---
title: "How to Value a Company: DCF and Comps with AI Help"
url: "https://goremy.ai/blog/articles/how-to-value-a-company-dcf-and-comps-with-ai-help"
author: Luis Chavez-Mattos
editor: Luis Chavez-Mattos
editor_title: Director, Product
published: "2026-11-19T10:20:11.641Z"
updated: "2026-11-19T10:20:11.641Z"
topic: Budgeting and Forecasting
description: "Learn how to value a company with a DCF and trading comps: the step-by-step method, a worked example with numbers, and AI prompts that check your work."
---

# How to Value a Company: DCF and Comps with AI Help

> Learn how to value a company with a DCF and trading comps: the step-by-step method, a worked example with numbers, and AI prompts that check your work.

Edited by Luis Chavez-Mattos, Director, Product · Published 2026-11-19

## How do you value a company?

You value a company by estimating what its future cash is worth today, then checking that estimate against what the market pays for similar businesses. The first part is a discounted cash flow (DCF): forecast free cash flow, discount it back at a rate that reflects risk, and add a terminal value for the years beyond the forecast. The second part is comparable company analysis ("comps"): apply the valuation multiples of similar companies, such as enterprise value to EBITDA, to your target. Run both, compare, and report a range, not one number.

## TL;DR

- **Use two methods, not one.** A DCF tells you what the business is worth on its own cash flows. Comps tell you what the market currently pays for similar businesses. When they disagree, the gap is the interesting part.
- **Start with enterprise value, then bridge to equity value.** Subtract net debt (and other claims like minority interest) from enterprise value to get what belongs to shareholders, then divide by diluted shares for a per-share figure.
- **The terminal value usually drives the answer.** In the worked example below, about three quarters of the value sits in the terminal value, so small changes to the discount rate or growth rate move the result a lot.
- **Show a range with a sensitivity table.** Executives need to see how the answer shifts when assumptions change, not a single falsely precise number.
- **AI is good at the checking and the drafting, not at the data.** Paste in numbers from filings yourself, and use AI to audit your logic, find inconsistencies and write the memo. Do not let it supply financial figures from memory.

## What are the main ways to value a company?

There are three families of method. Most real work uses the first two together.

| Method | What it answers | Best for | Weak spot |
|---|---|---|---|
| DCF (intrinsic) | What are the future cash flows worth today? | Stable businesses with forecastable cash | Highly sensitive to discount rate and terminal assumptions |
| Trading comps (relative) | What does the market pay for similar public companies today? | Quick, market-anchored view | Good comparables are hard to find; markets can be mispriced |
| Precedent transactions (relative) | What have acquirers paid for similar companies? | M&A and sale discussions | Includes control premiums; deals are old or sparse |

Other approaches exist: asset-based valuation (useful for holding companies or distressed businesses) and, for early-stage companies, revenue multiples or venture-style methods. This guide focuses on DCF and comps because they are what a model request usually means.

## How do you build a DCF, step by step?

1. **Define the forecast period and gather history.** Use five years for a mature company, up to ten for one still growing toward steady state. Pull three to five years of historical financials from the annual report (10-K for US public companies) so your forecast starts from reality.

2. **Forecast revenue.** Build it from drivers where you can: volume times price, customers times revenue per customer, or segment by segment. A growth rate that fades toward the long-run economy's growth rate is easier to defend than one that stays high and then drops.

3. **Forecast operating profit and unlevered free cash flow.** Unlevered free cash flow is the cash the business generates before paying anyone who financed it:

   EBIT, less taxes on EBIT, plus depreciation and amortization, less capital expenditure, less the increase in net working capital.

   "Unlevered" means you ignore interest, so the value does not depend on how the company is financed.

4. **Choose the discount rate (WACC).** The weighted average cost of capital blends the cost of equity and the after-tax cost of debt, weighted by how the company is financed. Cost of equity is usually built with the capital asset pricing model: a risk-free rate, plus beta times an equity risk premium. Use sources you can cite for each input, and state the date.

5. **Calculate terminal value.** Two common ways:
   - **Perpetuity growth:** final-year FCF times (1 + g), divided by (WACC minus g). Keep g at or below long-run economic growth, typically low single digits.
   - **Exit multiple:** final-year EBITDA times a multiple taken from comps.

   Cross-check one against the other. If the perpetuity method implies an exit multiple far from what peers trade at, one of your assumptions is off.

6. **Discount everything to today.** Divide each year's FCF by (1 + WACC) raised to the year number. Do the same to the terminal value using the final year's factor. Add them up. That sum is enterprise value.

7. **Bridge to equity value.** Enterprise value, less debt, plus cash, less other claims (minority interest, preferred stock, and similar), equals equity value. Divide by diluted shares outstanding for value per share. Use the latest balance sheet and include options and convertibles through the treasury stock method.

## What does a DCF look like with real numbers?

This is a hypothetical company, "Northfield Co.", with made-up figures so you can follow the arithmetic. Dollar amounts are in millions.

