How to Compare Companies Across 10 Years of Filings
To compare a company across a decade, normalise for restatements, segment redefinitions, and accounting changes first, so every year is measured on the same basis before you read the trend.
To compare a company across ten years of filings, first put every year on the same basis, then read the trend. Line up the same segment definitions, adjust for accounting-standard changes and demergers, keep the consolidated-versus-standalone choice constant, and rebase per-share figures to a common share count. Only when the series is genuinely apples to apples should you draw a conclusion from it.
This sounds obvious and is almost never done. Most long-run comparisons pull ten numbers off a screen, plot them, and treat the line as truth. The problem is that a reported figure is not a fixed fact across time. The basis underneath it moves, and if you do not account for that, you can read a trend that the business never actually lived.
Why raw ten-year series lie
When a company reports a quarter or a year, it does more than publish new numbers. It often restates the prior-year comparable so the two periods sit on the same footing. That restatement can come from an accounting-standard change, a demerger or spin-off, a discontinued operation, or a redefinition of how segments are drawn. Each of these can shift a reported line without a single thing changing in the underlying business.
The consequence matters for two different jobs. If you are reading a clean year-on-year change, the restated comparable is exactly what you want, because it holds the basis constant. But if you are studying a full decade, you are stitching together numbers that were each published on the basis of their own moment, and those bases are not the same. The history you see on a screen today is not always the history that was knowable on the date an older decision was made. Testing an idea on “as reported today” data can quietly flatter the result, which is the heart of lookahead bias and the reason point-in-time data matters.
A ten-year chart is only as honest as the basis under each dot.
So the work is not plotting the series. The work is making the series comparable before you plot it.
The five things that break a long comparison
Five recurring changes will distort a decade of filings if you ignore them. Each has a standard fix.
| What changes | Why it distorts the trend | How to normalise |
|---|---|---|
| Accounting standards | A new standard can move revenue, lease costs, or how items are grouped | Note the adoption date; read the transition disclosures; keep the pre and post years mentally separate |
| Segment redefinitions | Divisions get merged, split, or renamed, so a segment in 2016 is not the same box in 2025 | Rebuild the longest consistent segment map you can; group where you must |
| Demergers and spin-offs | A division leaves the group, shrinking the topline with no operational decline | Read continuing operations; strip the demerged unit from earlier years if you want a like-for-like path |
| Consolidated versus standalone | Mixing the two across years makes revenue, debt, and margins incomparable | Pick one basis and hold it for the entire window |
| Share count changes | Bonuses, splits, buybacks, and raises move per-share figures with no change in the business | Rebase EPS and per-share book value to a common share count |
None of this is exotic. It is the discipline of reading the notes rather than the headline figure, which is a theme running through how to read an annual report. The notes to the accounts are where a company tells you, often in a single dry line, that this year is not measured the way last year was.
Start with segments, because that is where mix hides
Segment reporting is usually the first thing to normalise, because a consolidated topline blends businesses that behave nothing alike. A single group can hold one segment that is enormous in revenue but thin in margin, and another that is small in revenue but very high in margin. Read only the group total and you see none of that, and a ten-year comparison of group revenue tells you almost nothing about which engine is actually driving the change.
To make the point concrete, look at one company’s public quarterly segment disclosure for the quarter ended September 2025. The figures below are approximate and shown only to illustrate the method, not to judge the company.
| Segment | Revenue (approx) | Segment EBIT margin (approx) |
|---|---|---|
| Oil to chemicals | ₹1,60,600 crore | 9% |
| Retail | ₹90,000 crore | 6% |
| Digital services | ₹42,700 crore | 52% |
| Oil and gas (upstream) | ₹6,100 crore | 83% |
The biggest revenue segment is one of the thinnest by margin, and the fattest-margin segments are small by revenue. Now imagine tracking this group over ten years while the segment boundaries were redrawn once or twice along the way. If you compare “revenue” year to year without holding the segment map constant, you are comparing different mixtures of these very different economics. The trend you read is partly a story about accounting definitions, not the business. This is why the longitudinal method leans so heavily on getting segments right, and it is the practical companion to the mechanics covered in segment analysis explained.
