How to Analyse a Company’s Financials

Reading a set of financials tells you what a company reported. To analyse a company’s financials is to do something harder and more useful: separate what the business already owned actually did from everything else mixed into the same number.

Four different things sit inside one set of results, and they behave nothing alike:

1. what the business the company already owned produced
2. what it bought
3. what accounting recognised without any cash moving
4. what management chose to do with the cash

A reported growth rate, a reported margin, a reported earnings per share figure: each one is a blend of those four. The blend is not a trick and nobody is hiding it. The split is almost always published, usually one table further into the same document a headline was taken from.

This guide teaches four questions that pull those things apart, and it works each one on a real Canadian company using only that company’s own filings.

1. Of the growth, how much did the business we already owned produce, and did the rest bring earnings with it?
2. Which margin moved, and on whose sales?
3. Did the earnings arrive as cash?
4. Where did the cash go?

The four questions do not pair off one to one against the four things. Question 1 separates the first two from each other. Question 2 is the consequence of those same two being mixed inside one denominator. Questions 3 and 4 take the third and the fourth.

The questions are the working order. What the page is teaching underneath them is a single family of defects, and every worked example below is a member of it. Each one has a name, and the names are the useful part, because once a defect has a name you start seeing it in filings you have never read before.

• Netting hides magnitude. A subtotal that moved 72 sitting on 1,856 of movement underneath it.
• A numerator with nothing in its denominator. An operating margin containing profit earned on sales that are not in the sales line.
• A denominator that is not what its own definition says. A published leverage ratio.
• One obligation, two lines. A currency loss and a revaluation charge that are two sides of the same euro-denominated liability.
• Two rows with the same name, nine times apart. Net income.
• Rounding larger than the effect being measured. Two per-share growth rates that rank the wrong way round once each is rounded to the cent.
• A prior year that moved. A restated comparative, and a prior-year debt figure that moved by nearly a third.

Here is the finding that ties the page together. Worked below on two real companies, on two of the four questions the analysed number points the other way from the reported one, and on the other two it turns out to be describing a different thing from the one the reported figure describes. In both companies the filings contained everything needed to work that out. Where this page does the working out rather than quoting a row the company printed, it says so. Analysis is arithmetic on disclosure. It is not detective work and it needs no data subscription.

What you will be able to do by the end

• Attribute a growth rate to its sources, and check that the pieces add back to the reported total.
• Name which sales a margin was computed on, and compute the same margin three defensible ways.
• Compare earnings with cash from operations, and name the specific lines that explain the gap, including the two that are one obligation seen from two sides.
• Read the definition of a denominator before trusting a ratio a company hands you, including one that contradicts its own footnote.
• Name which net income row and which share count a per-share figure was built from, and spot where the rounding is bigger than the difference.
• Read a capital allocation policy off a few years of one table.

This guide assumes you can already find your way around the three statements. If the balance sheet, income statement and cash flow statement are still unfamiliar shapes, start with our walkthrough of how to read financial statements and come back.

The two companies on the table, and the money each one reports in

Dollarama Inc. (TSX: DOL) carries questions 1, 2 and 4. Everything below comes from the company’s own fourth-quarter and fiscal-year news releases for four consecutive years: Fiscal 2023 (released March 29, 2023), Fiscal 2024 (April 4, 2024), Fiscal 2025 (April 3, 2025) and Fiscal 2026 (March 24, 2026). Dollarama’s fiscal year ends in late January or early February, and Fiscal 2026 ended February 1, 2026. Every Dollarama figure here is Canadian dollars, in thousands, as the releases print them.

Constellation Software Inc. (TSX: CSU) carries question 3 and the contrast on question 4, across five fiscal years, FY2021 through FY2025, from its own annual Management’s Discussion and Analysis. Constellation reports in US dollars even though its shares trade on the TSX in Canadian dollars. That is not a footnote to skip: it means a Constellation revenue figure and a Dollarama revenue figure are denominated in different money and must never be set side by side as though they were comparable.

Dollarama is one business plus an acquisition plus a stake in a third company. Constellation is a company that buys software businesses for a living. The same four questions work on both, and they expose different things.

Question 1: of the growth, how much did the business already owned produce

Two tables in the March 24, 2026 release hold everything this question needs: Selected Consolidated Financial Information and Selected Segmented Financial Information.

	C$ thousands
Fiscal 2025 sales, entirely Canadian	6,413,145
Fiscal 2026 Canadian segment sales	6,800,927
Fiscal 2026 Australian segment sales	454,827
Fiscal 2026 consolidated sales	7,255,754

Attribute the growth:

• Reported consolidated sales growth: 13.14%
• Of which the Canadian business: 6.05 percentage points
• Of which Australia, first owned during Fiscal 2026: 7.09 percentage points
• Australia’s share of the whole sales increase: 53.98%

Those pieces close exactly against the reported total, which is the check that tells you the attribution is right and not merely plausible. Always do it. If the pieces do not add up, you have the wrong rows.

So far this is only a description of where the sales came from. The move that turns it into analysis is to put earnings beside the same split, pulled from the same segment table.

	C$ thousands
Fiscal 2025 net earnings	1,168,545
Fiscal 2026 Canadian segment net earnings	1,309,683
Fiscal 2026 Australian segment net loss	(245)
Fiscal 2026 consolidated net earnings	1,309,438

The acquisition supplied 53.98% of the sales increase and a net loss of $245 thousand. Consolidated net earnings rose 12.06% in Fiscal 2026, and the whole of that increase came from the Canadian segment: Canada supplied $141,138 thousand against a consolidated increase of $140,893 thousand, with Australia subtracting the difference.

Read one row up before deciding the acquisition lost money

That net loss is the last line of a four-line chain, and the three lines above it are printed in the same segment column.

Dollarama Australia, Fiscal 2026	C$ thousands
Operating income	5,681
Net financing costs	(6,053)
Income taxes, a recovery	127
Net earnings (loss)	(245)

The chain closes. The acquired business traded at an operating profit of $5,681 thousand and was carried below zero by $6,053 thousand of net financing costs, which is the cost of the money used to buy it rather than anything that happened in the stores. Those are two findings with two different futures attached: an operating loss is a trading problem, and a financing cost is a balance sheet decision that can be refinanced, repaid or left alone. Quoting the negative $245 thousand on its own collapses them into one.

Then read what the same release says about the year ahead, because it is where that distinction stops being reassuring. Dollarama states that “the Corporation expects the Australian segment to incur a net loss in Fiscal 2027”, and sets out why: integration costs, a transformation of IT infrastructure, additional headcount and labour costs amounting to an incremental A$35.0 million to A$45.0 million of expenses in the aggregate. So the reconstruction above tells you what happened and the guidance tells you what the company thinks happens next, and the two are not the same story. Separating the lines is the skill. Reading only the favourable half of what separating them reveals is the trap.

Read what the Canadian segment contains too, before drawing the obvious conclusion from it. Footnote (1) to Selected Segmented Financial Information states that “The Canadian segment includes the contribution of the Corporation’s equity-accounted investments in Latin America.” Of the $141,138 thousand increase attributed to the Canadian segment, $61,631 thousand is the rise in equity-accounted earnings, which is Latin America rather than Canadian stores. The segment label is a reporting choice, and the footnote is where the label is defined.