**Assumptions:** WACC of 10%, terminal growth of 2.5%, current EBITDA of $20, debt of $40, cash of $10, 10 million diluted shares.

| Year | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|
| Free cash flow | 10 | 12 | 14 | 16 | 18 |
| Discount factor at 10% | 0.9091 | 0.8264 | 0.7513 | 0.6830 | 0.6209 |
| Present value | 9.1 | 9.9 | 10.5 | 10.9 | 11.2 |

- Sum of present values of years 1 to 5: about $51.6
- Terminal value: 18 × 1.025 ÷ (0.10 − 0.025) = $246.0
- Present value of terminal value: 246.0 × 0.6209 = about $152.7
- **Enterprise value: about $204.4**
- Net debt: 40 − 10 = $30
- **Equity value: about $174.4**, or roughly **$17.44 per share**

Note that the terminal value is about 75% of enterprise value. That is typical, and it is why you must show sensitivity. Holding terminal growth at 2.5%:

| WACC | Enterprise value |
|---|---|
| 9% | about $238 |
| 10% | about $204 |
| 11% | about $179 |

One point of discount rate moves the answer by roughly 15% in either direction. Any executive reading your model should see this table.

## How do you value a company with comps?

Comps work on a simple idea: similar businesses should trade at similar multiples. The steps:

1. **Pick the peer set.** Choose public companies in the same industry, with similar size, growth, margins, geography and business model. Five to ten is typical. Write one line justifying each; this is the part reviewers challenge first.

2. **Pull consistent financials.** Get market capitalization, debt, cash and the metrics you will use, such as revenue, EBITDA and net income. Make sure fiscal year-ends line up (calendarize if not) and that you use the same basis across all peers: last twelve months or forward estimates, not a mix.

3. **Calculate enterprise value for each peer.** Market cap plus debt, minus cash, plus minority interest and preferred stock.

4. **Compute the multiples.** Common ones are EV/EBITDA, EV/Revenue (for fast-growing or unprofitable companies) and P/E. Match the multiple to the business: EV/EBITDA for most industrials and services, EV/Revenue when profits are not yet meaningful.

5. **Take the median and the quartiles.** The median resists outliers better than the average. Look at the 25th to 75th percentile range to frame the answer.

6. **Apply to the target.** Multiply the target's metric by the multiple to get enterprise value, then bridge to equity value the same way as in the DCF.

**Worked example.** Suppose Northfield's five peers trade at EV/EBITDA of 9.0x, 10.5x, 11.0x, 12.0x and 13.0x. The median is 11.0x, and the middle range runs from 10.5x to 12.0x. With Northfield's EBITDA at $20:

- Median: 11.0 × 20 = **$220** enterprise value
- Range: 10.5x to 12.0x gives **$210 to $240**

Then ask whether Northfield deserves to sit above or below the median. If it grows slower or has thinner margins than the peers, a discount is reasonable. If it is better, a premium is. Explain which and why.

## How do you combine the two into one answer?

Line the results up side by side:

| Method | Enterprise value |
|---|---|
| DCF (WACC 10%, g 2.5%) | about $204 |
| DCF sensitivity range (WACC 9% to 11%) | about $179 to $238 |
| Trading comps (10.5x to 12.0x EBITDA) | $210 to $240 |

The DCF's $204 implies about 10.2x EBITDA (204 ÷ 20), which sits inside the peer range. That is a good sign: the two methods tell a consistent story. A reasonable conclusion would be an enterprise value of roughly $200 to $240, with the exact point depending on how much you trust the forecast versus the market's current mood.

If the methods disagree sharply, do not average them to hide it. Find the cause. Common ones: the forecast assumes growth the peers do not have, the peers are priced during a boom or slump, or the terminal assumptions imply an unrealistic exit multiple.

If you add precedent transactions, expect higher values, since buyers pay a control premium. Label that row clearly.

## Where can AI help with a valuation, and where should you not trust it?

AI is useful in a valuation workflow, but unevenly.

**Where it helps:**
- Reading a 10-K or earnings transcript and pulling out items that affect the forecast: segment trends, capital spending plans, one-time charges, debt maturities.
- Auditing your logic. It is good at spotting that your terminal growth rate exceeds your WACC, that your tax rate changes between years, or that you used a diluted share count in one place and basic in another.
- Suggesting why a company might belong in or out of a peer set, so you have a list to verify.
- Writing the first draft of the methodology section and the executive summary.
- Explaining a concept to a non-finance stakeholder in plain language.

**Where it should not be trusted:**
- **Figures from memory.** A chat model can state a company's revenue, beta or share count confidently and wrong, or out of date. Every number should come from a filing, a market data provider or your company's own systems, and you should paste it into the prompt.
- **Arithmetic across many steps.** Do the calculation in your spreadsheet and ask AI to review the formulas, not to produce the final result.
- **Current market data.** Interest rates, risk premiums and peer prices change. A model without live data access will guess. Check the date on every input.
- **Judgment calls.** Whether a growth fade is credible, or whether two companies are really comparable, is your call. AI can argue either side; you decide and defend it.

