What Are Primary Sources in Stock Research?

Investing writing varies wildly in how seriously it takes its own claims. Some pieces link every number to a filing. Some quote management straight from the transcript. Others assert a margin trend or a balance-sheet worry with no anchor at all - the words simply appear, and you're expected to take them on faith, like a fortune cookie with a Bloomberg terminal.

That difference isn't stylistic. It's the line between primary-source research and secondary commentary, and it's the single most useful distinction an investor can learn to see.

What a Primary Source Actually Is#

A primary source is a document the company itself filed, recorded, or published - the raw disclosure, before anyone else has interpreted it. In equity research, the canonical ones are narrow and well-defined:

• The 10-K. The annual report filed with the SEC. Audited financials, a description of the business and its segments, the risk factors management is willing to put in writing, and the MD&A - management's own discussion of results.
• The 10-Q. The quarterly report. Shorter and unaudited, but it carries the freshest income statement, balance sheet, and cash-flow statement, plus updated commentary on the quarter.
• The 8-K. Filed when something material happens between regular reports - a leadership change, a major contract, a restatement, an acquisition. If it matters and it's sudden, it shows up here first.
• Earnings-call transcripts. Verbatim records of prepared remarks and analyst Q&A. The same numbers appear in the press release, but the transcript is where you see what management was asked, what they answered, and - often more revealingly - what they declined to answer.
• The proxy statement. Filed before the annual meeting. The cleanest source on executive pay, board composition, and how management's incentives are actually wired.
• Press releases and prepared exhibits. Material announcements the company has put on the record itself.

The unifying property is provenance. Each document exists at a specific URL, on a specific date, with a specific filer. For U.S. companies they're all free on the SEC's EDGAR database - which means a claim sourced from one can be checked, line for line, by anyone willing to open it. (The SEC even publishes a plain-English guide to reading a 10-K, which is more than most people who quote one have done.)

What to Actually Mine From the Filing#

The two most valuable parts of a 10-K for an individual investor aren't the financial tables - they're the business description and the risk factors at the front. Read the financials elsewhere; read these two for the things a balance sheet can't tell you.

The business description answers: what does the company actually sell, to whom, and where? You'd be surprised how often the answer isn't what the brand suggests - a company can be incorporated in one country and earn most of its revenue in two others, or carry a famous consumer name while a quieter segment quietly pays the bills. From the narrative, pull:

• Segments and revenue mix - which part of the business is actually growing.
• Geography - where the revenue is earned, not where the company is registered.
• Customer concentration - one big contract behind half the revenue is a very different risk from millions of small customers.
• Supplier and supply-chain geography - so when a region hits a political crisis or a logistics shock, you already know whether this company is exposed.
• Seasonality - in some industries a soft quarter is just the calendar, not a problem. Without this, you'll over-read one bad print.
• The risks management itself names - if a company knows something could seriously hurt it, it has to disclose it. That section is the company telling on itself, in writing.

Good news for your weekend: the business and risk sections don't change much quarter to quarter. Read them properly once, then refresh - a quick re-read each quarter, a thorough one at least once a year. You're hunting for what changed, not re-reading a novel.

Why a Source Beats a Confident Guess#

This distinction turned urgent the moment it became trivial to generate confident-sounding paragraphs about a company without ever opening a filing. A general language model can write fluent prose about a 10-K it has never read, because it has read a great deal of writing that sounds like a 10-K. The output is plausible. It may even be partly correct. But nothing in the writing process ties any specific sentence to any specific document.

Primary sources break that pattern three ways:

1. They're concrete. A revenue figure on page 47 of a 10-K is either there or it isn't. A claim that quotes it can be verified in seconds; a claim that doesn't, can't.
2. They're dated. Every filing has a filing date, so stale information stops hiding behind present-tense prose.
3. They're attributable. A statement from a transcript belongs to the person who said it. A risk factor belongs to the company that filed it. Knowing whose claim it is, is itself information.

I learned to trust this the hard way. I once spent an evening on a 10-K I was ready to like, right up to a footnote three pages past where most people stop reading. One sentence about customer concentration reframed the whole business - the cheerful story on page one and the risk on page ninety did not agree, and the footnote was right. Now I read filings backwards as often as forwards. The headline is written to be read; the footnotes are written because they have to be.

That's the discipline that separates a study from a guess. I made the same argument from a different angle in Why Not Just Ask ChatGPT to Analyze a Stock? - un-anchored research is exactly the failure mode primary-source discipline prevents.

Read the End Date, Not the Fiscal-Year Label#

One trap lives inside the filings themselves and catches more investors than it should: the fiscal year is not the calendar year. A company gets to define when its own financial year starts and ends, and plenty of them don't run January to December. Retailers in particular love a fiscal year that closes in late January.