None of that is an accusation. Dollarama acquired The Reject Shop on July 21, 2025, so the acquired business sat inside Fiscal 2026 for a little over six months, and a retail chain does not get reshaped in that time. The point is narrower and more practical: it tells you what the 13.14% describes and what it does not, and it tells you how long the figure will keep behaving this way. An acquisition stops being growth after four quarters. Once the acquired business sits in both the current period and the comparative, its contribution to the growth rate drops to whatever it grew by, which is a completely different number from the one it contributed on the way in.

Count the weeks before comparing two years

Fiscal 2025 had 53 trading weeks. Fiscal 2024 and Fiscal 2026 had 52. A rate measured over 53 weeks of selling against a rate measured over 52 is not a like-for-like comparison, and on the reported rates Fiscal 2025 looks like a bad year that Fiscal 2026 rescued.

Per trading week, growth ran 7.24% in Fiscal 2025 against 15.31% in Fiscal 2026 consolidated. The figure underneath is the one worth having: the Fiscal 2026 Canadian segment alone grew 8.09% per trading week. The business the company already owned did accelerate, and by a great deal less than the consolidated rates suggest, because most of the consolidated acceleration is the acquisition.

The second layer sits inside the comparable sales number

Having isolated the Canadian business, decompose what it did. Comparable store sales in Canada rose 4.2% in Fiscal 2026 on a 52-week basis, “consisting of a 2.4% increase in the number of transactions and a 1.7% increase in average transaction size”. In the fourth quarter, the same measure rose 1.5%, “consisting of a 1.6% decrease in the number of transactions and a 3.1% increase in average transaction size”. Both quotations are from the March 24, 2026 release.

Same metric, opposite composition. The year was traffic-led. The quarter, on its face, was basket-led: fewer people came in and spent more each. A falling transaction count is the kind of thing that stops a reader cold, so keep reading, because the next sentences of the release are the ones that finish the job.

The company states: “Excluding the impact of the calendar shift, Comparable stores sales would have increased by 3.5%, including a 0.5% increase in the number of transactions”. It explains the shift itself: it removed a strong pre-holiday week from the quarter, added a weak late-January week, and dropped four pre-Halloween shopping days. And in the same sentence in which it names the shift, it names a second cause, “offset by the impact of the calendar shift and unfavourable weather conditions negatively impacting store traffic during historically strong sales weeks”.

Now read the corrected figure honestly rather than gratefully. On a like-for-like calendar, transactions rose, so the headline decline was mostly an artifact of where the fiscal quarter happened to end, and the release attributes more of it again to weather. But a 0.5% increase in transactions against 2.4% for the full year is not a quarter that was secretly fine. It is a soft quarter on traffic, with two disclosed reasons and nothing left over that looks like customers leaving.

That is the whole lesson of the second layer, and it is a better one than a clean exoneration would have been. Decomposing a metric into volume and price is not finished when you have the two components. It is finished when you have read the sentences that say which calendar they were measured on, what else the company blames, and what the adjusted figure looks like beside the full-year figure. A reader who stopped one sentence early would have written down falling traffic as a warning sign. A reader who stopped one sentence late would have written down no problem at all.

The same question at a company that acquires for a living

Constellation Software makes the point from the other direction. Revenue went from US$5,106 million in FY2021 to US$11,623 million in FY2025, up 127.63% across the window.

The company publishes its own organic growth rate in its revenue-by-type table, and for FY2021 through FY2025 it reads 7%, negative 1%, 5%, 2% and 4%. Chain those four year-over-year steps together and organic growth came to 10.27% in total across the window, against revenue up 127.63% over the same years. Those four years also cost US$6,825 million of cash spent acquiring businesses.

Constellation also publishes two organic growth rates for FY2025, not one. The MD&A gives full-year organic growth of 4%, or 3% after adjusting for the impact of changes in the valuation of the US dollar. Neither figure is the true one and neither is a correction of the other: one includes the effect of the currency the company reports in, and one strips it out. Quote whichever suits the question you are asking, and say which one it is.

This is not a criticism of the model either. It is the business, stated openly in every filing. Our ranking of the best Canadian growth stocks ranks on revenue per share rather than revenue, and on that measure Constellation is an unusual entry: its revenue growth and its revenue-per-share growth are the same figure, 22.8% a year on our arithmetic over the five-year series above, because the share count did not move. The expansion was not paid for by issuing new shares to the people who already owned some. A reader who values that 127.63% as though it were a same-store number, though, is valuing something that did not happen.

The line worth watching in that filing is the one Constellation breaks out itself. Recurring revenue went from US$3,611 million, or 70.72% of revenue, in FY2021 to US$8,700 million, or 74.85%, in FY2025. That is the portion of the revenue base that arrives again next year without a new acquisition or a new sale, and it is the part of a 127.63% increase that carries forward on its own.

Question 2: which margin moved, and on whose sales

A margin is a fraction. Analysing it means knowing what is in the numerator and what is in the denominator, and the surprising thing is how often those two are not measuring the same business.

Rebuild operating income first, because that is how you know you have the right table

Before computing any margin, reconstruct the subtotal from the lines above it. In every one of Dollarama’s four annual releases, on both the year and its comparative, this closes:

gross profit, less SG&A, less depreciation and amortization, plus share of net earnings of equity-accounted investments, equals operating income

For Fiscal 2026: 3,268,665 − 1,093,289 − 429,053 + 191,536 = 1,937,859 (C$ thousands). It closes to the dollar in all eight columns of the four releases.

That exercise does two things. It confirms you are reading the rows you think you are reading, and it forces you to notice every term you had to include, which is where the interesting one is hiding. There is a second check with the same purpose, and it is the one worth running first on any company that reports segments: the segment sales have to add to consolidated sales. Canadian sales of $6,800,927 thousand plus Australian sales of $454,827 thousand give $7,255,754 thousand, which is the consolidated figure, so the table in front of you is the consolidated one. A segment column and a consolidated column print the same row labels a few inches apart, and that addition is how you tell which one you are holding.

The fourth term of the reconstruction is the one to notice. Dollarama’s operating income contains its share of the profit of a business whose sales are not in Dollarama’s sales line. Dollarcity, held through the entities the company calls CARS and ICM, is a joint arrangement: IFRS 11 is the standard that classifies it as one, and IAS 28 is the standard that then tells Dollarama how to measure it, using the equity method. Dollarama books its share of Dollarcity’s profit as earnings, and none of Dollarcity’s revenue appears in Dollarama’s revenue. So the reported operating margin has a numerator with nothing in its denominator, which is the second defect on the list above, and that is the whole of this section.

Three readings of the same year, all defensible

Here is the operating margin for Fiscal 2026 against Fiscal 2025, computed three ways from the same release. The levels are carried to three decimals so that the change column reproduces from its own rows, which a two-decimal version will not do.