Also check your company's policy before pasting nonpublic financials into any AI tool. If the target is confidential, use an approved internal tool.

### Prompt 1: audit a DCF

> You are a skeptical valuation reviewer. Below are my DCF assumptions and outputs. Do not change my numbers. Check them for: (1) internal inconsistencies (tax rate, share count, units, sign errors), (2) terminal growth that is too high relative to WACC or long-run economic growth, (3) the percentage of enterprise value that comes from terminal value, (4) the EV/EBITDA multiple implied by my terminal value and whether it looks reasonable against the peer multiples I provide, (5) any step in the equity bridge that is missing. List problems in order of how much they would change the answer, and say what you would need to see to resolve each one.
>
> [Paste: forecast table, WACC inputs with sources and dates, terminal assumptions, net debt items, share count, peer multiples]

### Prompt 2: stress-test the peer set

> Here is my target company description and my proposed peer set with a one-line rationale for each. Act as a reviewer who disagrees with my choices. For each peer, tell me the strongest argument that it is not comparable (size, growth, margin, geography, business model). Then list types of companies I may have left out. Do not give financial figures; I will pull those from filings.
>
> [Paste: target description, peer list with rationale]

### Prompt 3: write the summary

> Using only the numbers below, write a 150-word valuation summary for a non-finance executive. State the enterprise value range, which method each end of the range comes from, the two assumptions that matter most, and one risk to the conclusion. Do not add any figures that are not in the data.
>
> [Paste: results table and sensitivity table]

After each answer, check that every figure in the output matches your source, and that the AI did not quietly "fix" an assumption you told it to leave alone.

## What mistakes make a valuation wrong?

- **Mixing enterprise value and equity value.** Applying an EV multiple and calling the result the share price skips the debt and cash bridge.
- **Mismatched cash flows and discount rate.** Unlevered FCF is discounted at WACC. Cash flow to equity is discounted at the cost of equity. Do not cross them.
- **Terminal growth above the economy's long-run growth.** A company cannot outgrow the economy forever.
- **Ignoring dilution.** Options, restricted stock and convertibles change the share count.
- **Stale or inconsistent comps.** Mixing last-twelve-months multiples with forward multiples, or including a peer in the middle of an acquisition, distorts the median.
- **Hockey-stick forecasts.** If growth jumps in year three with no driver behind it, the DCF is only reflecting hope.
- **One number, no range.** The model cannot be that precise, and readers who know it will trust you less.

## Can you build your own valuation tool instead of rebuilding a spreadsheet every time?

If you value companies regularly, you probably rebuild the same structure each time: a comps table, a DCF, a sensitivity grid, a place to store sources and assumptions. [Remy](https://goremy.ai) is a product agent that builds full-stack business apps from a plain-English description, so you could describe a small internal valuation tracker: a form for the assumptions, a comps table you maintain, a sensitivity grid that recalculates, and a log of who changed which input and when. Remy produces a spec, a deployed app with a database and sign-in, and the code sits in a git repository you own. It does not replace your judgment or a licensed market data source, so you would still supply the numbers. Pricing is on the site, with a 30-day free trial.

## Frequently Asked Questions

### What is the difference between enterprise value and equity value?

Enterprise value is the value of the whole operating business, regardless of how it is financed. Equity value is what is left for shareholders after you subtract net debt and other claims such as minority interest and preferred stock. DCF with unlevered cash flow and EV multiples both give you enterprise value first, so you must complete the bridge to get a share price.

### Which is better, DCF or comps?

Neither is better on its own. A DCF is grounded in the company's own cash flows but is sensitive to assumptions about the discount rate and the terminal value. Comps reflect what the market pays today but depend on how good your peer set is. Using both and explaining any gap between them is stronger than relying on either one.

### How do you value a company with no profits?

A DCF is hard when near-term cash flows are negative, because most of the value sits far in the future. Analysts usually lean on EV/Revenue comps, apply a multiple to a forward revenue estimate, and sometimes build a DCF with a longer forecast period that reaches a normal margin. State plainly that the uncertainty is much wider than for a profitable business.

### Can ChatGPT or another AI tool do a valuation for me?

It can help with reading filings, checking logic, and drafting the write-up. It should not be your source for financial data, because it may produce wrong or outdated figures with confidence. Paste in numbers you have verified, do the calculations in a spreadsheet, and use AI as a reviewer.

### How do I choose a discount rate?

Build WACC from sources you can cite: a risk-free rate such as a government bond yield, an equity risk premium from a published source, a beta from a market data provider, and the company's cost of debt. Record the date of each input, then show a sensitivity range around your result. Many companies and banks also have a house view on WACC, so check whether one exists before building your own.
