Comps Sanity Checklist for M&A Advisers

One person builds, another reviews and the partner decides what the page claims

A comps page gets read for one number and attacked on the detail behind it. The seller remembers the top of the range. The buyer's adviser looks for the row that does not hold, because one bad row gives them a reason to discount the rest.

The person who built a sheet cannot review it. They read what each cell was meant to say, not what it says. Give the review to someone else, with the backup open and the filings to hand.

1While building
Analyst
Writes the screen down, keeps the backup tab current and notes every judgement call as it is made.
2Before the partner sees it
Associate or director
Runs every check below with the filings open, and marks which rows were traced.
3Before the pitch or valuation note
Partner
Tests the range against a second method and the seller's number, and decides what the page claims.
4At each round of a live process
Reviewer again
Reprices, rolls LTM and re-marks peers. A page built for the pitch is not valid at final offers.

A sheet can look clean and still be wrong by most of a turn

Every error below is one we would expect a competent buyer's adviser to find in a first pass, and none of them shows on the final page. Together they lift the median by 0.8x. On Project Harbour's £12.0m of EBITDA, that is £9.6m of enterprise value the seller will expect and no bidder will pay.

Exhibit 1

Five errors lift the median from 9.6x to 10.4x, worth £9.6m on £12.0m of EBITDA

Trading comparables for Project Harbour, as built and as corrected. £m unless stated. Priced at close on 30 September 2026.

Company	As built	Corrected
Peer A	9.6x	9.6x
Peer B1	11.5x	10.0x
Peer C2	8.8x	9.2x
Peer D3	11.2x	10.0x
Peer E4	13.5x	n.m. (2)
Peer F5	7.9x	8.6x
Mean	10.4x	9.5x
Median	10.4x	9.6x
Rows counted6	6	5

Company	Basis	LTM to	Equity value	Claims (1)	EV	LTM EBITDA	EV / EBITDA as built	EV / EBITDA corrected
Peer A	IFRS 16	Jun-26	420.0	60.0	480.0	50.0	9.6x	9.6x
Peer B	IFRS 16	1Mar-26	310.0	35.0	345.0	30.0	11.5x	10.0x
Peer C	IFRS 16	Jun-26	2185.0	26.0	211.0	24.0	8.8x	9.2x
Peer D	3US GAAP	Jun-26	520.0	40.0	560.0	50.0	11.2x	10.0x
4Peer E	IFRS 16	Jun-26	690.0	55.0	745.0	55.0	13.5x	n.m. (2)
Peer F	IFRS 16	Jun-26	140.0	518.0	158.0	20.0	7.9x	8.6x
Mean							10.4x	9.5x
Median							10.4x	9.6x
6Rows counted							6	5

1
Stale LTM
Peer B runs to March while every other row runs to June. Rebuilt, EBITDA is £34.5m and the multiple falls from 11.5x to 10.0x.
2
Basic shares
Peer C uses a basic share count. Diluted by the treasury stock method, equity is £194.0m, not £185.0m.
3
Mixed lease basis
Peer D reports under US GAAP, so rent sits above EBITDA. On the IFRS 16 basis of the other rows it is 10.0x, not 11.2x.
4
Peer under offer
Peer E's price carries a bid premium. It tells you about that bid, not the sector, so it comes off the median.
5
Pension deficit left out
Peer F's bridge misses a £14.0m deficit. Added back, the multiple rises from 7.9x to 8.6x.
6
No count on the page
The sheet showed a mean and a median with no row count, so a reader could not see that one row was driving them.

(1) Net borrowings, lease liabilities, pensions and minorities, less associates. Equity value is share price times shares.

(2) Not meaningful: excluded under a rule set before the results were seen. Here, any company under offer on the pricing date.

Source: Sample data. Project Harbour and every company, customer and vendor shown are invented. Multiples are computed from the figures shown.

The bridge is the first place a buyer's adviser looks

Enterprise value is what it costs to own the whole business. The bridge is the list of everything that ranks ahead of the shareholders on the way to equity value, so it is where a sheet is most often incomplete. Build it line by line for every peer, and keep it where a reviewer can see it.

