Lesson 4 of 7
Normalising a capacity business
The average that hides the subject. Two defensible ways to normalise nine filed years, and the one question both of them answer badly.
By Javier Sanz
Founder of ValueScreener. Built Ninety Nine, a securities services provider, and sold it to a tier-one global bank; ran operations and internationalization at Alpaca, the brokerage API.
Who writes thisEditorial processLinkedIn (opens in a new tab)X (opens in a new tab)
In one minute
- Normalising means removing the part of a result that is the cycle, and at a cyclical business that is most of it.
- Two defensible methods here: a mean margin applied to current revenue, and a mean margin applied to a normalised revenue. They give different answers.
- Neither method can tell you whether the window contained a representative cycle, which at this company it did not.
The table is built and the absences are read. This lesson produces the one figure the next capstone needs, and it produces it twice, because there is more than one defensible way to do it.
Operating margin against its nine-year mean
- FY1711.13%
- FY1811.01%
- FY199.42%
- FY207.77%
- FY2111.61%
- FY2214.68%
- FY2317.26%
- FY2418.81%
- FY2515.69%
Nine-year mean, 13.04%: 13.04%
What normalising is for
To answer the question: what does this business earn in an ordinary year. Not the best year, not the worst, and not the most recent, which at a swinging business is whichever of the three the calendar happened to deliver.
The method is always the same shape: take a measure over a window, average it, and apply the average to something current. What differs is which measure is averaged and what it is applied to.
Method one: mean margin on current revenue
Average the nine annual operating margins: 13.04 percent. Apply it to the most recent year's revenue of 10,966.7 million dollars. That gives a normalised operating income of 1,430.1 million.
Reported operating income in that year was 1,720.6 million, so this method says the latest year earned about 290.5 million more than an ordinary year on the same revenue, which is 20.3 percent above normal.
The assumption inside it is that current revenue is itself ordinary, which at a business whose revenue is part of what cycles is a large assumption.
Method two: mean margin on mean revenue
Average the nine years of revenue: 8,792.9 million dollars. Apply the same 13.04 percent margin. That gives 1,146.6 million.
This method assumes neither the margin nor the revenue of the latest year is ordinary, which is more conservative and introduces a different problem: mean revenue over a window in which revenue nearly doubled is not a level the business was ever at, it is the midpoint of a trend.
The question neither method answers
Whether the window contained a representative cycle. The first lesson of this course established that it did not: nine years with two shallow down years, and a 52.5 percent contraction in the company's own earlier record sitting outside the window.
So both figures above are averages of an expansion. A normalised figure computed over a window without a bad year in it is not a cycle average, and calling it one would be the error the whole course is designed to avoid.
What to write down
The method, the window, the figure, and the sentence that says what the window lacks. Four items. The next capstone takes the figure and the sentence together, because a value built on a normalised figure inherits everything that figure does not contain.
Check yourself
What does normalising a cyclical business try to produce?
- The best estimate of next year's result
- What the business earns in an ordinary year, rather than the most recent year whichever kind that happened to be
- The average of the peak and the trough
Show the answer
Answer: B.
The method is a measure averaged over a window and applied to something current. Which measure and which current figure are what differ between methods.Two methods give 1,430.1 and 1,146.6 million against 1,720.6 reported. What should be published?
- The most conservative of the three
- All three with their methods, because the spread is the honest output
- The reported figure, since it is audited
Show the answer
Answer: B.
The highest is 50.1 percent above the lowest, all from the same nine filings. A single figure without the spread reports the least useful part of the work.What can neither method tell you?
- Which year was the peak
- Whether the window contained a representative cycle, which here it did not
- What the margin was in any single year
Show the answer
Answer: B.
Nine years with two shallow down years is an expansion. A normalised figure computed over such a window is not a cycle average.Sources
- Toll Brothers, Inc., annual reports on Form 10-K for fiscal years 2017 through 2025, for revenue and operating income; fiscal 2025 accession 0000794170-25-000112, filed 19 December 2025.
What this leaves out
- Both methods use an unweighted arithmetic mean of annual figures, which weights a small year the same as a large one. A weighted average gives different answers and is equally defensible if stated.
- The window contains no contraction of the kind this industry has experienced, so neither normalised figure is a cycle average and neither should be described as one.
- A normalised figure is an input to a judgement, not a forecast. Nothing here says what this or any company will earn in any future year.
- Nothing here is a view on Toll Brothers or on any security.