A comparable company table looks like the most objective document in a valuation pack. It is a grid of published figures, each one traceable to a filing or a screen, and the arithmetic between them is division. That appearance is the trap. Every comps table contains four judgements that no screen supplies — which peers were chosen, which period was used, what was added to or subtracted from the metric, and which average was taken — and a template that does not force those judgements into view produces a number nobody can check.
The template in this piece is built to expose them. It has ten fields, each of which exists because its absence breaks a specific thing a reader needs to do; it is applied to a five-peer set constructed for the arithmetic; and it ends in a range rather than a multiple, because the range is what the peer set actually supports.
Nothing in it is evidence about any real market. The peers, their market capitalisations, their debt and their earnings are constructed so that the additions can be followed, and a reader who wants a live example should build one from filings rather than read the figures here as anything but arithmetic.
The ten fields
A comps template is a row per peer and a set of columns per field. The temptation is to keep the columns a screen gives you — ticker, price, market capitalisation, a multiple — and that set omits the four things that decide the answer.
Table 1. The ten fields of a comparable company template, the reason each is in the template, and what breaks in the analysis when it is missing. Source: constructed for this piece; the third column is the diagnostic.
| Field | Why the template needs it | What breaks without it |
|---|---|---|
| Peer label and the reason it is in the set | Records the selection test applied, so the set can be re-run | The set becomes a list of familiar names, and a reader cannot tell whether it was chosen or remembered |
| Reporting period and currency of every figure | Fixes the date the multiple refers to | A multiple that mixes a December year-end with a June one is a comparison of two different economies |
| Market capitalisation, with the share count and the date | Makes the equity value reproducible | A market capitalisation without a share count cannot be checked, and a stale one misprices the peer |
| Debt: borrowings, lease liabilities, pension deficit, minority interests | Enterprise value is a claim on all capital, not just equity | Leases left out understate enterprise value, and the multiple with it, in exactly the peers that lease most |
| Cash and cash equivalents | Net debt is debt less cash, not debt | A peer holding net cash shows a debt figure that reverses the sign of its leverage |
| The reconciliation line: market capitalisation plus net debt equals enterprise value | Forces the two components to add to the figure used | A typed enterprise value with components that do not reconcile is the most common comps error |
| The metric, with its adjustments named | EBITDA reported, adjusted and normalised are three different numbers | An adjusted metric against a reported metric flatters whichever company adjusts most |
| The multiple, calculated in the cell from the two figures beside it | Makes the ratio auditable | A pasted multiple cannot be traced, and one transposition moves the median |
| The basis: trailing twelve months, forward, or both | States what the multiple is a multiple of | Mixing bases across peers compares a year that happened with a year that has not |
| The exclusion log: peers considered and dropped, with the reason | Makes the selection arguable rather than invisible | Dropping a peer changes the median in the direction the analyst wanted, and nothing records it |
The last field is the one that separates a template from a spreadsheet. A peer set of five implies a longer list that was considered; the ones dropped, and the reason — no comparable metric, a different business model, a controlling shareholder, a period that cannot be aligned — are part of the analysis, and a reviewer’s first question is always why the sixth peer is not there.
The peer set is the argument
The multiple is division. The peer set is the valuation, and five tests decide whether a company belongs in it.
The first is what the business does and how it earns. A peer should earn its revenue the same way the subject does: a subscription business is not comparable to a project business at the same margin, because the revenue is contracted differently and the multiple pays for the difference. Sector classifications are too coarse to settle this; the segment note in the filing, or its absence, is the evidence.
The second is size. Small companies trade at different multiples from large ones, for reasons of liquidity, index membership and the cost of capital, and a peer set spanning a factor of sixty in enterprise value is not a set. Where the subject is small, the peers should be small, or the multiple should carry a stated discount that the pack discloses.
The third is growth. An eight-times multiple on a business growing at 4 per cent and an eight-times multiple on one growing at 20 per cent are not the same price; if the set’s growth rates scatter, the multiple is measuring the scatter. A growth-adjusted multiple — enterprise value to EBITDA divided by expected growth, or a regression of multiples against growth — is the correction, and it is worth using only where the growth forecasts for the peers are on a comparable basis.
The fourth is leverage and the fifth is jurisdiction. Leverage matters because the multiple is on enterprise value and the enterprise value includes the debt: two peers with the same operations and different capital structures should share an enterprise multiple, and if they do not, something else is in the number. Jurisdiction matters because tax, accounting and market structure move both the metric and the multiple, and a peer set drawn across three regimes is a set with three accounting conventions in it.