Basis	Fiscal 2025	Fiscal 2026	Change
As reported, consolidated	26.675%	26.708%	+0.03 points
Consolidated, excluding equity-accounted earnings	24.649%	24.068%	−0.58 points
Canadian segment, excluding equity-accounted earnings	24.649%	25.594%	+0.95 points

Fiscal 2025 had no Australian segment, so its consolidated and Canadian figures are the same number. The reported row is Dollarama’s own. The other two rows are our arithmetic on the operating income, equity-accounted earnings and sales rows of the same tables.

Three answers to “did the margin improve”: flat, worse, better. None of them is wrong. The question you are asking decides which one is the answer.

• Is the consolidated entity’s margin profile improving? No. It bought a lower-margin business, and the average came down accordingly.
• Is the Canadian retail segment getting better at its job? Yes, by 0.95 points, with the equity-accounted line stripped out of both years.
• Did the reported operating margin move? Barely, because the equity-accounted line grew enough to offset the mix effect and hide both of the answers above.

Two readers can disagree about whether this company’s margins improved while both are quoting a real number out of the same release. The figure means nothing until one of them says which sales it was computed on.

What “a lower-margin business” means, in the rows that say it

“Mix” is easy to assert and rarely shown. The segment table shows it, on our arithmetic on the Fiscal 2026 segment rows.

Fiscal 2026, percent of segment sales	Canada	Australia	Difference
Gross margin	45.61%	36.61%	9.01 points in Canada’s favour
SG&A	14.42%	24.71%	10.29 points in Canada’s favour
Depreciation and amortization	5.60%	10.65%	5.05 points in Canada’s favour
Operating margin	25.59%	1.25%	24.35 points in Canada’s favour

The Canadian operating margin in that table is the ex-equity figure, so that both columns are measuring the same thing: what the stores earn on what the stores sell. Three gaps produce the fourth, and they have to close against it. Australia buys and sells at a gross margin 9.01 points worse, spends 10.29 more points of its sales running the business, and carries 5.05 more points of depreciation and amortization: 9.01 + 10.29 + 5.05 = 24.35, which is the operating margin gap exactly. The depreciation row is the one most likely to be skipped and it is a fifth of the whole difference, because the acquired chain carries nearly twice Canada’s depreciation burden per dollar of sales, 10.65% against 5.60%. Weighted by $454,827 thousand of Australian sales, 25.59% and 1.25% average to 24.07% on the same ex-equity basis, which is the middle row of the three-basis table above. Every point of that deterioration is the arrival of the acquired business rather than anything deteriorating.

The half worth practising is that the direction of the error is not predictable in advance. Mix can flatter a consolidated margin as readily as it can damage one, depending on which side of the core the acquired business sits, and only the segment table tells you which happened.

Two things that move an equity-accounted line, and only one is operations

Dollarama’s share of Dollarcity’s earnings went from $129,905 thousand in Fiscal 2025 to $191,536 thousand in Fiscal 2026, which the release states as a 47.4% increase. It attributes that increase to “the continued strong operational performance of Dollarcity” and to “the impact from the acquisition of an additional 10.0% equity interest in CARS on June 11, 2024”.

Read the ownership carefully, because it changes what the increase means. The prior-year figure reflected 50.1% of CARS for part of the year and 60.1% for the rest. The current figure is 60.1% throughout, plus 80.05% of ICM. Part of the $61,631 thousand increase is Dollarcity selling more product, and part of it is Dollarama owning a larger slice of the same product.

The release gives you the separator for the operating half. Dollarcity’s own sales rose 20.2% over the twelve months to December 31, 2025, with its store count going from 632 to 732, and that is measured on Dollarcity’s whole business before any ownership percentage is applied to it.

Then read to the end of the sentence the 20.2% sits in, because it names more than sales. Dollarcity’s performance “was mainly driven by a 20.2% increase in sales, supported by growth in the total number of stores (from 632 on December 31, 2024, to 732 on December 31, 2025) and the increase in gross margin as a percentage of sales from lower inbound shipping and logistics costs. This was partially offset by a slight increase in SG&A as a percentage of sales from costs associated with Dollarcity’s expansion plans in Mexico.” So four disclosed things are moving inside the one line Dollarama reports: sales growth, a gross margin improvement, an SG&A offset, and the ownership step. The figure you came for is a single number, and the sentence underneath it is four.

One trap sits in the way of that figure, and it is exactly the kind this page exists to teach. The same release states 28.3% for the fourth quarter alone, in the same sentence pattern: Dollarcity’s “strong fourth quarter performance was mainly driven by a 28.3% increase in sales, primarily attributable to an increase in Comparable store sales and in total number of stores”. Both sentences then carry the same two store counts, 632 at December 31, 2024 and 732 at December 31, 2025, because both quote the same year-over-year pair. Two sales growth rates, one sentence pattern, one set of store numbers between them, and only one of them belongs in a twelve-month comparison. Read the period before the percentage, every time.

One further basis note on the same line, because it is the kind of thing that quietly corrupts a model: the $191,536 thousand was earned over January 1, 2025 to December 31, 2025, which is not Dollarama’s fiscal year.

Question 3: did the earnings arrive as cash

Earnings are an opinion formed under accounting rules. Cash from operating activities is closer to a fact. When the two disagree, the disagreement is the interesting part, and Constellation Software’s FY2025 is about as clean a teaching case as exists.

Five years from the company’s own annual MD&As, US dollars in millions except the per-share line:

US$ millions	FY2021	FY2022	FY2023	FY2024	FY2025
Revenue	5,106	6,622	8,407	10,066	11,623
Net income to common shareholders	310	512	565	731	512
Net income, total	169	551	62	767	586
Cash from operating activities	1,300	1,297	1,779	2,196	2,732
Free cash flow available to shareholders	883	853	1,160	1,472	1,683
Reported EPS (US$)	14.65	24.18	26.67	34.48	24.15

Now the single page that carries the whole story. From the Results of Operations section of Constellation’s FY2025 Management’s Discussion and Analysis, page 3, in the company’s own variance columns:

• Revenue up 15%
• Net income to common shareholders down 30%, from 731 to 512
• Net income in total down 24%, from 767 to 586
• Earnings per share from US$34.48 to US$24.15
• Cash from operating activities up 24%, from 2,196 to 2,732
• Free cash flow available to shareholders up 14%, from 1,472 to 1,683

Three of those statements describe a bad year and three describe a good one, and all of them are on the same page. The last one is worth its own beat: the company’s own measure of the cash a shareholder could have been paid rose in the year reported earnings per share fell 30%, and across the five years in the table above it has fallen once, in FY2022, while reported earnings per share has fallen once as well, in the year the page is about.

The first thing to rule out is the share count, because a per-share figure can move for reasons that have nothing to do with the business. Constellation’s weighted average share count was 21.2 million in every one of the five years. Per share and in total tell exactly the same story here, and nothing in the divergence is a share-count artifact. If the mechanics of why a share count moves a per-share figure are not yet second nature, our explainer on what a stock actually is builds it from the ground up.

The charges were enormous. The pre-tax line barely moved.

The intuitive next step, having found a fall in earnings, is to look for the charges that caused it. Constellation’s Results of Operations table lets you do better than that: it prints the effect of every line, so the whole year can be laid out at once. US$ millions, positive helping pre-tax income and negative hurting it, taken from the company’s own printed variance column.

