In the US, 87% of firms with more than $100 million in revenue are private, according to an April 2024 analysis by Apollo's chief economist using S&P Capital IQ data. Most companies worth an investor's time never file a 10-K, so a method built only for public filings breaks at the first hurdle.
To research a company, write down the decision you need to make, confirm the exact legal entity, read what the company says about itself, check those claims against filings or independent data, scan outside signals such as news and hiring, and finish with a one-page memo that argues against your own thesis.
This guide works for public and private targets. It draws on SEC investor guidance and on akta.pro's company news benchmark, which scored 71,408 articles across 133 companies in June and July 2026 and showed that the most common research failure is reading about the wrong company.
Key takeaways
- Decide first. A research question and two exit tests save more hours than any tool.
- Resolve the entity before you read anything. Namesakes and parent companies contaminate research quietly.
- Treat the company's own materials as claims, because the SEC does not vouch for the accuracy of what is filed.
- Public targets have filings. Private targets have notices, registries, and signals. Match the source to the target.
- End with a memo that contains the case against your own thesis.
The 7 steps to research a company at a glance
| Step | Question it answers | What you produce |
|---|---|---|
| 1. Decision and exit tests | What am I deciding, and what would make me walk away? | One sentence and two or three exit tests |
| 2. Entity check | Which exact company is this? | An entity card with legal name, domain, parent, status |
| 3. Company claims | What does management say is true? | A numbered list of claims to verify |
| 4. Verification | Do filings or independent data support those claims? | The claims list, annotated with sources |
| 5. Outside signals | What happened that the company did not headline? | A dated timeline of events |
| 6. Peer comparison | Is this good, or only normal for the industry? | The same metrics for three peers |
| 7. Memo | What do I conclude, and what would change my mind? | One page |
Each step ends in an artifact, so a colleague can audit your work without redoing it.
Step 1: Start with the decision and the exit tests before you research a company
When you research a company, start with the decision, not the company. Name the action you might take (invest, pass, monitor, or escalate to diligence) and write two or three facts that would make you walk away. Research with no decision collects trivia, and research with no exit tests drifts toward whatever you found first.
Exit tests must be specific and checkable. Examples: an auditor's report that raises doubt about the company continuing, one customer accounting for most of revenue, or a founder with a prior company that ended in investor litigation.
Write them before you see any data. Afterward, you will bend them to fit the story you like.
Step 2: Confirm the exact entity before you open a document
Confirm whether the target is public, private, or a subsidiary, then record its legal name, website, headquarters, and parent. Do this first because name searches return look-alikes, and every later step inherits the error. A careful analysis of the wrong company is still wrong.
University library guides flag the traps. A search for Hilton returns both Hilton Hotels and its holding company, Hilton Worldwide. The financials for a subsidiary such as Marvel sit inside its parent's annual report, which is Disney's.
Private companies make the problem worse. The akta.pro benchmark deliberately included namesake companies, and the three news platforms it tested attributed only 32% to 58% of returned articles to the correct company. akta.pro reached 93%. One of the study's headline findings reads: "Precision is the whole game."
Fill in an entity card before moving on: legal name, aliases, domain, headquarters, founding year, ownership type, and parent. For private targets, a resolved ID from a database such as akta.pro's company data API gives every later lookup a fixed anchor. If you want the mechanics, see how entity resolution affects company data quality.
Step 3: Read what the company says about itself, then treat it as claims
Start with the company website, the investor presentation, and the Business section of the 10-K or offering document. These show how management wants to be seen. Convert each statement into a claim to verify, because the SEC itself does not vouch for the accuracy of a 10-K or 10-Q.
Extract five claims: what the company sells, who buys it, how it earns money, why it wins, and what it plans next. Label each as numeric or qualitative.
Numeric claims, such as growth rates and customer counts, are the easiest to check. They are also the most revealing when they fail.
Step 4: Verify the claims with filings or alternative data
Match the source to the company type. US public companies file 10-Ks, 10-Qs, and 8-Ks on EDGAR. Private companies leave thinner traces, such as Form D notices, registry records, and funding and hiring signals. Whichever applies, check each claim from step 3 against a source the company did not write.
| Target | Primary sources | What to pull | Limit |
|---|---|---|---|
| US public company | 10-K, 10-Q, 8-K, proxy statement, and Form 4 on EDGAR | Business, risk factors, MD&A, audited financials, insider trades | Filings look backward |
| Subsidiary | Parent company's annual report | Segment results and notes | Detail may be thin |
| US private company raising capital | Form D on EDGAR, state registry records | Offering amount, related persons, first sale date | Not every raise produces a Form D |
| Any private company | Structured company data, funding history, headcount and job postings | Estimated revenue, headcount trend, investors, leadership | Estimates are ranges, not audited facts |
Public targets are about reading in the right order. The SEC's guide to the 10-K includes Business, Risk Factors, MD&A, and the audited financial statements. Read them in that sequence. The SEC also fixes the order of topics in a 10-K, so you can compare this year's risk factors with last year's. A risk that is new, reworded, or dropped shows what management's lawyers now consider material.
