GeminIQ
Subscribe

How to Value a Company From Filings Alone

Chad Hartman

By

Published · Last updated

Every valuation model — a DCF, a comparable-multiples analysis, a dividend discount model, a simple book-value screen — is built from two fundamentally different kinds of inputs, and almost no valuation discourse bothers to separate them. One kind traces directly to a specific line in a specific filing: revenue, cash flow, share count, historical margins. The other kind is a forward-looking judgment a reasonable analyst could disagree with: a discount rate, a terminal growth rate, a peer multiple, a projected margin five years out. A finished valuation presents both as a single number, with a single level of confidence — and that's the part worth unlearning.

This guide separates the filed from the assumed across every major valuation approach: what a 10-K and 10-Q actually give you directly, where every model has to reach past the filing for something the filing doesn't contain, and a general framework for auditing any valuation — yours or someone else's — by asking which half of it is fact and which half is judgment.

Start your 7-day free trial →

Table of Contents

What Filings Alone Can Actually Tell You

A 10-K and its preceding 10-Qs give you a complete, auditable record of what already happened: historical revenue and its growth rate, historical margins at every level of the income statement, historical free cash flow, the current balance sheet in full, share count, historical dividends paid, and enough history to compute trailing returns on capital. Every one of these traces to a specific XBRL tag in a specific filing, which means every one of them can be independently verified by anyone willing to open the document.

What filings cannot give you, under any circumstances, is anything about the future. That's not a limitation of any particular company's disclosure — it's structural. A filing reports what happened during a period that has already closed. Valuation is inherently a forward-looking exercise. The gap between those two facts is where every assumption in every valuation model has to live.

Why Every Valuation Model Needs Something Filings Don't Have

No valuation methodology gets around this, because the problem isn't a gap in any particular model — it's a gap between what a filing is and what a valuation needs. A filing answers "what did this business generate." A valuation needs to answer "what is this business worth," which requires a view on what it will generate, and how much a dollar of that future generation is worth today. Filed data alone cannot answer either half of that question.

What varies by model is how much of the valuation gets built from the reliable, filed half before the assumption enters, and how exposed the final number is to a single judgment call. That's the actual difference between valuation methodologies worth understanding — not which one is "more accurate," but which one asks you to trust the least amount of unfiled judgment to get to an answer.

The four approaches covered below split like this:

Valuation model Inputs that are filed Inputs that are assumed Where the assumption concentrates
Discounted cash flow Base-year free cash flow, margin structure, capital intensity — reconstructible from the cash flow and income statements Future revenue growth, future margin trajectory, terminal growth rate, and the discount rate (risk-free rate, equity risk premium, beta) The discount rate — the final output is as sensitive to that one rate as to every growth assumption combined
Comparable multiples The company's own earnings, EBITDA, or revenue, and each peer's own reported figures Which companies qualify as comparable, which period's figures to use, and the peer group's current multiple — a live market price, never reported in any filing The multiple
Asset-based / book value Every asset and liability line on the balance sheet Whether filed book value approximates present worth — largest for goodwill and intangibles carried at historical acquisition price Goodwill and intangibles, and historical-cost physical assets
Dividend discount Every dividend actually paid, and the growth rate of that history, from the financing activities section The future dividend growth rate and the required rate of return Split across both — and the model only applies to companies with an established, stable dividend policy

DCF: The Cash Flows Are Filed, the Discount Rate Isn't

A discounted cash flow model's historical foundation — the free cash flow figures used to establish a base year, the margin structure, the capital intensity of the business — is entirely filed data, reconstructible from the cash flow statement and income statement directly. Our full DCF walkthrough builds exactly this foundation from a real 10-K, step by step.

Everything past that base year is assumption. Future revenue growth rates, future margin trajectories, and the terminal growth rate applied to the final projected year are all judgment calls a filing cannot supply, however carefully they're reasoned. The discount rate carries the same problem in a more concentrated form: WACC requires a risk-free rate, an equity risk premium, and a beta — none of which appear anywhere in a 10-K, because none of them are company data at all. A DCF's final output is exactly as sensitive to that one rate as it is to every growth assumption combined, which is why the discount rate is the single highest-leverage assumption in the entire model. Platforms that publish a fair-value estimate have already made that choice on your behalf rather than exposing it — how GeminIQ compares to Finbox covers that tradeoff directly.

Comparable Multiples: Your Company's Numbers Are Filed, the Multiple Isn't

A comparable-multiples valuation looks deceptively simple: take the company's own earnings, EBITDA, or revenue — all filed — and multiply by whatever a peer group is currently trading at. The company's half of that equation is as reliable as any figure on this list. The multiple itself is not.

Three separate judgment calls sit inside "the multiple." Two are visible immediately: which companies actually qualify as comparable, and which period's figures to use for the peer group. The third gets skipped past most often: the peer group's current multiple is itself a live market price, not a filed number. A peer's own earnings are filed. What the market is currently willing to pay for those earnings is observed, continuously, in trading, and was never reported by anyone in a periodic filing. Applying "the industry trades at 15x" to a company's own filed earnings imports a market judgment wholesale, however precise the arithmetic looks once it's applied — and that imported judgment is itself an expectation about the future, priced into the multiple rather than filed anywhere.

