A private company has no price. It has a set of accounts, a cap table, a plan, and someone who wants to know what it is worth — for a transaction, a share transfer, a financing, a tax filing, a family settlement, or two shareholders who have stopped agreeing. Each of those purposes produces a different number from the same company, and the first task in any valuation is to establish which question is being answered.
What follows is therefore not a lookup. A valuation is an argument: a method, the inputs the method requires, the arithmetic that carries the inputs to a figure, and the bridge that turns an enterprise value into a price per share of the thing actually being sold. Its output is a range with the assumptions named and a table showing what moves the answer, because a single figure resting on three contested inputs claims a precision the inputs cannot support.
This piece covers the four ways a private company differs from a listed one and what each difference costs in arithmetic, the three routes and the question each answers, an illustrative pack worked to the cent with its DCF and its bridge, the measured cost of building one at published rates, and the decision that separates a valuation from a number with no argument behind it.
Why a private company is a different problem
Four differences carry most of the weight, and each has an arithmetic consequence rather than a stylistic one.
There is no price to check against. A listed valuation can be set beside a market capitalisation, and the gap between them is information: either the model is wrong or the market is, and the reconciliation is the work. A private valuation has no such test. The only available checks are internal — a second route to the same figure, and a bridge that reconciles the components of the answer — which is why a pack that offers one method and no cross-check is not merely thin. It is unverifiable by construction.
The cap table is a set of contracts, not a share count. A private company’s equity is usually shares plus options, plus preference with a liquidation preference and sometimes a participation right, plus shareholder loans that behave like debt, plus earn-outs contingent on events after the valuation date. The enterprise value is the value of the business; the price per share is what remains after every claim between the enterprise and the ordinary shareholder has been deducted in the right order. Getting the DCF right and the bridge wrong produces a headline figure that is defensible and a per-share price that is not.
The statements are management accounts. They are unaudited, they may not be on the same basis year to year, related-party transactions are often material and rarely disclosed as such, and the owner’s remuneration is frequently a distribution in disguise. Every input taken from them is an adjusted input, and the adjustment has to be written down, because the same revenue figure adjusted two ways supports two valuations.
The purpose fixes the standard. Value to a controlling buyer, value to a minority holder with no market for the shares, value for a tax authority that wants fair market value, value in a dispute where the question is what a shareholder would have received — these are not the same exercise, and the assumptions that fit one misfit another. A valuation that does not state its purpose cannot be tested against its own premises.
Three routes and what each is for
Every valuation method is one of three approaches: income, market, or cost. The choice is a question of what evidence exists, not a matter of preference, and a usable pack runs at least two of them.
Table 1. The three routes to a private-company valuation, the input each requires, the question it answers and the way it fails. Source: constructed for this piece; the failure column is the diagnostic, not a ranking.
| Route | Input it requires | Question it answers | How it fails |
|---|---|---|---|
| Income — discounted cash flow | A forecast of free cash flow, a discount rate, and a terminal assumption | What is the business worth if the plan happens and the discount rate is right | The terminal value carries most of the answer, so a small change in two assumptions moves the figure more than any year of the forecast |
| Income — capitalised earnings | A maintainable earnings figure and a capitalisation rate | What is the business worth if this year’s earnings continue unchanged | The maintainable figure is a judgement, and growth is either ignored or smuggled into the rate |
| Market — comparable companies | A peer set with reconciled enterprise values and a common multiple | What are similar businesses being priced at today, on a public market | The peers are rarely comparable, their multiples embed their own growth and leverage, and a private company has no observable price at all |
| Market — transaction multiples | Completed deals in the same sector with disclosed terms | What have buyers paid for businesses like this | Deals are few, terms are partial, and the premium a buyer paid for control is inside the multiple without being separable from it |
| Cost — net asset value | A balance sheet at fair value, with intangibles considered | What would be left if the business stopped trading and the assets were realised | It values what is on the balance sheet, and for a service business that is close to nothing |
The income route is the anchor for a going concern with a forecast, and it is the only one that forces the assumptions into the open where a reader can disagree with them one at a time. The market route is the discipline on the income route: a DCF can be made to produce any figure by moving the discount rate a point and the terminal growth half a point, and a multiple computed from five reconciled peers is a narrower target than that. The cost route is the floor, useful in distress, in property-heavy businesses, and as a test of whether the other two have produced something physically impossible.
The mistake that recurs is using the market route with unreconciled peers. A multiple is enterprise value over a metric, and the enterprise value is market capitalisation plus net debt — not market capitalisation. A peer quoted at 8.0 times on enterprise value and the same peer quoted at 8.0 times on market capitalisation are different claims, and the difference is the peer’s leverage. For a peer holding net cash the enterprise value sits below the market capitalisation, so the multiple computed on market capitalisation is the higher of the two: it overstates the peer’s enterprise multiple, and a set built that way raises its own median and inflates every value derived from it.
The arithmetic of one income route
The income route in its fullest form is a discounted cash flow: an explicit forecast period, a terminal value beyond it, both discounted to today, summed into an enterprise value. The structure below is not a forecast of any company. It is a five-year construction designed so that the arithmetic can be checked in one pass.
