10-K vs 10-Q: How to Read a Filing Nobody Wrote for You

A primary source in stock research is a document the company itself filed or said out loud: the 10-K or 10-Q in the United States, the periodic and current reports filed through ESPI in Warsaw, the earnings-call transcript. It is not better than a take on it because it is official. It is better because it carries a date, a fixed place in a document and a signature, and every intermediary on the way to you drops at least one of the three.

The advice to read the original is old and nobody argues with it, usually for a reason about honesty: a regulated filing does not lie, because lying in one is expensive, while a post online costs nothing. That is a large piece of the truth. The annual report is signed by management, audited by a firm that carries liability, and the risk section is drafted by a lawyer who can picture the lawsuit.

But a filing is not more truthful than a take on it. It is checkable, which is a different virtue entirely.

What a primary source actually is#

A primary source existed before anyone summarised it. In the American set that means the forms the company files with the regulator: the annual 10-K, the quarterly 10-Q, the ad-hoc 8-K and the proxy statement, all of them free on EDGAR the moment they are filed. In Warsaw the same set has different names: the annual, half-year, quarterly and current reports, filed through the ESPI system. Press releases and investor presentations belong here too, since the company publishes them itself. Everything built on those documents afterwards, from an analyst note to a sell-side equity research report, is already secondary.

The earnings-call transcript is just as primary, but do not go looking for it on EDGAR. A transcript is not a filing type, and it lands there only in the rare case a company furnishes one as an exhibit to an 8-K. Normally you get it from the call itself, from the company's investor-relations page, or from a service that writes those calls down.

Everything else is secondary: the article about the filing, the analyst note, the forum thread, the summary a language model wrote. Secondary does not mean bad. It means somebody already decided for you what was interesting, which is not the same as what was important.

You can check it in five minutes#

Take any sentence from someone else's analysis and try to check it in five minutes. It works when the sentence carries three things: when, where and who. When, because a filing has a filing date and a period-end date, so a figure from last year cannot pose as the present. Where, because the number sits on a specific page of a specific document and either it is there or it is not. Who, because a sentence from a call belongs to the person who said it and a risk factor belongs to the company that signed it.

The intermediary drops those three in order. A headline keeps the who and loses the where. An analyst note keeps the number and loses the period it refers to. A model-written summary can read well and carry none of them, because it was assembled from text that sounds like a filing rather than from the filing.

The earnings call shows this best. The numbers said on the call are already in the press release, so why bother with the transcript? For the questions. A press release never carries the question somebody asked, or the answer management declined to give, and that is usually where the evening's real information sits. "We'll come back to that next quarter," offered in response to a question about one segment's margin, tells you more than the whole strategy slide.

So the useful question is not whether a claim is true but where it came from, because the second one can be answered on the spot. A citation in a report matters more than its tone because a source link turns a claim into something you can knock down yourself.

ESPI filings and the Polish 10-K#

A Warsaw-listed company does not file a 10-K and has nothing that maps onto it exactly. It files an annual report containing the management report on the company's activity, and that report does the work the narrative sections of a 10-K do. The regulation on current and periodic information requires it to carry an "opis istotnych czynników ryzyka i zagrożeń, z określeniem, w jakim stopniu emitent jest na nie narażony" – a description of material risks and threats, stating how exposed the issuer is to each. Which means a Polish issuer, like an American one, has to write the incriminating paragraph about itself. It just sits in a different chapter.

There is one thing in a Polish annual report that a 10-K does not have. The same regulation opens it with a letter from the chief executive. That letter has to cover the issuer's "najważniejsze dokonania lub niepowodzenia emitenta w danym roku obrotowym i perspektywy rozwoju działalności emitenta na najbliższy rok obrotowy": its achievements or failures in the financial year, and where it expects to go in the next one. The word for failures is there in the statute, and most letters walk politely around it. When one does not, you get the shortest passage in the whole report that settles anything, and you learn whether this management can name a bad year out loud.

The deadlines differ, and you feel it in the calendar. That same regulation gives a Warsaw issuer up to 60 days for a quarterly report, three months for the half-year and four months for the year. The SEC gives the largest filers 40 days for a 10-Q and 60 days for a 10-K.