The discipline that makes all five visible is one sentence per peer. A row in the exclusion log that says why a company was left out is worth more than a paragraph defending the ones that were kept, because it is the only part of the exercise a reader can independently check.
The table, computed
The set below is constructed. Each peer’s enterprise value is assembled from a market capitalisation and a net debt figure, the multiple is computed from that enterprise value, and the arithmetic reconciles in every row.
Table 2. A five-peer comparable company table, constructed for this piece. Amounts in constructed currency units; enterprise value is market capitalisation plus net debt, where a bracketed net debt figure is net cash; the multiple is enterprise value divided by EBITDA. Source: constructed; no real company is described.
| Peer | Market capitalisation | Net debt | Enterprise value | EBITDA | EV / EBITDA |
|---|---|---|---|---|---|
| Peer A | 1,050.00 | 150.00 | 1,200.00 | 150.00 | 8.00× |
| Peer B | 960.00 | (60.00) | 900.00 | 120.00 | 7.50× |
| Peer C | 2,100.00 | 300.00 | 2,400.00 | 300.00 | 8.00× |
| Peer D | 660.00 | (60.00) | 600.00 | 110.00 | 5.45× |
| Peer E | 1,700.00 | 150.00 | 1,850.00 | 250.00 | 7.40× |
| Set | median 7.50× |
Read the two net cash peers first, because they are where the reconciliation earns its place. Peer B has a market capitalisation of 960.00 and net cash of 60.00, so its enterprise value is 900.00, which is 7.50 times its EBITDA of 120.00. Computed on market capitalisation instead, Peer B would appear at 8.00 times — half a turn higher — and the median of the set would move. Peer D is the same story from the other end: 660.00 of market capitalisation against 60.00 of net cash is 600.00 of enterprise value and 5.45 times, where the incorrect market-capitalisation multiple would be 6.00 times.
That is the argument for the reconciliation column in one line. Two peers out of five hold net cash, and both appear dearer than they are if the multiple is computed on equity value: 8.00 against 7.50, and 6.00 against 5.45. In a set where the subject company carries debt, the bias runs in one direction: it makes every net-cash peer look dearer, lifts the median, and overstates the enterprise value the median produces for the subject.
The set’s multiples, in order, are 5.45, 7.40, 7.50, 8.00 and 8.00. The median is 7.50 and the mean is 7.27, and the four-tenths of a turn between them is the whole question of which statistic to publish, taken up below.
From a multiple to a price
Applying the set to a subject company requires the subject’s own figures and its own bridge. The construction here is the shortest honest one: the subject’s EBITDA is multiplied by the chosen multiple to give an enterprise value, its net debt is deducted, and the equity value is divided by the diluted share count.
Table 3. The multiple applied to a constructed subject company — EBITDA 180.00, net debt 150.00, 100.00 fully diluted shares — at the set’s low, median, mean and high multiples. Constructed for this piece. Source: Table 2 for the multiples; the subject’s figures are constructed.
| Basis | Multiple | Enterprise value | Less net debt | Equity value | Per share |
|---|---|---|---|---|---|
| Low — Peer D | 5.45× | 981.82 | (150.00) | 831.82 | 8.32 |
| Median | 7.50× | 1,350.00 | (150.00) | 1,200.00 | 12.00 |
| Mean | 7.27× | 1,308.76 | (150.00) | 1,158.76 | 11.59 |
| High — Peers A and C | 8.00× | 1,440.00 | (150.00) | 1,290.00 | 12.90 |
Four observations follow from that table, and each is a decision the template forces rather than a fact it reports.
The range is the output. Between the low peer and the high peers the price per share moves from 8.32 to 12.90, which is 55.0 per cent of the low figure. A pack that publishes 12.00 from the median and describes the peer set as supportive is publishing one point from a range that spans more than half its own value.
The median is the defensible statistic on a set of five, and the mean is the one that misleads. Peer D’s 5.45 times is not a comparable price for the subject; it is a peer priced as if it were worth less, and averaging it into the set drags the mean to 7.27 and the answer to 1,308.76, which is 41.24 below the median and no more accurate for it. The median ignores the size of the outliers; the mean pretends they are information.
The choice of peers moves the multiple as much as the arithmetic does. Drop Peer B and Peer D — the two with net cash, whose multiples both sit below the median — and the remaining three have a median of 8.00 times, giving 1,440.00 of enterprise value and 12.90 a share. That is 6.7 per cent on the multiple from two exclusions, and the exclusion log in the template is the only place the change would be visible.