Effect on FY2025 income before income taxes, US$ millions
Revenue	+1,557
Expenses	−987
Amortization of intangible assets	−139
Foreign exchange (gain) loss	−181
IRGA / TSS revaluation charge	−257
Finance and other expense (income)	+169
Bargain purchase gain	0
Impairment	−15
Redeemable preferred securities	+58
Revaluation of equity-method investment to cost	−260
Finance costs	−17
Reported change in income before income taxes	−72

That column foots exactly, and it is worth seeing what happens if you build it the other way instead, because the failure is the demonstration of a rule rather than a quibble. The filing computes its variance column before rounding, and its levels are each rounded to a million. Subtract the rounded levels and three rows come out a unit light: amortization of intangible assets gives 138 where the company prints 139, foreign exchange gives 180 against 181, and finance and other expense (income) gives 168 against 169. A column built that way comes to −71, and the reported change in income before income taxes is −72. The company’s column closes and ours does not. That is the whole reason for the rule: when you quote a change, quote the company’s change. When you quote levels, quote the levels.

Now read what the table says. The adverse movements in the year total 1,856, of which 852 is marks and currency: amortization of intangibles, the foreign exchange swing, the IRGA charge, the revaluation of an equity-method investment to cost, and impairment. Earnings to common shareholders fell 219. The adverse movement in those lines is 3.89 times that fall. They do not explain part of the decline. They over-explain it several times over, and the reason pre-tax income fell only 72 is that 1,557 of revenue growth landed in the same year and absorbed nearly all of them.

That is the first defect on the list at the top of this page, and the one most likely to be got backwards. Netting hides magnitude. A small change in a subtotal is not evidence that nothing large happened underneath it. It is evidence that the large things cancelled, and the only way to know which way each one ran is the line-by-line table the company already printed. The same reasoning applies to any subtotal in any filing, which is why this is the habit worth taking off the page.

The chain below pre-tax has its own movers

Pre-tax income is not where a headline earnings figure comes from. Two further steps sit between them, and in FY2025 both of them ran against the company.

US$ millions	FY2024	FY2025	Change
Income before income taxes	1,011	939	−72
Income tax expense	244	353	+109
Non-controlling interests	37	74	+37
Net income to common shareholders	731	512	−219

A 72 fall at the pre-tax line became a 219 fall at the shareholder line, because income tax expense rose 109 and the minority holders of Constellation’s subsidiaries took 37 more.

Work that bridge with a calculator and it lands a unit out: 72 plus 109 plus 37 is 218 against the printed 219, and the FY2024 column gives 1,011 less 244 less 37 equals 730 against a printed 731. The filing says to expect that, in its note on presentation at the front of the document: “Due to rounding, certain totals and subtotals may not foot and certain percentages may not reconcile.” Every term is rounded to a million before you see it, so a bridge rebuilt from the printed rows lands within a unit of the printed total rather than on it. One unit is the rounding. Anything larger is a wrong row, and that is the distinction worth holding on to for every bridge further down this page.

The tax step is not one you have to estimate, because the company states it. Constellation gives its consolidated effective tax rate in respect of continuing operations as 24% for the twelve months of FY2024 and 38% for FY2025. Our own arithmetic on the two printed rows, 244 over 1,011 and 353 over 939, gives 24.1% and 37.6%, which is a check that we read the right two rows rather than a better figure than the company’s. The MD&A gives the drivers in general terms only: the rate is “affected by the realization and anticipated relative profitability of our operations in those various jurisdictions, as well as different tax rates that apply and our ability to utilize tax losses and other credits”. It is natural to wonder whether the large marks were non-deductible, and the filing does not say so, which means a 14-point move in a company’s tax rate is not something to attribute by guess. The non-controlling interest step is a question of who owns the subsidiaries, which has nothing to do with how the subsidiaries traded.

So a reader who wants to explain the 30% fall in earnings per share needs both halves. Above the tax line, an enormous adverse swing in marks and currency, mostly absorbed by revenue growth. Below it, a tax rate with its own drivers and a minority split that turned what was left into something three times its size. Either half on its own produces a confident and wrong account of the year.

The non-cash lines, and the two that are one obligation

Now the charges themselves, from the same variance columns.

Amortization of intangible assets went from US$1,044 million to US$1,182 million, up 13%. This is the accounting cost of the software businesses Constellation bought, spread over years. No cash leaves the company when it is recorded, and for an acquisitive company the line is large by construction and grows as the acquisitions accumulate.

The IRGA and TSS membership liability revaluation charge went from US$183 million to US$440 million, up 140%. It is a revaluation of a liability Constellation owes under an agreement covering its Topicus.com interest, and no cash moves when it is recorded.

It is worth knowing what that charge actually tracks, because “an accounting mark” is the kind of phrase a reader files away and stops thinking about. The MD&A says: “The IRGA / TSS membership liability revaluation charge relating to the investment in equity securities of Sygnity and Asseco was $144 million and $252 million for the three and twelve month periods respectively. The fair value of these investments … is determined by their respective share prices at the end of each reporting period.” So US$252 million of the US$440 million charge, 57.3% of it, is the mark on the shareholdings in two listed Polish companies, set by where those two shares closed at the period end. That is a concrete thing with a concrete driver, and it behaves like one: it can reverse in a later period without anybody at Constellation doing anything.

Two more lines moved against the company in the same year: a revaluation of an equity-method investment to cost, and a swing in foreign exchange to a loss of US$154 million. Four separate causes, then, until you read the note on the liability.

The MD&A explains that the liability rose “78% or $541 million over the twelve month period ended December 31, 2025 from $693 million to $1,234 million as a result of the revaluation charge of $440 million and a $101 million foreign exchange loss. The IRGA / TSS membership liability is denominated in Euros and the Euro appreciated 10% versus the US dollar”.

Liability of CSI under the IRGA, US$ millions
December 31, 2024	693
December 31, 2025	1,234
Stated increase	541
of which the revaluation charge	440
of which a foreign exchange loss	101

So US$101 million of the US$154 million foreign exchange loss is the same euro-denominated Topicus obligation that produced the US$440 million revaluation charge. One obligation, two separate lines of the income statement, each recorded once by the company, and the two add to the whole $541 million movement the company states.

The error waiting there is not arithmetic, it is attribution: reading the foreign exchange line as an unrelated event when two thirds of it is the same obligation as the revaluation charge. Both lines belong in a tally of what ran against the company in the year, which is why the 852 of marks and currency above counts both of them. What the reader needs is to know that the two are one exposure, because they will move together again next year, and a model that treats them as independent will be wrong twice in the same direction.

This is the most useful habit on the page and the hardest one to build, because it requires reading past the table you found the number in. Two lines on an income statement can be two sides of one economic event, and the only way to know is the note that explains what moved the liability.

The charge that runs the wrong way

The IRGA charge is worth a paragraph because of what Constellation does with it. It is an accounting mark, not a payment, and the company deducts it anyway from its own free cash flow measure, because the obligation is real. A company that marks its own free cash flow down by a non-cash charge is not trying to flatter the number.

