An hour spent reading three years of statements side by side, looking for the same handful of things in each, finds more than three separate hours spent reading each year in isolation. Trend, not level, is what a first pass should be optimized to see.
Key takeaways
The instinct on a first read is to work through one year of statements at a time, cover to cover, before moving to the next. That is thorough and it is slow, and it tends to miss the thing a fast first pass is actually looking for: whether a number is drifting in a direction across three years, which a single-year read cannot show by definition. The faster and more useful approach is to pick a short list of items and read all three years for each one before moving to the next item.
The single most common way an hour of reading goes wrong is comparing an SDE-basis year to an EBITDA-basis year as if they were the same measure. Seller’s discretionary earnings “starts from a business’s pre-tax profit and adds back interest, one owner’s compensation and benefits, and discretionary or non-recurring expenses the current owner ran through the business,” while EBITDA “does not add back owner compensation in the same way, on the assumption that the business already pays, or would need to pay, a market wage to whoever runs it.” (deavo.ai/insights/sde-vs-ebitda…) The same source is explicit that “a multiple applied to SDE is not directly comparable to a multiple applied to EBITDA for the same business, since the earnings base itself is calculated differently” — and the same logic applies to a plain year-over-year margin comparison, not just a valuation multiple. Before comparing three years, confirm all three were prepared, or can be restated, on the same earnings basis, or the trend the reader sees may be an artifact of the accounting method rather than the business.
Deavo’s own red-flag list for reading a target’s financials is built for exactly this kind of scan: owner compensation, benefits and personal expenses run through the business; one-time or non-recurring items “such as a lawsuit settlement or a one-off equipment sale”; related-party or non-arm’s-length pricing “including rent paid to a property the owner also owns”; a gap between the statements and what was filed with CRA or for GST/HST; accounts receivable growing faster than revenue; rising inventory without a matching increase in sales; loans to or from related parties; and margins “that move significantly from year to year without an obvious explanation.” (deavo.ai/insights/reading-financial-statements-before-you-buy) Reading down this list once, checking each item against all three years side by side, takes roughly the same time as reading one full year of statements — and it surfaces the trend items a single-year read cannot. The same source is explicit that none of these flags “signal on their own that a business is a bad opportunity” — the goal of the hour is a specific question list for the vendor, not a verdict.
Where a seller or their advisor has already prepared a deal report, it is a useful shortcut but not a substitute for the statements themselves. “There is no single standard format” for a deal report, ranging “from a short summary to a fairly detailed data room,” and it typically bundles two to three years of historical financials plus current interim figures, a normalized add-back earnings summary, an asset list, and customer or revenue concentration. (deavo.ai/insights/reading-a-deal-report-whats-inside) The source is direct about its limits: “a deal report is not a valuation and is not a guarantee that the numbers will hold up,” and the marketplace hosting it “does not prepare, verify, or vouch for the accuracy of any financial or legal information in a deal package.” It also makes a useful secondary point: “the quality of a deal report also tends to say something about how organized the seller’s own business is” — a thin or disorganized report is itself a data point worth noting before the hour is over.
Three years of statements for a service business show revenue of $1.1M, $1.3M and $1.5M — growth that looks strong at a glance. Reading the six items across all three years: owner compensation was $0 in year one, $40,000 in year two and $85,000 in year three, meaning the owner began drawing a real salary partway through the period — the earnings basis is not consistent across the three years and needs restating before margins are compared. AR-to-revenue moved from 8% to 11% to 17%, a clear widening trend. Gross margin held flat at 38–39% across all three years once the compensation add-back is normalized. The read in an hour: revenue growth is real, margin is stable once normalized, but the AR trend is worth a specific question — is collection slowing, or is a large invoice sitting unpaid at year-end — before deciding how much diligence budget this target earns.
A fast read finds trend and asks the right question; it does not verify anything. The widening AR ratio in the example above could be a genuine collections problem, a single large year-end invoice from a normally reliable customer, or a change in payment terms negotiated with a key account — the hour surfaces which of these is worth asking about, not which one is true. The same limit applies to every item on the red-flag list: it flags where to look closer, and it is not a substitute for verifying an add-back against a receipt, or a related-party rent figure against an actual lease. Treating the hour as a screening pass that decides how much deeper diligence a target has earned, rather than as diligence itself, is what keeps it honest.
Related: ratios that tell you whether to keep going, the first ten questions to ask a vendor, and the case file on three years of statements, two sets of books.
Include the notes wherever they exist — related-party disclosures and contingent liabilities, two of the items on the red-flag list above, are often stated only in the notes rather than visible on the face of the statements. Skipping them to save time defeats the purpose of the read.
Treat it as a starting point, not a finished answer. Deavo’s own material notes that two accountants can reasonably normalize the same borderline item differently, so the useful exercise is checking how consistently the seller’s own add-backs were applied year to year, not simply accepting the final normalized number.
A short call is enough to walk through what a first pass at your target’s statements actually found.
Nobody publishes Canadian transaction data, so every valuation in this country quotes an American benchmark. We are building the Canadian one — multiples, asking-to-sale spreads and days on market, by sector and by city. Leave an email and you will see it first.
No pitch, no listings. One email when the first report lands.