Which means you wait up to three weeks longer for a full quarter in Warsaw, and two months longer for the year, than a reader of the 10-K. On top of that a Polish issuer owes no quarterly report for the second and last quarter of its financial year, because the half-year and annual reports close those periods. Four sets of numbers a year, then, but only two of them quarterly in the American sense.

One clause of that regulation pays for itself immediately: by the end of the first month of the financial year, the issuer has to publish, as a current report, the fixed dates of every periodic report for the whole year. In January you already know which days will have something to read, and you do not have to guess.

The current report is the local 8-K and goes out without delay once something happens: a change on the board, a large contract, a tender offer, a restatement. An annual report holds two documents at once: the one carrying the auditor's seal and the one that is a speech.

10-K vs 10-Q#

The whole difference comes from one thing: the 10-K is audited and the 10-Q is not. That is why the 10-K is long, narrative and yearly, while the 10-Q is short, numerical and filed three times a year. You read the 10-K to understand the business. You read the 10-Q to see what changed in it since last time.

So why read both when half the numbers repeat? Because the 10-Q is written on the assumption that you already know the 10-K. Its management discussion does not re-explain the business model, it points at the previous period, and without the annual context you are reading the middle of a conversation. Start with the 10-Q and you end up knowing what fell without knowing whether it mattered.

The most valuable part of a 10-K for an individual investor is not the financial tables at all. It is the business description and the risk factors near the front. The numbers exist in ten other places. Those two sections exist nowhere else, because that is the company describing itself in its own words and under its own signature. The regulator publishes a plain guide to reading a 10-K that takes less time than one argument about the company online.

From the business description, pull five things:

• Segments and revenue mix – which part of the business is actually growing.
• Revenue geography – where the money is earned, not where the company is registered.
• Customer concentration – one contract behind half the revenue is a different risk from a million small customers.
• Supplier geography – so that on the day a region blows up you already know whether this company sits there.
• Seasonality – in some industries a soft quarter is the calendar, not a problem.

The reading rule is short: description and risks properly once a year, after that only for the difference.

The fiscal year is not the calendar#

One trap lives inside the document and catches more people than it should. The company, not the calendar, decides when its financial year ends. Retailers like to close in late January so that Christmas and the clearance that follows land in the same period.

The result is that a quarter reported in the spring of 2026 can be labelled "Q1 FY2027", because that company's fiscal 2027 began before calendar 2027 did. The filing date says 2026, the label says 2027, and both are correct. A reader who glanced only at the label walks away believing the numbers are a year fresher than they are.

Treat "FY" as a name, not a timeline. Every filing carries a period-end date and a filing date, and those are the only two dates worth trusting here.

The footnote that changed my mind#

One evening I was working through the 10-K of a company where everything looked right: good margins, a coherent story, every segment in its place. Three pages past the point where most people stop reading sat a short footnote about customer concentration. One sentence, and the picture of the business stopped holding together. The optimism on page one and the risk on page ninety were saying two different things, and the footnote was right.

The difference sits in the signature. A press release comes out of investor relations with nobody's name under it. An annual report is signed by the board, and its numbers pass an audit firm before anyone outside sees them. That does not make one side honest and the other dishonest. It changes the price of being wrong, and the price of being wrong is the only reason anyone writes badly about themselves.

That asymmetry is the entire edge. A filing on EDGAR costs nothing, the transcript costs nothing, access is universal, and almost nobody reaches the end anyway. Not because the document is secret, but because it is long, dull and written for nobody in particular. An advantage you can pay for is common. An advantage you get by being patient is not.

What secondary sources are for#

If the original is sitting there free, what is the commentary for? For telling you the filing exists in the first place, and for the ten years of history somebody has already compressed into a paragraph. That is a real saving.

The trouble starts with the length of the chain. A post quoting a site quoting an analyst note quoting a transcript sits four steps from the source, and at every step somebody rounded a number or dropped a condition from a sentence.

Watch it happen to one sentence. Management says it expects margin to improve if input prices stabilise. An analyst writes that margin should improve in the second half. A news site turns that into an expected margin recovery. By the time it reaches a forum post, margin is recovering. The condition evaporated at step two and the future tense at step four. The fix is not to ban secondary sources; it is to make sure the chain ends in a document you can open yourself.