The same restraint shows up elsewhere in the filings. Across all five annual MD&As there is no adjusted net income, no adjusted EPS, no adjusted diluted figure and no adjusted earnings. A company carrying US$1,182 million of acquisition amortization has every excuse in the world to publish a figure that adds it back, and it publishes none.

Where the cash actually came from, in the company’s own sentence

The question this section asks is whether the earnings arrived as cash, and Constellation answers it directly in one sentence of the MD&A. The sentence closes.

US$ millions
Net income, total	586
Plus adjustments to net income	2,709
Less cash used in non-cash working capital	6
Less taxes paid	556
Builds to	2,733
Reported cash from operating activities	2,732

The one-unit gap is the filing rounding each term to a million. The same sentence names what is in the adjustments: “primarily amortization of intangible assets, depreciation, IRGA/TSS Membership liability revaluation charge, finance and other income, revaluation of investment accounted for using the equity method to cost, finance costs, and income tax expense”.

Then test the obvious guess against it. A software company with US$8,700 million of maintenance and other recurring revenue looks like a company that collects subscription cash up front, which would put cash ahead of earnings on billing timing. For this year the filing says otherwise. Deferred revenue did rise, by $248 million, “mainly due to acquisitions made since December 31, 2024 and the timing of maintenance and other billings versus performance and delivery under those customer arrangements”, and non-cash working capital in total still used $6 million of cash. The whole of the distance between 586 of net income and 2,732 of operating cash is the add-backs. Taxes, meanwhile, were heavier in cash than in the accounts: 556 paid against the 353 expensed, a gap of 203, in the same year the company-stated effective rate moved from 24% to 38%. The tax step that turned a 72 pre-tax fall into a 219 fall at the shareholder line arrives here a second time, as cash out of the door rather than as an expense in the accounts.

That is what a check is for. The intuition was reasonable, the filing settles it in one sentence, and the answer is not the intuition. It also sets up the thing to look for when a gap runs the other way.

When earnings rise and cash does not, look in the inventory line

Constellation’s gap runs the reader-friendly way: cash ahead of earnings, explained on the page the fall appears on. The gap worth worrying about runs the other way, earnings ahead of cash, and the usual mechanism is the working capital that was not the explanation above. Money spent on inventory that has not sold yet, or on a receivable a customer has not paid yet, is an outflow of cash that never touches the earnings line.

Dollarama’s balance sheet shows how to size it. Inventories went from $921,095 thousand at February 2, 2025 to $1,103,175 thousand at February 1, 2026, up 19.77%, against sales growth of 13.14% over the same year. That is a gap of 6.63 points, and a company whose inventory grows faster than its sales is absorbing cash into stock on shelves.

Be careful about what that does and does not prove. A year in which a company acquires a retail chain will carry more inventory for exactly that reason, and Fiscal 2026 was that year. So the figure is a question to ask rather than a verdict to deliver. What would turn it into a verdict is the same gap repeating, in a year with no acquisition to explain it, while cash from operations trails net earnings. That is three observations, not one, which is why this question needs several years of filings rather than the latest one.

Earnings and cash at different moments, at Dollarama

Two smaller cases from the March 24, 2026 release, each a place where earnings and cash belong to different periods.

The equity method, again. Dollarama recorded $191,536 thousand of Dollarcity earnings in Fiscal 2026. Cash from those earnings arrives only when Dollarcity declares a dividend. On February 5, 2026, after the fiscal year ended, CARS declared a US$125.0 million dividend, of which Dollarama’s share was US$75.1 million (C$102.2 million), expected in the first quarter of Fiscal 2027. Real earnings, real cash, different years.

An unrealized gain. A $10,348 thousand unrealized gain on a derivative on equity-accounted investments lifted the EBITDA margin by 20 basis points and diluted EPS by $0.03, on the company’s own reckoning. Dollarama publishes a line called EBITDA excluding unrealized gain from derivative on equity-accounted investments to isolate it: $2,397,878 thousand, against EBITDA of $2,408,226 thousand. That reconciliation is not voluntary courtesy. Under National Instrument 52-112, a Canadian issuer presenting a non-GAAP measure has to give the definition and, in the words of Dollarama’s own footnote, “where applicable, their reconciliation with the most directly comparable GAAP measure”. Two things are worth carrying out of that. The qualification is real, and the release itself files its measures under separate headings, with the leverage ratio sitting under one headed “(B) Non-GAAP Ratios” rather than with the financial measures, which matters for that ratio later on this page. For a measure like EBITDA excluding one named gain, the practical effect is the useful one: the unadjusted figure is in the same document, so you never have to take the adjusted one on faith.

Understanding why reported earnings fell while cash generation rose tells you what happened inside a business. It does not tell you what the shares did afterwards, which depends on what the market already expected and on a great deal else besides. Our guide to what moves a stock price covers that gap between a result and a reaction, and the two questions are worth keeping separate in your head.

Check which net income you are holding

Two lines on the same page of the same filing both answer to the name net income, and in Constellation’s FY2023 they differ by roughly nine times.

Total net income in FY2023 was US$62 million. Net income attributable to common shareholders of CSI was US$565 million. The difference is non-controlling interests absorbing a loss, which belongs to other people’s stakes in subsidiaries rather than to Constellation’s own shareholders. Reported earnings per share is calculated on the attributable figure.

A reader who grabs the total row for FY2023 gets a company that barely broke even on US$8,407 million of revenue, and every ratio built on top of that row is wrong: the margin, the return on equity, the price-to-earnings multiple, all of them. A reader who grabs the attributable row gets the company whose per-share figures the market actually quotes.

This is not a rare edge case. Any company with meaningful minority interests has both rows, they are printed adjacent to each other, and the gap between them is not stable from year to year. Total net income is attributable to common shareholders plus non-controlling interests, and the FY2025 table above shows the addition: 512 plus 74 is 586. So the sign of the gap tells you what the minority interests did. Look at the FY2021 column of the five-year table: total net income of US$169 million against US$310 million attributable, a total below the attributable figure, which is the minority interests absorbing a loss. Then FY2022, where the total of US$551 million sits above the attributable US$512 million, because that year the minority interests earned a profit rather than taking a loss. The gap changes sign. There is no rule of thumb that saves you from checking the label.

Before computing anything per share, write down which of the two rows you are holding. It is the single cheapest error to avoid on this page.

Question 4: where did the cash go

The cash flow statement’s financing and investing sections are where a management team’s actual priorities are recorded, as opposed to its stated ones. A few years of one table will tell you the policy, and the order of the numbers by size is the order of the priorities. Our guide to how to research a stock lays that table out for Dollarama by fiscal year, with buybacks, dividends and capital expenditure side by side, if you want the policy in one place before reading the two things below.

The price paid rose every year. Dollarama’s weighted average repurchase price went from $77.28 in Fiscal 2023 to $188.47 in Fiscal 2026. A buyback is a purchase, and a purchase is good or bad depending on what was paid relative to what the thing was worth. “The company bought back stock” is not by itself good news, any more than “the company issued stock” is by itself bad news.