The same bridge comes back in the sale itself. Each debt-like item a bidder finds is deducted from the equity price, so an item missed here is a price chip later.

Exhibit 2

Peer C: five items take £220.0m of enterprise value to £194.0m of equity value

Enterprise value to equity value for Peer C, corrected. £m. Balance sheet at 30 June 2026. Axis starts at £170.0m.

220.0

Enterprise value

(14.0)

Net borrowings

(9.0)

Lease liabilities

(4.0)

Pension deficit

(2.0)

Minority interests

3.0

Stake in associate

194.0

Equity value, diluted

How the equity value is built

Share price, £	2.00
Basic shares, m	92.5
Market cap on basic shares, £m	185.0
Options, 9.0m at £1.00: net new shares by treasury stock method, m	4.5
Diluted shares, m	97.0
Equity value on diluted shares, £m	194.0

Note: The net claims of £26.0m in the comps table are the five items in the bridge: 14.0 + 9.0 + 4.0 + 2.0 less 3.0.

Source: Sample data. Project Harbour and every company, customer and vendor shown are invented.

Bridge item	Where the reviewer looks	How it shows up in a buyer's adviser's review
Diluted shares	Share capital note and option table in the annual report, plus any awards since.	They rerun the treasury stock method. If your count is basic, every equity value in the table becomes suspect.
Net borrowings	Latest balance sheet and debt note, after any deal or raise since that date.	They ask why acquisition debt is missing when the acquired earnings are in the numbers.
Lease liabilities	Lease note, and which accounting framework the company reports under.	They take the liabilities out or put them in, then reset EBITDA to match. A table that mixes both falls apart in one step.
Pension deficit	Retirement benefit note. Check whether the figure is before or after tax.	They add it to enterprise value as debt-like, then raise the same item against your target's equity price.
Minority interests	Balance sheet and subsidiary note. Book value can understate a valuable stake.	They ask whether EBITDA is fully consolidated. If it is and minorities are missing, your multiple reads low.
Associates	Equity-accounted line, and whether those earnings sit in EBITDA.	They deduct the stake if its earnings are outside EBITDA, and flag any double count if they are not.

One company, two multiples, depending on the accounting

Lease accounting now splits peer tables in a way that did not matter a decade ago. Rent sits above EBITDA under some frameworks and below it under others, so the same company shows different earnings and a different enterprise value. Pick one basis, restate every row to it and print the basis on the page.

Private targets add a second trap. UK companies on FRS 102 move most leases on balance sheet for periods beginning on or after 1 January 2026, so a target's EBITDA can step up with no change in the business. Its history then needs restating before any multiple is applied to it.

Exhibit 3

The same business reads 11.2x or 10.0x depending on where rent sits

Peer D on two lease bases, and how each framework treats rent. £m unless stated. LTM to June 2026.

As reported, US GAAP

Enterprise value: 560.0
LTM EBITDA: 50.0

11.2x

Restated to IFRS 16 basis

Enterprise value, adding lease liabilities of 60.0: 620.0
LTM EBITDA, adding back lease cost of 12.0: 62.0

10.0x

Framework	Who uses it	Rent on an operating lease	Liability on balance sheet
IFRS 16	Listed UK and Canadian peers	Below EBITDA	On
US GAAP	US peers	Above EBITDA	On
FRS 102 from 2026 (1)	UK private companies	Below EBITDA, for most leases	On
FRS 102 before 2026	UK private company history	Above EBITDA, for operating leases	Off
ASPE	Canadian private companies	Above EBITDA, for operating leases	Off

(1) Periods beginning on or after 1 January 2026. Low-value and short-term exemptions apply.

Source: Sample data. Project Harbour and every company, customer and vendor shown are invented. Treatment by framework: FRC, IFRS Foundation and FRAS Canada, linked under Sources.

Errors matter most when they hide a gap to the seller's number

A wide, inflated range is dangerous for a specific reason: it makes the seller's number look reachable. Once the rows are clean, the gap appears, and it is far cheaper to discuss before the mandate is signed than after the first offers land.