Asset-Based Valuation: Almost Entirely Filed, With One Open Question

Book value, net asset value, and liquidation-style approaches come closer than any other methodology to being fully derivable from a filing, because the inputs — every asset and liability line on the balance sheet — are filed by definition. This is the valuation approach with the smallest assumption-dependent surface area of anything on this list.

The one open question that remains is whether the filed book value of an asset approximates what it's actually worth today. Goodwill and intangibles carry the largest gap between book value and any real-world sense of value, since they represent historical acquisition prices rather than current market worth, and depend entirely on whether impairment testing has kept pace with reality. Physical assets carried at historical cost less depreciation can run either direction — understated for appreciating real estate, overstated for equipment nearing obsolescence. None of this requires a discount rate or a growth projection, which is exactly why this approach is the most filing-grounded of the four covered here.

Dividend Discount Model: The History Is Filed, the Future Isn't

A dividend discount model's history — every dividend a company has actually paid, and the growth rate of that history — is filed data, directly reconstructible from the cash flow statement's financing activities section across as many periods as a company has reported. That history is also the least useful part of the model, because a DDM's output depends entirely on two things a filing cannot supply: the future dividend growth rate and the required rate of return applied to discount it.

The model is also structurally narrower than the others on this list — it only produces a meaningful answer for companies with an established, relatively stable dividend policy, which rules it out entirely for a large share of public companies regardless of how good the filed history looks.

A Framework for Auditing Any Valuation

Every input in every model on this list falls into one of two categories, and naming which one applies to each line is the entire exercise. A filed input traces to a specific tag in a specific filing, is identical no matter who calculates it, and won't change until the next filing posts. An assumed input requires a judgment call a reasonable, informed analyst could make differently, and will shift depending on who's making it and when.

Applying that split to a finished valuation — anyone's, including your own — means asking, line by line: does this number come from a document I could open and verify, or does it come from a choice someone made? A valuation built mostly on the first kind deserves real confidence. A valuation that looks precise but is mostly the second kind, dressed up in enough decimal places to look objective, deserves exactly the confidence its weakest assumption can support — no more.

What This Means for Reading GeminIQ's Metrics

GeminIQ's Calculated Metrics are built entirely from the filed half of this framework, deliberately. Return on Invested Capital, historical margins, and growth rates all trace to specific XBRL tags with one published, consistent method — the same number, calculated the same way, for every company. GeminIQ does not calculate WACC, a target multiple, or a projected terminal value, for the same reason none of those numbers belong on this side of the line: doing so would mean silently picking a beta, a peer group, or a growth rate on the user's behalf and presenting the result with the same confidence as a figure that traces to an actual filing. ROIC vs. ROE vs. WACC covers why two of those three are reconstructible from filings and the third isn't.

The raw filed inputs every one of these models needs — cash flow, share count, capital structure, historical growth — are available directly. The judgment calls stay exactly where they belong: with the person building the valuation, made visibly rather than buried inside a number that looks more objective than it is.

Frequently Asked Questions

Can a company be valued using only its SEC filings, with no assumptions at all?

No. Filings report what already happened; valuation requires a view on the future, which no filing contains. What varies by methodology is how much of the final number rests on filed history versus forward-looking judgment — asset-based approaches lean furthest toward filed data, DCF and comparable multiples lean furthest toward assumption.

Which valuation model relies least on unfiled assumptions?

Asset-based and book-value approaches, since nearly every input is a balance sheet line item filed directly. The tradeoff is that they answer a narrower question — what the assets are worth — rather than what the ongoing business is worth as a going concern.

Why doesn't GeminIQ calculate a target price or fair value estimate?

A target price necessarily blends filed data with assumptions — a discount rate, a multiple, a growth projection — that require analyst judgment rather than being extracted from a filing. GeminIQ provides every filed input those models need and calculates every metric that's fully reconstructible from as-filed data, without asserting a single "correct" answer for the parts of any valuation that are inherently a judgment call.

Is a comparable-multiples valuation more or less reliable than a DCF?

Neither is inherently more reliable — both blend filed and assumed inputs, just in different places. A comparable-multiples valuation concentrates its assumption in the multiple selected; a DCF spreads its assumptions across the discount rate, the growth trajectory, and the terminal value. Understanding where each model's assumption lives matters more than picking one model as universally superior.

Two valuations can use the same filed data and reach very different answers, and the entire difference will trace to the assumptions, not the facts. Knowing which parts of a number are which is the actual skill.

Start your 7-day free trial →

Wall Street's data. Main Street's price.

Institutional terminals charge thousands a year for as-filed accuracy. GeminIQ gives you the same thing for a fraction of the cost: financials built directly from raw SEC EDGAR filings, not third-party APIs, with full XBRL traceability back to the original 10-K or 10-Q. No normalized guesswork, just calculated metrics, charts, screeners, and watchlists built on numbers exactly as the company reported them. Start researching now at GeminIQ.com.

Disclaimer: The content in this blog is for educational and entertainment purposes only and does not constitute financial, legal, or tax advice. Investing involves risk, including the loss of principal. The views expressed are my own and not intended as financial advice or a guarantee of future performance.