There is a check on that pair of rows a reader can run in a minute. Divide each year’s buyback dollars by the number of shares repurchased that year, not by the company’s total share count, and compare the answer with the weighted average price the company states in its own prose. In three of Dollarama’s four years it lands exactly on the stated price; in Fiscal 2024 it gives $92.05 against a stated $92.04, because the dollar totals are published to one decimal in millions and a rounded numerator cannot do better. A cent of disagreement confirms the two rows describe the same programme. A dollar of disagreement would mean they do not.

The dividend is not where the money went. Declared dividends per share across the four years were $0.2212, $0.2832, $0.3680 and $0.4232, against diluted EPS of $2.76, $3.56, $4.16 and $4.73. That is a payout ratio of 8.01%, 7.96%, 8.85% and 8.95%: a dividend that rises steadily and consumes almost nothing. If you want to see what a payout at that level does to an income stream over time against one at fifty or sixty percent, our dividend income calculator will run the arithmetic on any starting position and growth rate. How much of a Canadian dividend actually reaches you also depends on the account it lands in and the credits attached to it, which is the subject of our guide to how investment income is taxed in Canada.

The EPS bridge, and which share count it used

Reported earnings per share growth blends two things: what the business earned, and how many shares the earnings were divided by. Separating them takes one line.

Dollarama, Fiscal 2026:

• Net earnings up 12.06%
• Diluted weighted average share count down 1.47% (280,819 thousand to 276,684 thousand)
• 1.1206 ÷ 0.9853 = up 13.73%
• Reported diluted EPS growth: up 13.70%

The bridge closes against the reported figure, which is the check. Part of the 13.70% was bought rather than earned, and the size of that part is the distance between the first and last lines above. Neither sentence is a complaint. Buying back shares at a price below their worth transfers value to the holders who stay, and that is a real return. But it is a financial return rather than an operating one, and forecasting it requires assumptions about future buyback prices rather than about future sales.

Then write down which share count you used, because the release prints two of each, one row apart from the other.

Dollarama	Fiscal 2025	Fiscal 2026	Change
Basic EPS	$4.18	$4.75	+13.64%
Diluted EPS	$4.16	$4.73	+13.70%
Basic shares, thousands	279,825	275,611	−1.51%
Diluted shares, thousands	280,819	276,684	−1.47%

The basic count fell further than the diluted count, so basic earnings per share should be growing the faster of the two. In those printed rows it is the slower, by 0.06 points. Rebuild each one from net earnings over its own share count without rounding anything and the order comes back: basic runs $4.17598 to $4.75104, up 13.77%, and diluted runs $4.16120 to $4.73261, up 13.73%. The printed pair ranks the other way because each per-share figure is rounded to the cent, and $4.17598 prints as $4.18, which lifts the prior-year base and flattens the growth rate measured off it. A per-share figure a company prints is rounded, and at two decimals the rounding can be larger than the thing you are trying to measure. That is the three-decimal discipline of the margin table above, arriving at the denominator instead: two figures from one table with the same name and different contents, and nothing in either figure tells you which one you are holding or how much of the difference between them is real.

The contrast: a company that does the opposite with the same coherence

Constellation Software, five years, from its annual MD&As:

• Dividend declared: US$1.00 per quarter, US$4.00 a year, unchanged every year.
• Weighted average share count: 21.2 million, unchanged every year.
• Cash used to acquire businesses: US$1,337 million, US$1,782 million, US$1,847 million, US$1,683 million and US$1,513 million.

And the policy in the company’s own words, from the MD&A: “our objective is to invest all of our FCFA2S in acquisitions which meet our hurdle rate”.

Two coherent capital allocation policies pointing in opposite directions, both legible from four or five years of a single table before reading a word of narrative. One returns almost everything and builds slowly. One returns a token amount and reinvests the rest. Which is preferable is a question about the opportunities each company faces. That it is knowable at all, from documents anyone can download, is the point.

What the filings cannot tell you is how a share price behaves while a policy is running, and that is a separate question worth keeping separate. Our ranking of the best Canadian blue chip stocks puts both of these companies through the same test, the drawdown each one handed its holders across three market shocks, and they come out of it in opposite places, one of them among the sturdiest names on that page and the other among those that fell further in a rising market than they did in the crash.

A ratio whose definition contradicts its own footnote

Every answer above produces a ratio, and a ratio is only as meaningful as the two things inside it. Companies define the ones they publish, usually in a footnote under the table, and the footnote is where the analysis is. Dollarama’s leverage disclosure at February 1, 2026 is the sharpest example in either company’s filings, because the definition and the footnote do not agree with each other, in the same press release.

Here is the definition, from the release’s Non-GAAP section:

“Adjusted net debt to EBITDA ratio is a ratio calculated using adjusted net debt over consolidated EBITDA for the last twelve months.”

Consolidated EBITDA. Unambiguous, and the consolidated figure is printed elsewhere in the same release: $2,408,226 thousand for Fiscal 2026.

Here is the footnote to the ratio itself, building the denominator it actually used:

Denominator of the stated ratio	C$ thousands
Consolidated EBITDA of the Corporation	2,408,226
Plus Dollarama Australia’s EBITDA before Dollarama owned it	37,761
Denominator actually used	2,445,987

The footnote describes that second line as Dollarama Australia’s EBITDA “for the period between February 3, 2025 until closing of the TRS Transaction on July 21, 2025 (as calculated and reported by Dollarama Australia)”.

Read that again, because it is easy to skim. The denominator of the ratio is not consolidated EBITDA, which is what the definition in the Non-GAAP section of the same release says it is. It is consolidated EBITDA plus earnings from a business Dollarama did not own during the period those earnings were generated, calculated and reported by that business rather than by Dollarama. On consolidated EBITDA alone, the same adjusted net debt is 2.10 times against the stated 2.07 times.

What the adjustment is for is in the footnote’s own dates, and it is worth working out rather than filing away as carelessness. The numerator of this ratio is a balance sheet at a single instant, February 1, 2026, by which date the Australian business was owned outright and whatever was borrowed against it sits in the debt figure in full. The denominator covers a period, and consolidated EBITDA includes that business only from the July 21, 2025 closing. Adding Dollarama Australia’s own earnings for the months before closing puts twelve months of flow on the same perimeter as the stock it is being divided into. That is our reading of the dates and the amount, which is what the filings give; neither the definition nor the footnote states a rationale. The general shape of the problem is worth more than this one instance: a ratio that divides a stock by a flow has to decide which period the flow covers, and that decision lives in the footnote rather than in the definition.

The distance between 2.07 and 2.10 changes nothing about Dollarama, and nothing here suggests it was meant to. That is precisely why it is the best example on the page. The adjustment is defensible on its own terms, it is fully disclosed, a reader can rebuild it from figures in the same document, and it is still not the ratio the definition promised. Nobody had to hide anything for a published number to mean something other than what it says. Somewhere in your investing life a company will hand you a leverage multiple, a return on capital or a margin whose denominator has been assembled the same way, and the gap will matter more than this one does. The habit that catches it is not scepticism. It is reading the definition, then reading the footnote, then checking that they describe the same thing.