The bridge is where the multiple ends. The enterprise value at the median is 1,350.00 and the equity value is 1,200.00, which is a deduction of 11.1 per cent for net debt. A comps analysis that stops at enterprise value has not answered the question the client asked, which is the price of a share.
Where the multiple stops
The honest limitation is that a comparable company multiple is a ratio of two numbers about other companies, and its authority is borrowed from a set that the analyst chose.
Four limits follow.
The peers are not comparable, and cannot be made so. Every additional test of comparability — business model, size, growth, leverage, jurisdiction — shrinks the set, and a set small enough to be genuinely comparable is too small to have a median. Five is already generous; three is honest; one is a story. The template makes the trade visible, and it does not resolve it.
The multiple is a price, not a value. It says what the market paid for similar claims on similar earnings on a stated date. Where the market’s view of the sector is moving — and it always is — the multiple imports that movement into the valuation of a private company that has no market at all.
No control premium or minority discount is measured. The peers are priced as minority claims on listed companies, and the subject may be a controlling stake in a private one. The premium a buyer pays for control, and the discount a minority holder accepts for the absence of a market, are conventions applied on top of the multiple, and a pack that applies one should show the figure without it.
The metric is not standard across the set. EBITDA as reported, as adjusted by the company and as normalised by the analyst are three numbers per peer, and the choice moves every multiple in the table. Where a peer’s adjustments are large, the multiple computed on its adjusted metric is measuring the adjustment.
What the table costs
The compute is published and the drafting is measured, and the split between them is the useful part of the figure.
The multiples operation is published at $0.006125 on Base, and a peer set of five needs five calls for the multiples and five more if each peer’s enterprise value is assembled separately: ten calls, $0.061250. That price is published rather than measured, because the analyst services are pending review and return no 402 quote today.
The drafting is 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 carrying a filing; Qwen quoted $0.00875 at both lengths. The reading of five peer filings is long-prompt work, and there Qwen’s flat $0.00875 is 8.4 per cent below DeepSeek’s long-prompt figure, while the short work — the screening notes, the reconciliation checks, the table’s caption and the selection rationale — is 20 per cent cheaper on DeepSeek.
One illustrative build: three short calls to define the selection tests and the metric’s definition, at $0.00700, is $0.02100; five long calls to read five peer filings, at Qwen’s $0.00875, is $0.04375; five short calls to compute the multiples and run the reconciliations is $0.03500; four short calls to write the table’s caption, the source block and the selection rationale is $0.02800; three short calls to rebuild the table after the reviewer challenges two peers is $0.02100. Twenty calls, $0.14875, or $0.00744 a call.
Three comparisons bound it. Reading the filings on DeepSeek’s long-prompt rate instead of Qwen’s takes the table to $0.15275, 2.7 per cent more. At the metered ceiling of $0.00955 the twenty calls cost $0.19100, 28.4 per cent above the measured total. And at the $0.00875 that circulates for this service they cost $0.17500, 17.6 per cent more than measured, which is the correction that matters most in a table of this size: the $0.02625 difference is 43 per cent of the engine calls the whole analysis runs on.
That is the arithmetic worth keeping. Ten engine calls at the published rate are $0.061250; the drafting around them is $0.14875 measured; the document parts are unpriced. A comps table is a document about other documents, and most of its cost is in the reading.
The decision
Two things decide whether a comps table can be relied on, and neither is the multiple.
Publish the exclusion log with the table. Every peer considered and dropped, with one sentence on why, on the same page as the multiple. It is the only part of the analysis that lets a reader test the selection rather than accept it, and it takes a line per peer. Where the log cannot be produced because the set was assembled from memory, the table is a recollection with arithmetic attached.
Then report the range and the multiple that produced it, and reconcile the enterprise value to equity in the table itself. A single figure from a five-peer median is a point on a range that in this construction spans 55.0 per cent; the reader’s decision depends on where in that range the subject sits, and the pack’s job is to show the width rather than to select from it.
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 runs the multiples operation and reconciles the set
- 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 the prompt bands
- The cost context: x402.review/2026/10/31/inference-endpoints-compared/ — the two endpoints against each other
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 and did not step between the two lengths. The multiples operation at $0.006125 is the published catalogue price 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. Every peer, market capitalisation, debt figure, EBITDA figure, multiple, the subject company and its share count are constructed for this piece and are labelled illustrative where they appear; no real company, transaction or market multiple is described, and none of the figures is evidence about any sector. Totals and percentages are computed from the figures shown, rounded to the cent and to one decimal place respectively, so a recomputation from unrounded values may differ in the last digit. Nothing here is investment, legal or tax advice, a valuation of any company, 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: comparables, multiples, valuation, template
