, six items#

The second six ask about something that appears in no line of any statement: why the customer stays. This is where it is easiest to fool yourself, because a decent answer can be written for almost any company if you define the market broadly enough. The mechanism and its limits are in the piece on what an economic moat actually is.

1. Essential or optional. Imagine switching the product off at the customer tomorrow. If they notice within a quarter rather than within an hour, it is optional.
2. Cost of switching. Count months of migration and people to retrain. Under a quarter is not a switching cost, it is inconvenience.
3. Network effects. Check whether the hundredth user makes the product more valuable to the first. If not, you have ordinary scale, which works on costs and does not defend price.
4. Recurring revenue. Find the share of revenue that renews without a fresh customer decision. Seventy percent and thirty percent are two different businesses at identical sales.
5. Scaling without capex. Divide capital spending by the increase in revenue over three years. The more assets a dollar of new sales needs, the less reaches the owner.
6. A difference the customer can name. Say in one sentence why the customer does not pick the competitor. Three adjectives do not count as a sentence.

Four of those six collapse into one question: what does leaving cost. If nothing, there is no moat, there is a price list.

The industry and management, nine items#

The last nine measure people, and people have no line in the balance sheet. They do leave a trail where they are required to. A US filer whose risk factor section runs past fifteen pages has to put a summary of no more than two pages at the front - and what the company picks for those two pages says more than the full fifteen. It reads faster than an investor deck, and what to take from it is in three risks the company had to disclose.

1. Fragmented competition. Count the players holding three quarters of the market. Three means the weakest of them sets the price the moment cash gets tight.
2. Regulation. Work out whose cost the new rules raise: the company's, or that of anyone wanting into its market. Same statute, opposite conclusions.
3. Answering the hard question. Find the question about falling margins in the transcript. If no number appears in the answer, you have your answer.
4. Consistency of strategy. Set the strategy from two years ago against today's. A new slogan where an old target used to be is usually an abandoned target, not a new idea.
5. Capital allocation. Check the buyback table for the average price management paid for its own business. Buying at the peaks and going quiet at the lows is a reading on discipline, not bad luck.
6. Pay. Read what bonuses are attached to: revenue, earnings per share, the share price, or nothing measurable.
7. Insider ownership. Separate shares bought with their own money from shares granted under an incentive plan. Only the first cost anything.
8. Guidance against delivery. Line up the last eight guidance numbers with what the company actually delivered. A gap that widens three quarters running is a pattern, not a run of bad luck.
9. Behaviour in a downturn. Check how far revenue fell in the last trough and how many quarters it took to recover. Back in a year and back in three years are two different businesses.

Three checks on this list - the ninth from the first ten, the fifth and eighth from these nine - carry a hard number that almost nobody counts. The next three chapters are about those. The rest of the management question reduces to lining up promise, action and evidence.

How many customers hold up this revenue#

A customer worth 5% of revenue negotiates the price. A customer worth 20% sets it.

That is not a metaphor, it is the balance of power. A buyer above twenty percent knows that walking away wrecks two years of planning, so at every renewal they take a slice of margin and the company hands it over. In the table the margin still looks like the company's. It belongs to the buyer.

In its report for the quarter ended 27 July 2025, NVIDIA wrote: "sales to one direct customer, Customer A, represented 23% of total revenue; and sales to a second direct customer, Customer B, represented 16% of total revenue" (segment note, Form 10-Q). Which means two buyers accounted for 39% of the quarter's revenue. That is not a charge against the company, it is a number easy to find and easier to skip.

I read the threshold like this: one customer above 10% is a question, two is a risk, three is the business model. Outside the US the same information sits in the segment and revenue note, under major customers.

The trap hides in the word "direct". A note like that covers who places the order, not who ultimately uses the product. Buyers of record are sometimes integrators or distributors, so the real dependence can sit one level further down - or be smaller than it looks, if a hundred end users stand behind one purchase order. What the note names is what the company knows about its invoices. The rest is your work.

What do you do with the answer once you have it? Come back in a year, to the same note. Concentration alone settles nothing, because a company with one enormous buyer can earn decently for years if that buyer has nowhere to go. Direction settles a great deal: a largest customer whose share rises three years running means the company wins fewer deals, only bigger ones. In a deck that looks exactly like success, and it ends differently.