The numerator has two versions too

Net debt at February 1, 2026 was $2,293,552 thousand. The company also publishes adjusted net debt, at $5,063,206 thousand, and lease liabilities are almost all of the difference. They sit outside the line called debt but they are a contractual obligation to pay cash for years, which is why IFRS 16 brought them onto the balance sheet, and why a retailer’s adjusted figure is more than twice its unadjusted one.

Our guide to financial ratios explained works that lease trap in detail on this same company at a later date, along with the DuPont decomposition that shows where an extraordinary return on equity actually comes from. The stub added to the denominator here exists only while the acquisition sits part-way inside the twelve months, so by the second quarter of Fiscal 2027 the company’s own last-twelve-month EBITDA needs nothing added to it.

Which multiple you quote off that numerator matters, and so does whose multiple it is. The release states 2.07 times for Fiscal 2026 and 2.16 times for the prior year, and no other leverage multiple appears in it. Net debt over the same denominator comes to 0.94 times on our arithmetic, not on the company’s, and anyone repeating it should say so. Two ratios on one balance sheet, more than two times apart, and the only thing separating them is which obligations the numerator was told to count.

One more denominator worth checking on any company that buys back a lot of stock: Dollarama’s shareholders’ equity at February 1, 2026 was $1,455,888 thousand against total assets of $7,558,352 thousand, an equity multiplier of 5.19 times, which is a description of a capital structure rather than a grade awarded for operating skill.

Red flags, and four things that look like one and are not

Pattern recognition is the part of analysis that comes last, because it needs several years of practice on real filings. Here is a starting set, drawn from the two companies above.

Genuine warnings

Earnings and cash diverging in the same direction for years. One year is a charge. Constellation’s FY2025 gap is one year, runs in the reader’s favour, and is fully explained on the page it appears on. A company whose net income rises every year while cash from operations does not is a different animal entirely, and the inventory and receivable lines are where to look first, using the growth comparison worked above.

A prior year that moved, which is to say a restated comparative. The Fiscal 2026 release states Fiscal 2025 capital expenditures as $243,450 thousand. The Fiscal 2025 release stated $246,869 thousand for the same year. A $3,419 thousand restatement is immaterial in itself, and that is exactly the point: it proves prior years move, and a series stitched together from several filings must take the latest published version of each figure.

Constellation’s version of the same defect is not immaterial at all. Its FY2024 MD&A presented December 31, 2024 debt with recourse to CSI at US$2,159 million. Its FY2025 MD&A presents the same date at US$1,466 million. Cash and debt without recourse are identical across the two filings, and the difference, US$693 million, is to the million the IRGA and TSS liability at that date, which the FY2024 MD&A lists as one of the five components it builds its $2,159 million total from. That identity is our reading of the two tables, and it is exact. Neither filing states that the item was reclassified, or why.

So a prior-year debt figure moved by 32% of the way it was first presented, with no number in the business changing, and the whole of the move is one obligation sitting inside a line called debt in one filing and outside it in the next. That is the leverage question above arriving as a comparative: which obligations a numerator was told to count, asked this time of a figure you might have written down a year earlier and never checked again. A $3,419 thousand move and a $693 million move make the same point at both ends of the scale. Never quote a prior-year figure from the older filing when a newer one carries it.

Falling transactions with a rising basket, repeated across quarters, with nothing in the release to explain it. One quarter of it is noise, as the Dollarama example above shows. Several quarters of it, with no shift in the fiscal calendar and no weather or pricing to point at, is a business whose customers are visiting less often, and the average basket will not hide it forever.

An ownership percentage moving inside a growth rate, as in the CARS step from 50.1% to 60.1%. Any time the share of a business being consolidated or equity-accounted changes, part of the year-over-year change is the ownership change rather than the business.

Not red flags

A single quarter’s transaction decline the company explains. Dollarama’s fourth quarter is the worked example: transactions down 1.6% as reported, up 0.5% once the shifted weeks are taken out, with a calendar shift and unfavourable weather both named in the release. A soft quarter, and a specific one, rather than a trend.

Tiny equity from buybacks. A 5.19 times equity multiplier is a chosen capital structure, not distress. Distress is a structure the company cannot service, which is a question about cash flows and covenants rather than about the size of the equity line.

A large gap between total and attributable net income. It usually means significant minority interests, which is a structural fact about who owns the subsidiaries, not something concealed.

No adjusted EPS. Constellation publishes none. That is the more conservative choice, not an omission.

An acquisition that dilutes consolidated margins. It tells you the acquired business earns less than the acquirer, which was knowable before the deal closed. Whether the purchase was a good one is a different question from whether the average moved.

What reading the growth rate and stopping actually costs

Worked on both companies at once. Someone who read only the headline in each case came away with:

• Dollarama: a company growing sales 13.14%, up from 9.30%.
• Constellation: a company whose earnings per share fell 30%.

Both statements are true. Both conclusions drawn from them are wrong, and for different reasons.

The Dollarama reader has the direction right and the size and the reason wrong. Only 6.05 of those 13.14 percentage points came from the business the company already owned. The other 7.09 came from an Australian segment first owned during the year, which traded at an operating profit of $5,681 thousand, finished at a net loss of $245 thousand after the cost of the money used to buy it, and stops flattering the comparison after four quarters.

The Constellation reader has it backwards twice over. The company generated 24% more cash in the year its earnings fell, for reasons disclosed on the same page as the fall, and the share count dividing both figures did not move. And the charges were not a modest contributor to be netted off: 852 of marks and currency landed against a 219 fall in earnings to common shareholders, and the pre-tax line moved only 72 because revenue rose 1,557 in the same year and absorbed nearly all of it, with 169 from finance and other expense (income) and 58 from the redeemable preferred securities line covering most of the rest. Read the small subtotal as a quiet year and you have missed everything that happened.

What it costs, concretely. Projecting 13.14% forward for a company whose own business contributed 6.05 points of it overstates sales five years out by 38.22%, on our arithmetic compounding each rate over five years, and that error compounds into every valuation built on top of it. If the mechanics of why a small annual difference becomes a large one over five years are not intuitive yet, our explainer on how compounding works is the shortest route to making it intuitive, and it is the same arithmetic running in both directions.

Selling a company because reported earnings fell 30% in a year its cash generation rose 24% is the same error with the sign reversed. In both cases the correction took one table out of a document the company had already published.

The routine, in order

Run these before any consolidated percentage goes into a model.

1. Write down the period and how many weeks it contains, and the comparative period and how many weeks that contains. A 52-week year against a 53-week year is not a like-for-like growth rate. Where the counts differ, put both years on a per-week basis before saying which way anything moved.
2. Check which net income you are holding, total or attributable to common shareholders, and which share count sits under any per-share figure, basic or diluted. Write both down before computing anything per share, and remember that each printed per-share figure has already been rounded to the cent.
3. Compare net income with cash from operating activities, for as many years as you have. One year apart is a charge. Several years apart in the same direction is the finding. Cash from operating activities lives in the interim and annual financial statements, which are a separate filing from the news release and sit beside it on SEDAR+, so this step means opening the statements rather than the release. Knowing which document a figure lives in is part of the routine.
4. Decompose the change in the subtotal, not just the subtotal. A pre-tax line that barely moved can sit on top of very large offsetting movements, and the company’s own effect-by-line table is where they are.
5. Name the lines that explain the gap, then check whether any two of them are one obligation seen from two sides. Amortization, revaluations, foreign exchange and impairments all sit in the same reconciliation, and the note underneath will sometimes tell you that two of them moved for one reason.
6. Read the definition of the denominator of every ratio the company hands you, in the footnote under the table it appears in, and check that the definition and the footnote describe the same thing. Then recompute it on the plain consolidated figure and see whether the answer changes.
7. Take every prior-year figure from the most recent filing that carries it. Prior years move, quietly, and a series stitched from originals will not tie to anything.