A language model adds a difficulty of its own. It can produce a fluent paragraph about a filing it has never opened, because it has read a great deal of writing that sounds like a filing. The output looks intelligent and is sometimes partly right, only nothing in that process ties any specific sentence to any specific document. The failure shows most sharply on two numbers: ask a chat window for a quarter's revenue and margin, then for the line and the date it took them from.

What a filing will never tell you#

You could stop here if the primary source were neutral. It is not. The company writes the filing about itself, and it writes it to satisfy a requirement rather than to help you.

The risk section can be a list of sentences carried over from last year, where a real threat sits next to a line about possibly failing to retain key employees. I do not know how much of it is management's judgement and how much is the lawyer's caution, and from the outside that cannot be settled. Something else can: whether a sentence changed year over year. Put this year's risk section next to last year's and read only the differences, because a new paragraph about a single customer or a debt covenant does not appear there by accident.

The second thing a filing will not give you is a competitor seen through the competitor's eyes. The company describes the market as it sees it, and the fact that a rival just cut prices appears in the rival's filing or nowhere. Two filings from the same industry read better than one.

The third gap is the sharpest, and it is usually described wrongly. Forward-looking sentences are in a filing. A forecast is not. That same Polish regulation makes the chief executive's letter discuss "perspektywy rozwoju działalności emitenta na najbliższy rok obrotowy", and makes the management report set out the strategy "wraz z opisem perspektyw rozwoju działalności emitenta co najmniej w najbliższym roku obrotowym". In a 10-K, management's discussion carries the same load.

The trouble is that these are statements of direction, not forecasts. They carry no figure, no date and no condition that would make them false, so a year later there is nothing to hold them against. Anyone hunting a forecast in a primary source usually mistakes exactly those sentences for one, which is worse than finding nothing.

The order that protects you#

Sources are not equal, and the order you take them in decides whose conclusion is in your head when you start doing arithmetic.

1. The company's own document. Business description and risk factors from the 10-K, or the management report from the annual report. Your picture first, somebody else's second.
2. The earnings-call transcript. This is where you see what the analysts actually pressed on, and which question management stepped around.
3. Outside analysis, third and never first. By the time a note is written, its author has picked a side and quietly discounted the facts that did not fit. You read the original earlier so you can see what the note left out.
4. Search and models last, pointed at yourself. The best instruction is not "tell me why this is a good company" but "make the strongest case that I am wrong." That kind of sparring partner has a stopping rule and a prompt of its own.

There is a calendar habit that goes with this. Periodic reports arrive on dates the company published in advance; news arrives whenever somebody needs the traffic. Read a company on its own schedule, because on a daily drip of headlines even a steady business feels like a permanent crisis.

How long does it take at the start? Budget two evenings per company: one for the business description and risk factors, one for the latest transcript. After that the cost drops to a quarter of an hour a quarter, because you are only reading differences. That first stretch is the only hard part of the method, which is precisely why so few people get through it.

One number, two sources#

Say you want to know whether gross margin fell last quarter, and why. Both figures below are invented for the example.

The secondary version reads: "Analysts flagged softer gross margin this quarter, pointing to cost pressure." You do not know which analysts, how much softer, or whose phrase "cost pressure" is.

The primary version reads: "Gross margin fell from 64% to 62% quarter over quarter (10-Q, p. 12). On the call, the CFO attributed the move to higher freight and packaging costs that he expects to normalise over two quarters (Q3 transcript)."

That is the three conditions from the top of this piece, and the two-point difference stops being an impression and becomes a fact with an address. That is the standard I ask of any equity research report, ours included: on a management remark the Transcript opens from the Earnings brief, so you read the sentence in the original instead of trusting a paraphrase.

What I ask of a report#

The rule I apply to anything written about a company fits in one question: can I click through to the source. If I can, I am reading research. If I cannot, I am reading somebody's hunch in a suit, occasionally a correct one.

A Full report is built the same way. It comes from the company's documents and the transcript of the most recent earnings call, rather than from a general impression of what a company like this should look like, and material claims carry a link to the document they came from. Where the document says nothing, the report says so. What those links look like in a finished piece is easiest to see in the sample reports.

A headline says what the company wants you to know. A footnote says what it had to admit.