Then test five numbers that any analyst asked to research a company should be able to produce:
-
Revenue growth over at least three periods
-
Gross or operating margin trend
-
Operating cash flow compared with net income
-
Cash compared with debt
-
Customer concentration, where disclosed
Cash flow that trails profit for several periods deserves a line in your memo.
Private targets are about triangulation. A company that raises under Regulation D must file a Form D within 15 days after the first sale, which often makes it one of the earliest public traces of a private raise. When registries run dry, structured data fills the gap. akta.pro's database covers 20M+ companies with 75+ data points each, including funding detail, management profile, and headcount range. Record every estimate as an estimate, with its source and date.
Step 5: Check outside signals, from news to insider trades
Read news by event type, not by headline. Filter the last 12 to 24 months for leadership changes, legal and regulatory actions, funding, layoffs, and product launches, and confirm each article concerns your entity. Filings show what a company chose to report. News shows what happened around it.
Source choice matters more than most analysts expect. In its June to July 2026 test, akta.pro compared ten providers on the same 133 companies:
| Provider | Type | Right-company precision | Recall | F1 | Cost per 1,000 correct articles |
|---|---|---|---|---|---|
| akta.pro | Company news API | 93% | 72.7% | 81.3 | $0.50 |
| GPT-5.5 | LLM with web search | 81% | 55.9% | 62.5 | $34.42 |
| SerpAPI | News platform | 58% | 50.5% | 52.6 | $0.67 |
| Perigon | News platform | 39% | 73.7% | 48.6 | $0.73 |
| NewsAPI | News platform | 32% | 59.4% | 41.6 | $0.72 |
Source: akta.pro company news retrieval benchmark, July 2026. akta.pro designed and ran this benchmark, so treat it as a vendor result. The harness and company list are published for re-testing.
Broad feeds find volume and still miss the point. Perigon posted the highest recall in the study, with only 39% right-company precision. For research, a smaller feed that is about the right company beats a bigger feed you must clean by hand. In the same study, akta.pro surfaced 31 stories that no other provider found, a reminder that one feed is rarely the whole picture.
Add three people checks. For public companies, insiders must file a Form 4 within two business days of a transaction, so a cluster of sales or purchases is a timely signal. For every company, read the executives' prior roles and any litigation tied to them. For private companies, watch job postings and headcount, because hiring freezes show up there before they show up anywhere official. To automate this layer for private targets, see how to monitor private company news using an API.
Step 6: Compare against three peers
Pick three direct peers and run the same five numbers from step 4. A 12% operating margin means little alone. It means a great deal if peers sit at 5%, and very little if they sit at 25%.
Library research guides make the same point: how you interpret a company's financials depends heavily on its industry. If you cannot find peers with filings, use public competitors of a private target as a proxy and say so in your notes.
Step 7: Write the one-page memo and argue the other side
Write the memo last, and make it a page. Force yourself to state the strongest case against your view. Charlie Munger's rule was that you must "force yourself to consider arguments on the other side", and a memo with an empty "evidence against" section is a sign the research stopped early.
| Memo section | What goes in it |
|---|---|
| Decision and thesis | The action under consideration and the reason in two sentences |
| Evidence for | Each point with its source and date |
| Evidence against | The strongest counterpoints, including any exit test that nearly triggered |
| Open questions | Claims you could not verify, and who could answer them |
| Verdict and triggers | Proceed, pass, or monitor, plus the events that would change your mind |
The last test is honesty about your own limits. Warren Buffett wrote in his 1996 letter that the size of your circle of competence is not important, but "knowing its boundaries, however, is vital". If you cannot explain what would break the thesis, you are outside it. Mark the target as monitor and move on.
Frequently asked questions
What is company research?
Company research is structured work to learn how a business operates, earns money, is run, and competes, so that you can make one defined decision about it. For investors that decision is usually to invest, pass, or investigate further.
How do you research a private company?
Resolve the exact entity first. Then use Form D filings on EDGAR where the company raised capital under Regulation D, state registry records, court records, news signals, and structured company data for funding, headcount, and leadership. Expect estimates instead of audited financials, and record the source and date of every figure.
Where can I find a public company's financial statements for free?
The SEC states that all 10-Ks and 10-Qs filed with it are available to the public on EDGAR. Most companies also post them in the investor relations section of their websites.
What is the difference between company research and due diligence?
Company research decides whether a target deserves more work. Due diligence comes later and verifies material facts, often with non-public documents and management access, before capital is committed.
The next time you research a company, run steps 2, 4, and 5 on one live target and count how many of its claims survive. For a private company, you can search profiles across 20M+ private companies or pull a 12-month signal feed through akta.pro news signals.