Then rebuild the series you actually want
Once the basis is under control, you rebuild a small set of series that survive a decade of scrutiny. Growth, so you can see whether it is steady or lumpy. Operating margin, so you can see whether the business held its economics through good years and bad. Return on capital, so you can see whether returns persisted or faded. And cash conversion, because profit is an opinion and cash is a fact, so a decade of profit that never turned into cash is telling you something a single year would hide.
Read those four together and over a cycle. A cycle matters because a single year flatters whoever had the good year and punishes whoever had the bad one. Ten years lets one-off periods reveal themselves as one-offs rather than trend. It also lets you see the difference between a business whose margins wobble a few points around a stable centre and one whose margins are slowly, quietly eroding.
Company against itself, then against peers
There are two comparisons folded into this, and they are not the same job.
The first is the company against its own past. This is the cleaner of the two, because you are holding one accounting philosophy roughly constant and mainly fighting restatements and definition changes within one filer. The question is whether the business is better, worse, or the same than it was, measured honestly.
The second is the company against peers over the same decade. This is harder, because now you are reconciling two or more filers who may sit on different accounting choices, different segment maps, and different consolidation decisions. Everything in the single-company case still applies, and on top of it you have to line the two histories up with each other. The snapshot version of this, comparing two companies as they stand today, is covered in how to compare two companies; the ten-year version simply insists that both histories be normalised before either column is trusted.
The failure mode is the same in both cases. Someone puts two ten-year lines on one chart, the lines cross somewhere, and a conclusion gets drawn from the crossing point. If the two series were never on the same basis, the crossing point is an artifact of accounting, not a fact about the businesses.
A short, repeatable routine
The method reduces to a sequence you can run on any name.
First, fix the window and the basis: choose consolidated or standalone and commit to it for all ten years. Second, walk the notes for standard adoptions, demergers, discontinued operations, and segment redefinitions, and write down every date one occurred. Third, rebuild the longest consistent segment map the disclosures allow. Fourth, rebase per-share figures to a common share count. Fifth, and only now, plot growth, margin, returns, and cash conversion, and read the trend. Sixth, if you are comparing peers, repeat all of the above for each one before you overlay them.
It is slower than pulling a screen. It is also the difference between a chart that documents a business and a chart that documents a filing system.
The quiet payoff
A decade of filings, read carefully, answers a question a single year cannot: is this a durable machine or a good moment? The answer lives in whether margins held, whether returns persisted, whether growth was steady, and whether profit kept turning into cash, all measured on a basis that does not shift under your feet. The normalisation work is unglamorous, and it is exactly the part that separates a comparison you can rely on from one that merely looks rigorous. Do the boring part first, and the trend you read afterward is one the company actually lived.
Frequently asked questions
How do you compare a company across 10 years of filings?
Build the history on a consistent basis before you read it. Line up the same segments, adjust for accounting-standard changes and demergers, use the same consolidated-versus-standalone choice throughout, and put per-share numbers on a common share count. Only after the series is apples to apples should you look at the trend in growth, margins, returns, and cash.
Why can't I just use the numbers as reported each year?
Because reported numbers change basis over time. Companies adopt new accounting standards, redefine segments, and spin off divisions, and each of those can move the reported figure without anything changing in the actual business. A raw ten-year series can mix apples and oranges and produce a trend that never really happened.
What is a restated comparable?
When a company reports a period, it usually also republishes the prior-year figure on the new basis so the two are comparable. That republished prior-year number is the restated comparable. It is useful for a clean year-on-year read, but it means the history you see today is not always the history that was on the page when an older decision was made.
What can a ten-year history show that a single year cannot?
Durability. One year shows a level; a decade shows whether margins hold through a cycle, whether growth is steady or lumpy, whether returns on capital persist, and whether profit reliably becomes cash. It also exposes one-off years for what they were, which a snapshot cannot.