Table 2. An illustrative five-year discounted cash flow. Free cash flow to the firm in constructed currency units, discounted at 12.00 per cent, with terminal growth of 2.00 per cent. Components are rounded to the cent and the totals are the sums of the rounded components. Source: constructed for this piece; it describes no company.
| Year | FCFF | Discount factor at 12.00% | Present value |
|---|---|---|---|
| 1 | 100.00 | 0.892857 | 89.29 |
| 2 | 106.00 | 0.797194 | 84.50 |
| 3 | 112.36 | 0.711780 | 79.98 |
| 4 | 119.10 | 0.635518 | 75.69 |
| 5 | 126.25 | 0.567427 | 71.64 |
| Sum of the explicit period | 401.10 | ||
| Terminal value at year 5 — 126.25 × 1.02 ÷ 0.10 | 0.567427 | 730.70 | |
| Enterprise value | 1,131.80 |
Two figures in that table are worth reading twice, because they decide whether the pack is an argument or a decoration.
The first is the terminal value’s share. Of the 1,131.80, 730.70 arrives after the forecast period, which is 64.6 per cent of the answer. That is normal for a growing business and it is also the honest statement of what the valuation is: a bet on a perpetuity that nobody forecast, priced by a formula that assumes growth continues at 2.00 per cent for ever. A pack that presents the five forecast years in detail and the terminal value in one line is presenting 35 per cent of its work and 65 per cent of its conclusion with unequal care.
The second is the discount rate’s reach. The rate is not a fact about the company; it is an assumption about the return an investor requires, built from a risk-free rate, an equity risk premium, a beta and a size or liquidity premium where the valuer chooses to include one. Each of those is contestable, and the derived figure inherits all of it. Two valuers with the same forecast and rates of 11.00 and 13.00 per cent produce 1,260.79 and 1,026.35 on the same cash flows — 22.8 per cent apart — and neither is wrong on the arithmetic.
The forecast itself has to hold together as a set of statements, not a column of hopeful numbers. The test that catches most errors is that free cash flow must be rebuilt from its components rather than typed in: earnings before interest and tax, less tax at the statutory rate on that figure, plus depreciation and amortisation, less capital expenditure, less the increase in working capital. A forecast in which capital expenditure sits below depreciation for five years is forecasting a business that shrinks while it grows, and that shows up in the components and nowhere else.
The bridge that turns a value into a price
An enterprise value is not a price. It is the value of the operating business, before the claims that sit between the business and the ordinary shareholder. The bridge applies them in order, and the order matters.
Table 3. The illustrative bridge from the enterprise value in Table 2 to a price per share, with a cap table of 100.00 fully diluted shares. Constructed for this piece. Source: Table 2 for the enterprise value; the remaining lines are constructed.
| Line | Amount |
|---|---|
| Enterprise value (Table 2) | 1,131.80 |
| Less debt, including shareholder loans treated as debt | (210.00) |
| Add cash and cash equivalents | 60.00 |
| Equity value | 981.80 |
| Divide by fully diluted shares — 100.00, after the option pool | 9.8180 |
| Price per share | 9.82 |
Four conventions inside that table decide more than the arithmetic does, and each is a place where a pack either states its position or hides it.
Net debt is debt less cash, not debt alone, and the cash figure must be the cash the business needs to trade rather than the cash in the account on the valuation date. Where the company runs a minimum operating balance, the excess is the only part that belongs in the bridge.
Preference shares are deducted at the higher of their liquidation preference and their conversion value, because the holder takes the better of the two and the ordinary shareholder bears the difference. Where the preference participates alongside the ordinary after the preference is redeemed, the bridge becomes a waterfall rather than a subtraction, and a pack that subtracts a participating preference once has overstated the ordinary share’s price.
The option pool is handled by the treasury stock method: options are assumed exercised, the company receives the strike proceeds, and the shares issued net of the proceeds repurchased are added to the diluted count. The shortcut of dividing by the undiluted count is the most common bridge error in private-company work, and it overstates the price per share by roughly the pool’s share of the total.
Contingent consideration and earn-outs are deducted at their expected value if the valuation date is after the obligation was created, and they are not deducted if the buyer, not the company, will pay them. The distinction depends on the transaction documents, which is why the bridge cannot be completed from the accounts alone.
What the answer is worth
The honest limitation is structural rather than technical: a private-company valuation cannot be validated against a price, so it is only ever as good as its stated assumptions and its internal reconciliation.
Three consequences follow.
The range is the answer, not a hedge around it. Two routes with defensible inputs will not agree, and the disagreement is the useful output: if the income route says 1,131.80 and a reconciled peer set says something materially below it, the pack’s job is to say which assumption is doing the work, not to average the two into a figure neither method supports.
No adjustment for lack of marketability or for control is a measurement. The discounts and premiums used in practice — for illiquidity, for minority status, for control — are conventions drawn from studies of other transactions, and a pack that applies one should say so and show the figure without it. Their size varies widely across practitioners, and a valuation that moves 20 or 30 per cent on a convention is reporting the convention.