Exhibit 4

The seller's £140.0m sits above every corrected method, and the as-built comps hid it

Implied enterprise value of Project Harbour on LTM adjusted EBITDA of £12.0m. £m. Priced at close on 30 September 2026.

Trading comps, as built

94.8162.5

7.9x to 13.5x

Trading comps, corrected

103.2120.0

8.6x to 10.0x

Precedent transactions

102.0132.0

8.5x to 11.0x

DCF

105.0128.0

8.8x to 10.7x

Financial buyer

96.0114.0

8.0x to 9.5x

Seller's number, £140.0m: £8.0m above the top of the corrected methods. Fixing the comps moved the median by 0.8x, or £9.6m of value.

Note: Multiples for precedents and the financial buyer view are sample inputs. DCF is a sample range, shown as its implied multiple.

Source: Sample data. Project Harbour and every company, customer and vendor shown are invented.

29 checks, in the order that saves the most rework

The early groups change who is in the table, so run them first. There is no point polishing a row that is about to come out.

Peer set

Most arguments about a multiple are really arguments about who is in the table.

The screen is written down.Sector, business model, size band, geography and exclusions sit in a footnote or a backup tab. Any name removed by hand has its reason recorded.
Why it mattersA written screen lets a reviewer challenge the logic rather than the names, and lets the team rerun it when the market moves.
What going wrong looks likeThe buyer's adviser asks why a close competitor is missing. Nobody remembers, and the page starts to look selected.
Each peer earns its place on how it makes money.Recurring against project revenue, asset-light against asset-heavy, and distribution against manufacture matter more than a shared sector label.
Why it mattersMultiples price the durability of margin and the need for capital, and both come from the business model, not the sector code.
What going wrong looks likeA contract manufacturer sits beside a services firm. The range widens and the top end is easy to dismiss.
Size is dealt with, not ignored.Listed peers are usually far larger than the target. The page says so, and says how the range was adjusted, or why it was not.
Why it mattersListed scale brings liquidity and diversification that a private company does not have.
What going wrong looks likeA seller anchors on a listed median, the first offers land well below it, and the gap becomes your problem.
Awkward peers are marked.A company under offer, in distress or thinly traded has a price that reflects its situation rather than its business. Mark it or remove it.
Why it mattersThat price tells you about one situation, not about the sector.
What going wrong looks likePeer E in Exhibit 1 is under offer at 13.5x. The other side strips it out in one line and your range collapses.

Dates and currency

One pricing date for every row.All market caps use the same close, and that date is printed on the page.
Why it mattersA multiple is a snapshot. Mixed dates turn a market move into a false difference between companies.
What going wrong looks likeOne row was refreshed for a later meeting. A sharp reader spots it from a share price they happen to know.
LTM really is the last twelve months.Rebuild it from the filings: last full year, plus the current stub, less the prior-year stub. A database field can lag a recent filing.
Why it mattersBuyers price the latest twelve months, and a stale LTM on a growing peer inflates its multiple.
What going wrong looks likePeer B in Exhibit 1 shows 11.5x on LTM to March. Rebuilt to June, it is 10.0x.
Year ends are calendarised.Peers with different year ends are brought to a common period before forward multiples are compared.
Why it mattersForward multiples only compare the same months of expected earnings once the periods align.
What going wrong looks likeA March year-end peer is set against a December one, and the faster grower looks cheap for no real reason.
Forward numbers name their source and date.Consensus has a provider and an as-of date. Thin coverage is flagged, because one broker is not a consensus.
Why it mattersAn unnamed forecast cannot be defended, and consensus goes stale fast around results.
What going wrong looks likeThe page cites consensus from before a profit warning. The other side has the current figure.
Each multiple is in one currency.Value and earnings are in the same currency before dividing. Watch companies that trade in one currency and report in another.
Why it mattersSterling value over dollar earnings gives something that looks like a multiple and is not one.
What going wrong looks likeA UK-listed peer that reports in dollars is out by the exchange rate, and it survives because the number looks plausible.