The price management pays for its own stock#

A buyback is the one investment decision management makes that you can judge on execution price, because the price is printed in the document.

A US filer buying its own stock has to show those purchases in every quarterly and annual report: shares month by month and the average price paid per share. Nothing else in a set of accounts works like that. There management says what it thinks the company is worth and pays for the opinion in cash rather than in a slide.

In the same report, NVIDIA disclosed that in the first half of fiscal 2026 it had repurchased 193 million shares for $24.2 billion, which works out to roughly $125 a share. That figure is not a verdict on its own. It becomes one next to what the company earned per share then and earns now, because only then do you see the valuation at which management judged its own business worth owning.

The reading rule here is comparative, not absolute. Management that buys hard at the highest multiples in the company's history and stops once the price halves has told you what its discipline is worth. The reverse order is rare and worth remembering.

The trap: dollars spent are not the same as shares removed. If the share count sits still despite billions running through the buyback, the programme is not returning capital, it is filling in dilution from incentive plans.

You check that in one line - shares outstanding at period end, eight quarters back.

A dividend does not pass this test, because it carries no price. A company paying two percent a year does the same thing whether its shares cost thirty times earnings or ten. A buyback is a transaction, so it has a price, a moment, and an alternative management walked away from. That is why it is the one point in the whole capital allocation section where you can set a hard threshold instead of an opinion about the sensible use of cash.

Eight quarters of guidance against delivery#

Guidance is the only number in the reporting pack that management picks itself. Everything else is the result of events. That statement, set against delivery, measures the people running the company rather than the company.

You will find it in the earnings release and the last minutes of the call, usually as a range for next quarter's revenue and margin. Write down the midpoint, not the bottom, because the bottom is there to be beaten. Come back three months later and write the actual next to it.

Say a company looks like this:

Each of those quarters can be explained on its own, and management will explain each one. Four side by side explain themselves: the promise rises, delivery falls, the gap widens every quarter. That is no longer a run of bad luck, it is a way of conducting a conversation with the market.

Eight quarters, not four, because an annual cycle straightens out two weak quarters and leaves you looking at seasonality instead of a pattern. Eight covers two full years, so at least one change in conditions management had to answer for.

The trap runs the other way round from what you would expect. Management that reliably promises little and delivers more is also telling you something, and nobody writes that down under risks. I do not know where the line between prudence and expectations management sits, and I doubt it can be set on a number. I do know that the direction of the gap over eight quarters is visible to the naked eye, and that a single quarter says nothing.

Where this checklist fails#

The thresholds on this list are industry-specific and there is no way around it. Net debt to EBITDA of three is an alarm at a software company and normal at a water utility, because one has nothing to pledge and the other has pipes and a tariff. One threshold for every industry is a list for one industry.

The second weakness is worse, because it looks like a strength. A score of 22 out of 25 sounds like a measurement. It is a sum of things that do not add up: concentration and "strong brand" each get a point, though the first can be counted and the second is an opinion. A total always looks more precise than its parts.

The third problem is out of reach. No threshold measures whether management is honest, and honesty decides the quarter in which every number has gone stale, meaning the quarter in which something goes wrong.

Does that make the list useless? No, it means the list has a range. It does one thing well: it makes you write down what you were looking for before you saw what you found. Everything past that you supply yourself, and a list promising more is selling you a sense of control at the price of twenty-five ticked boxes.

So the score is not the answer. The answer is the checks the company did not pass, and whether any of them dismantles the reason you were looking at it in the first place.

How to evaluate a stock in an hour#

I think three of the 25 checks genuinely decide: customer concentration, the price paid in buybacks, and eight quarters of guidance against delivery. Not because the other twenty-two do not matter. Because those three carry a number nobody can talk around, and the rest can be defended with a good deck, and usually is.

All three fit into an hour, off one quarterly report and one annual, and every figure has been public since the day it was filed.

At Taufolio this list is the skeleton of the Full report: the case for a company broken into 25 points, each pointing back at the document it came from. Out of the finished report comes an Investment thesis, a record of why you hold a company, split into conditions you can check against the next results. The gist of the report sits at the top if you do not have forty minutes. On the Free plan you get 100 credits a month and a Full report costs around 100 credits, so the first company costs nothing.

Read it from the points that were deducted. A score of 22 out of 25 does not say the company is good; it says where to start arguing with your own thesis, and that is what the list is doing inside the product.