A valuation is not advice, and no recommendation to transact follows from it. It states what a set of assumptions produces. Whether a price is acceptable, whether a shareholder disposes of a holding, and whether a transaction should proceed are decisions for the parties, taken with the tax, legal and financing consequences in front of them, which are outside the valuation entirely.
What one pack costs
The prose around the arithmetic is bought by the call, and the arithmetic itself is bought from the compute services. At the published $0.006125 an operation, a pack that runs statements normalisation, multiples, a DCF and a cross-verification is four engine calls and $0.02450, or sixteen calls and $0.09800 if it is rebuilt twice after review. Those prices are published rather than measured: the analyst services are pending review and return no 402 quote, so the figure is the catalogue price for a capability that does not answer today.
The model calls are measured. On Base, deepseek-v4-flash quoted $0.00700 for a short prompt at max_tokens: 1000 on 31 October 2026, and $0.00955 for a prompt of about 3,000 tokens — the band a filing falls into. Qwen quoted $0.00875 for the same requests at both lengths and did not step. The consequence for a valuation pack is small but real: the reading of a data room is long-prompt work, and on that work Qwen is 8.4 per cent cheaper than DeepSeek’s long-prompt figure, while the drafting of the note is short-prompt work where DeepSeek is 20 per cent cheaper than Qwen.
One illustrative pack, in four stages: four long-prompt calls to read and normalise the statements, at Qwen’s $0.00875, is $0.03500; three short calls to structure the model, at $0.00700, is $0.02100; five short calls to draft the note’s sections is $0.03500; four short calls to run the cross-check and rewrite what the client’s answers change is $0.02800. Sixteen calls, $0.11900, or $0.00744 a call.
Four comparisons bound it. Reading the statements on DeepSeek’s long-prompt rate instead takes the pack to $0.12220, 2.7 per cent more. Pricing all sixteen calls at the metered ceiling of $0.00955 takes it to $0.15280, 28.4 per cent more. Pricing all sixteen at the $0.00875 that circulates for this service — the figure the catalogue’s tier row carries, and Qwen’s actual rate — takes it to $0.14000, 17.6 per cent above the measured total. And the 0.50 USDC opening balance is 71 calls at $0.00700, which is four packs of this size with change.
The reading stage is 29.4 per cent of that total and the drafting 71 per cent, which is the opposite of what the number looks like it should be. Reading a data room is expensive because the prompt carries the room.
The decision
Two things decide whether a valuation pack is usable, and neither is the choice of method.
State the purpose and the standard of value in the first line, before any arithmetic. A controlling-interest valuation for a transaction, a minority-interest valuation for a share transfer and a fair-market-value valuation for a tax authority are three different exercises; a pack that names its purpose can be tested against it, and one that does not cannot. If the purpose cannot be written in a sentence, the valuation is not ready to start.
Then build the answer from two routes and publish the bridge. The second route is the only check a private company allows, and the bridge is where the figure becomes a price. Write the bridge line by line with the cap table beside it, deduct the preference at the higher of its two values, dilute for the option pool, and show the discounts you applied and the figures without them. A reader who can re-derive the per-share price from the enterprise value has everything the exercise can honestly deliver.
Where this goes
- The pillar: x402.review/2026/10/31/finance-and-markets/ — three engines, four jobs, one price
- The role: studio.x402.press/roles/financial_analyst — the agent that builds the pack, with its parts and its refusals
- The service it buys: x402.press/services/deepseek/ — the canonical price, per model and per chain
- The second model: x402.press/services/qwen/ — the flat rate that does not step with prompt length
- The review: x402.review/2026/10/31/deepseek-review/ — the measured handshake and what it does not promise
- The cost context: x402.review/2026/10/31/inference-endpoints-compared/ — the two endpoints against each other, by prompt length
Method note
The model prices were read from the live 402 challenge on Base on 31 October 2026 at 08:50 UTC, outside DeepSeek’s weekday peak windows of 01:00–04:00 and 06:00–10:00 UTC, with one short prompt and one of about 3,000 tokens at max_tokens: 1000; Qwen was read the same way. The engine price of $0.006125 is the published catalogue figure for the analyst services, which are pending review and return no 402 quote, so it is published rather than measured. Call counts are planning assumptions stated in the cost section. The company, its cash flows, discount rate, cap table, debt and share count are constructed for this piece and are labelled illustrative wherever they appear; no client, comparable, transaction or market figure is described, because none was measured. Table totals are the sums of the rounded components, so a reader recomputing from unrounded values may differ by a cent. Nothing here is a valuation of any company, investment, legal or tax advice, or a recommendation to buy, sell or hold anything.
x402.review is written and edited by N.P. Vincent. The publication is connected to x402.press, which sells the services reviewed here; that interest is disclosed on every piece that touches a service, and it is stated here rather than left to inference.
Filed under Jobs · Tags: valuation, private-company, dcf, comparables