Enterprise value bridge

Shares are diluted, not basic.Options and warrants are counted by the treasury stock method at the pricing-date price. In-the-money convertibles sit in shares or in debt, never both. Treasury shares come out.
Why it mattersOptions and awards are real claims on the equity, and basic counts understate them.
What going wrong looks likePeer C on basic shares shows £185.0m of equity. Diluted it is £194.0m, and its multiple moves from 8.8x to 9.2x.
The balance sheet is the latest one.Net debt comes from the most recent published balance sheet, and any acquisition, disposal or raise since then is reflected.
Why it mattersEnterprise value has to match the capital structure on the pricing date.
What going wrong looks likeA peer bought a business after its balance sheet date. The bridge misses the acquisition debt and the multiple reads low.
Every claim on the business is in the bridge.Minority interests, preferred shares, pension deficits and other debt-like items go in. Stakes in associates come out if their earnings sit outside EBITDA.
Why it mattersAnything that ranks ahead of the equity is part of what a buyer pays for.
What going wrong looks likePeer F leaves out a £14.0m pension deficit and shows 7.9x, not 8.6x. It is added back in the first pass.

Earnings

The same business can show different EBITDA under different accounting rules.

Lease treatment matches across the table.Under IFRS 16, lease cost moves to depreciation and interest, below EBITDA. Under US GAAP, an operating lease stays a single cost above EBITDA. Put every row on one basis and treat lease liabilities in EV the same way.
Why it mattersThe same company otherwise shows two multiples, and the table compares different measures.
What going wrong looks likePeer D shows 11.2x on its US GAAP numbers and 10.0x on the IFRS 16 basis used by every other row.
The target's accounting basis is named.A UK private company on FRS 102 brings most leases on balance sheet for periods beginning on or after 1 January 2026. A Canadian private company on ASPE keeps the split between capital and operating leases. State the basis, and restate any history that straddles a change.
Why it mattersA multiple has to be applied to earnings on the basis it was built on.
What going wrong looks likeEBITDA steps up in the target's first year on the new rules. Applied at the old multiple, the value is overstated and a buyer takes it straight back.
Adjusted means the same thing in every row.Share-based pay, restructuring, acquisition costs and capitalised development spend get one treatment across the table. Read what a peer adds back before taking its adjusted figure.
Why it mattersEach company defines adjusted EBITDA for itself, so without alignment you are comparing definitions.
What going wrong looks likeOne peer adds back share-based pay and the next does not. The median rests on a definition nobody chose.
Recent deals are pro forma.A peer that bought or sold a business during the period needs a full-year view, or its earnings will not match its EV.
Why it mattersA deal changes debt in full on day one and earnings only from then on.
What going wrong looks likeA mid-year acquirer shows all the debt and half a year of target earnings, so its multiple reads high.

Precedent transactions

Headline value is split.Cash at completion, deferred consideration, earn-out and rolled equity are shown separately.
Why it mattersDeferred and contingent value is not cash at completion, and some of it may never be paid.
What going wrong looks likeA headline at maximum earn-out sets the seller's expectation, and no bidder matches it in cash.
The numerator is enterprise value.Announcements often quote an equity price, or a headline that leaves debt unclear. Confirm the basis, and gross up a part stake to 100%.
Why it mattersEquity value and enterprise value in one column make the multiples meaningless.
What going wrong looks likeA press release price for a stake goes into the table unadjusted, and the multiple is understated.
Undisclosed stays undisclosed.Most private deals publish no value: 55% to 62% in the latest Canadian adviser tables used in our deal market study. An estimate is labelled as one, with who made it.
Why it mattersPrivate deal data is thin, which makes estimates tempting and labelling essential.
What going wrong looks likeA database estimate is shown as fact. The buyer's adviser worked on that deal and knows the real number.
Earnings are dated at the deal.The multiple uses trailing earnings at announcement, not the latest year on a database or a later restated figure.
Why it mattersThe price was set on what the parties could see at the time.
What going wrong looks likeA later restatement makes an old deal look cheaper or dearer than it was.
Unusual deals are marked.Distressed sales, minority stakes, related-party deals and anything struck in a very different rate environment carry a different price. Mark them or keep them off the median.
Why it mattersThese prices reflect the circumstances of the deal, not the value of the business.
What going wrong looks likeA distressed sale drags the low end, and the seller asks why it is in the table at all.
Trading and transaction multiples stay apart.Precedents carry a control premium and often synergies. Trading multiples carry neither.
Why it mattersOne is the price of a minority share on an exchange, the other the price of control.
What going wrong looks likeA single blended range invites the obvious question: which is it?