So a quarter a company actually reported in the spring of 2026 can be stamped "Q1 FY2027," because that company's fiscal 2027 began before calendar 2027 did. The filing date says 2026. The fiscal label says 2027. Both are correct - and a reader who skims only the year on the cover walks away thinking the numbers are a year newer than they are.

The rule of thumb is short: trust the dates, not the fiscal tag. Every filing carries a period-end date and a filing date - read those, and treat the "FY" number as a name, not a timeline. It's the same dating discipline that makes a balance sheet readable across years.

Where Secondary Sources Fit#

Secondary sources are everything that comments on a primary source without being one: news articles, sell-side notes, podcasts, forum posts, blog summaries, AI overviews. They aren't useless - they're often where you first hear a filing exists, or where a long history gets condensed into a paragraph.

But they're commentary. A news article describing a 10-K is not a 10-K. A sell-side note quoting management is not the transcript. Each step away from the source adds a layer where a number can drift or a quote can be paraphrased into something the speaker didn't quite say.

And the risk compounds. A blog post quoting a news article quoting an analyst note quoting a transcript is four layers from the source - a financial game of telephone, and by the end the original sentence is often unrecognizable. The fix isn't to ban secondary sources; it's to make sure the chain of evidence eventually terminates in a primary document you could open yourself.

The Order to Read Them In#

Sources aren't equal, and the order you read them in protects you from inheriting someone else's bias. The sequence that works:

1. The filing narrative first. Business and risk sections of the 10-K. Form your own picture from what the company says about itself before anyone frames it for you.
2. The earnings-call transcript second. It explains the context behind the numbers and shows you what the analysts - the same people who publish the price targets - thought was worth pushing on. Reading between the lines of that transcript is its own skill.
3. Third-party analysis third, never first. A good note can condense a long history or surface something you missed. But by the time someone writes the take, they've picked a side, and they'll quietly underweight the facts that don't fit it. Read the primary source first precisely so you can see what got left out.
4. Search and AI last, to attack your own view. Use them for category trends, demand, and competitors - and point them at your own conclusion. The most valuable prompt isn't "tell me why this is a good company," it's "make the strongest case that I'm wrong." If it can't, you've learned something; if it can, you've learned more.

One more habit worth installing: prefer a monthly or quarterly read of analysis over a daily drip of news. The internet is a content factory - more headlines mean more traffic means more ad revenue, and fear travels fastest, so the feed skews negative. Read a stable company's news every day and you'll feel a permanent crisis that the filings simply don't show. The primary sources move on the company's schedule. The news moves on the algorithm's.

A Concrete Example: The Same Insight, Two Sources#

Say an investor wants to know whether a company's gross margin compressed last quarter, and why.

The secondary-source version reads: "Analysts noted softer gross margins this quarter, with several attributing the move to input-cost pressure." Non-trivial, but you have no idea which analysts, what magnitude, or whether "input-cost pressure" is management's framing or a journalist's gloss.

The primary-source version reads: "Gross margin declined from 64.2% to 61.8% quarter over quarter (10-Q, p. 12). On the call, the CFO attributed the change to 'higher freight and packaging costs that we expect to normalize over the next two quarters' (Q3 transcript, prepared remarks)." Not more elegant, not even shorter - just checkable, dated, and attributable. (Those figures are illustrative, not a real company. The point is the shape, not the number.) That is the entire game.

This is the standard a credible report has to meet - one of the structural features in What Is an Equity Research Report?.

How Taufolio Uses Primary Sources#

Taufolio is built around this discipline, in roughly that same order. When we generate a company report, the pipeline pulls the relevant filings from SEC EDGAR, opens the actual transcript of the most recent earnings call, and constructs the report from those documents - not from a general impression of what such a company should look like.

Meaningful claims are anchored. A revenue figure traces back to the line item it came from. A management quote traces back to the speaker and the call. When the source doesn't say something, the report says so, rather than inventing a fluent guess - refusing to manufacture coverage where the filing is silent is exactly what separates the work from commentary that happens to mention the same ticker. For the longer view of how those documents fit a full diligence pass, see How to Analyze a Company Before You Invest.

The Practical Takeaway#

The shorthand for judging stock research is simple: ask where its claims come from. If the writing routinely points back to filings, transcripts, and proxy statements - and you can follow the pointer - it's research. If the claims float on their own, fluent and unattributed, it's commentary. Both can be useful, but only the first can be checked, and only what can be checked is worth a real position.

Primary sources are not glamorous. They're long, dense, and written by lawyers who bill by the clause. They're also the ground truth - the only thing that keeps the rest of the research stack honest. Read the filing, argue with it, and if you'd rather not spend the evening with a 10-K and its footnotes, that's exactly what Taufolio reports are for: Short Recap keeps you current on the last month of company events, while Deep Research works through the business from the ground up, with material claims tied to the document they came from. You can see what that looks like in our sample reports.