Both companies’ filings, and every other Canadian reporting issuer’s, are on SEDAR+, which is the statutory filing system and free to search. Dollarama’s own fourth-quarter and fiscal 2026 results release is the document that carries nearly every Dollarama figure on this page, and it is a reasonable first document to run the routine against because the segment, capital allocation and leverage disclosures all sit in the same file. Our rankings of the best Canadian stocks are where this routine gets pointed at a wide list of real companies rather than at two teaching examples.

Frequently asked questions

What is the difference between reading financial statements and analysing them?

Reading tells you what was reported. Analysing separates the reported number into its parts, because four unlike things are blended inside every one of them: what the business already owned produced, what it bought, what accounting recognised without cash moving, and what management did with the cash. Dollarama reported 13.14% consolidated sales growth in Fiscal 2026, which ended February 1, 2026. Reading stops there. Analysing splits it into 6.05 percentage points from the Canadian business and 7.09 from an Australian segment first owned during the year, then reads one row further up the segment column to find that the Australian business earned $5,681 thousand of operating income and was carried to a $245 thousand net loss by $6,053 thousand of net financing costs. Every one of those figures was in the same release.

How do I tell how much of a company’s growth is organic?

Find the segment table and any organic or comparable measure the company publishes itself, and attribute the total growth across existing business, new capacity and acquisitions until the pieces add back to the reported figure. Sometimes the company does most of it for you: Constellation Software publishes its own organic growth rate in its revenue-by-type table, and for FY2021 through FY2025 it reads 7%, negative 1%, 5%, 2% and 4%, compounding to 10.27% across the four year-over-year steps. It publishes the FY2025 figure on two bases, 4% as stated and 3% after adjusting for the impact of changes in the valuation of the US dollar, so quote whichever answers your question and say which one it is. Reported revenue over the same window rose 127.63%, from US$5,106 million to US$11,623 million, with US$6,825 million of cash spent acquiring businesses.

Why would a company’s earnings fall while its cash flow rises?

Because some costs are recognised in earnings without any cash leaving, and because the distance between an operating result and the earnings attributable to common shareholders contains tax and minority interests as well. Constellation Software’s FY2025 is the textbook version: net income to common shareholders fell 30%, from US$731 million to US$512 million, while cash from operating activities rose 24%, from US$2,196 million to US$2,732 million, on revenue up 15%. The adverse movements above the tax line came to 1,856 in the year, of which 852 was marks and currency, including amortization of intangible assets and a US$440 million IRGA revaluation charge. Income before income taxes still fell only from 1,011 to 939, because revenue rose 1,557 in the same year and absorbed nearly all of it. Income tax expense then rose from 244 to 353 and non-controlling interests from 37 to 74, which is how a 72 fall at the pre-tax line became a 219 fall at the shareholder line.

Which net income should I use?

Whichever one matches what you are computing, but you must know which one you are holding. Constellation’s FY2023 filing carries total net income of US$62 million and net income attributable to common shareholders of CSI of US$565 million, in the same table, in the same year, with the difference being non-controlling interests absorbing a loss. Reported earnings per share is calculated on the attributable figure, so that is the row that belongs in any per-share or valuation work on the common shares. A reader who grabs the total row gets a company that barely broke even on US$8,407 million of revenue. The gap does not even keep its sign: total net income is the attributable figure plus non-controlling interests, so in FY2022 the total of US$551 million sat above the US$512 million attributable because the minority interests earned a profit that year, while in FY2023 the total sat far below it because they absorbed a loss.

Is a share buyback good for shareholders?

It depends entirely on the price paid, because a buyback is a purchase, and Dollarama’s weighted average repurchase price rose every year from $77.28 in Fiscal 2023 to $188.47 in Fiscal 2026. The effect on reported per-share growth is real and measurable: in Fiscal 2026 net earnings rose 12.06%, the diluted weighted average share count fell 1.47% from 280,819 thousand to 276,684 thousand, and 1.1206 divided by 0.9853 gives 13.73% against reported diluted EPS growth of 13.70%. Part of that per-share growth was bought rather than earned. Whether it was value created depends on what the shares were worth at each of those prices, which the buyback table cannot tell you.

Why are Dollarama’s 60.1% and 80.05% Latin American holdings not consolidated line by line?

Because control, not ownership percentage, decides how an investment is presented. Dollarama holds 60.1% of the entity it calls CARS and 80.05% of ICM, and the release describes the arrangement as “a joint arrangement using the equity method”. IFRS 11 is the standard that classifies an arrangement as a joint one, which means two or more parties share control under an agreement, and IAS 28 is then the standard that says how to measure it. A majority of the shares does not by itself give the unilateral control that consolidation requires. So the investor books its share of the other company’s profit as earnings, and none of that company’s revenue appears in the investor’s revenue. Dollarama recorded $191,536 thousand of Dollarcity earnings in Fiscal 2026 on that basis, earned over January 1, 2025 to December 31, 2025 rather than over Dollarama’s own fiscal year.

How many years of financials should I look at?

Four or five, taken from the most recent filing that carries each figure. One year cannot show you a shape, and the shape is usually the finding. The reason to take prior years from the latest filing rather than the original is that they move: the Fiscal 2026 Dollarama release states Fiscal 2025 capital expenditures as $243,450 thousand where the Fiscal 2025 release had stated $246,869 thousand, a $3,419 thousand restatement that is trivial in size and useful as a warning. Constellation’s December 31, 2024 debt with recourse to CSI moved from US$2,159 million in one MD&A to US$1,466 million in the next, which is the same warning at a size that would change a reader’s view of the balance sheet. Check the period lengths across the years you gather, too, because a 53-week year inside a four-year series will bend the trend on its own.

What is the first thing to check on a set of results?

Find the segment table, because if the company reports more than one segment then every consolidated percentage you are about to read is a blended average of businesses that are not alike. Dollarama’s reported operating margin barely moved, from 26.675% to 26.708%, while the same margin excluding equity-accounted earnings fell from 24.649% to 24.068% and the Canadian segment on that basis came in at 25.594%. The reason is visible one table further in: the Canadian stores ran a 45.61% gross margin against Australia’s 36.61%, spent 14.42% of sales on SG&A against Australia’s 24.71%, and carried 5.60% of sales in depreciation and amortization against Australia’s 10.65%. Three defensible answers, one year, one release.

Neither company named on this page is a recommendation, nothing here is a view on either one’s shares, and every figure is tied to the specific fiscal period and filing named beside it.