The page itself

The count is on the page.Mean, median and the number of rows appear together.
Why it mattersA median of four is not a median of twelve.
What going wrong looks likeA tight range rests on three names, and the reader only learns that by asking.
Outliers follow a rule set in advance.Decide the rule before seeing the result, show excluded rows as n.m. and define n.m. once.
Why it mattersRemoving outliers after seeing them is how a range gets engineered.
What going wrong looks likeThe one low row goes and the one high row stays, with no rule. The logic cannot be defended in a meeting.
A few rows are traced to filings.Pick rows at random, including one you have no reason to doubt, and trace price, shares, net debt and EBITDA to the annual report.
Why it mattersDatabase fields misclassify leases, minorities and one-offs, and the same error repeats across rows.
What going wrong looks likeA reviewer finds one wrong row in a meeting, and the whole page loses its credibility.
Like is applied to like.An LTM multiple goes on LTM earnings, a forward multiple on forward earnings, and the target's number is on the same adjusted basis as the peers.
Why it mattersThis is the last step, and a correct table can still produce a wrong value here.
What going wrong looks likeAn LTM multiple lands on next year's budget and quietly adds a year of growth to the value.
The range survives a second method.Cross-check against a DCF or what a financial buyer could pay. If the comps are the only support for the number, say so.
Why it mattersComps are market evidence, not a valuation on their own.
What going wrong looks likeThe range sits above what any financial buyer could fund, and the first-round offers prove it.
The seller's number has been tested.Put the seller's expectation on the same page as the corrected range, and raise any gap now.
Why it mattersA valuation gap is cheapest to surface before the mandate is signed, when the seller can still choose not to sell.
What going wrong looks likeIn the UK and Ireland, sellers' expectations were too high in 49% of processes and the gap ended 24% of those, on Dealsuite data in our deal market study.
Every row can be rebuilt.Provider, filing and date are footnoted, so someone who has never seen the sheet can reach the same number.
Why it mattersThe sheet will outlive the person who built it.
What going wrong looks likeA week before a board meeting the analyst has moved on, and nobody can say where a number came from.

How good teams still get this wrong

The mistake	What it costs	Do this instead
Reviewing the output page rather than the backup.	Bridge and LTM errors never surface, because they are invisible on the final page.	Open the backup first. Trace before you read.
Reusing the last mandate's peer set.	The table answers an old question, and the other side asks about the names you missed.	Rerun the screen each time, and keep only the names that still pass.
Trusting adjusted figures from a database.	One provider's adjusted EBITDA is not another's, and neither is the company's.	Start from reported figures and apply your own adjustments in one place.
Refreshing prices without rolling earnings.	Multiples move for the wrong reason: a new price over an old LTM.	Refresh price, shares, debt and earnings together, or not at all.
Showing one number instead of a range.	A point estimate invites a single argument, and sets an expectation you then have to walk back.	Show the range, its count and its basis, and say what would move it.
Correcting errors without saying what moved.	The partner or the seller hears a new number with no story behind it.	Keep a short change note: what moved, by how much and why.

Comps sanity checklist

One person builds, another reviews and the partner decides what the page claims

A sheet can look clean and still be wrong by most of a turn

The bridge is the first place a buyer's adviser looks

One company, two multiples, depending on the accounting

Errors matter most when they hide a gap to the seller's number

29 checks, in the order that saves the most rework

Peer set

Dates and currency

Enterprise value bridge

Earnings

Precedent transactions

The page itself

How good teams still get this wrong

Read with this

Sources

Rather hand this work over?