Five illustrative peers, one outlier, and the drop-one test that shows how fragile a small comparable set really is.
Use the median, and show the interquartile range beside it. In a set of five illustrative peers trading at 7.1x to 13.6x EBITDA, one outlier lifts the mean to 9.14x against a median of 8.20x; on 6.5m of EBITDA that is an enterprise value of 59.41m against 53.30m, and an equity value 15.5 per cent higher on the mean. Drop any single peer and the mean moves across a 20.2 per cent range, the median across 7.5 per cent.
Worked in full in Business Valuation by Julian R. Sterling, with every figure reproduced in a free workbook.See the book on Amazon →
Multiples are ratios, and ratio distributions are skewed: there is a floor near zero and no ceiling. One peer with depressed earnings, a takeover bid in its price or a growth story the others do not share drags the average up. The median ignores how far the outlier sits from the rest; the mean does not.
| Peer | EV/EBITDA | Implied EV on 6.5m |
|---|---|---|
| Peer A | 7.1x | 46.15 |
| Peer B | 7.8x | 50.70 |
| Peer C | 8.2x | 53.30 |
| Peer D | 9.0x | 58.50 |
| Peer E (recent bid speculation) | 13.6x | 88.40 |
The subject company has EBITDA of 6.5m and net debt of 14.0m.
Step 1: the four central measures.
Mean = (7.1 + 7.8 + 8.2 + 9.0 + 13.6) / 5 = 9.14x
Median = middle value of the sorted set = 8.20x
Harmonic mean = 5 / Σ(1 / multiple) = 8.69x
Trimmed mean (drop highest and lowest) = (7.8 + 8.2 + 9.0) / 3 = 8.33x
In Excel: =AVERAGE(B2:B6), =MEDIAN(B2:B6), =HARMEAN(B2:B6), =TRIMMEAN(B2:B6,0.4), and =QUARTILE.INC(B2:B6,1) and ,3) for the quartiles.
The harmonic mean is the average of the earnings yields, 11.51 per cent, turned back into a multiple. It is the right average when the ratio has price on top, because averaging yields gives each company's earnings equal weight per unit of value. It still sits between the median and the mean here, because the outlier keeps some pull.
Step 2: apply each to the subject and bridge to equity.
| Measure | Multiple | Enterprise value | Equity value |
|---|---|---|---|
| First quartile | 7.80x | 50.70 | 36.70 |
| Median | 8.20x | 53.30 | 39.30 |
| Trimmed mean | 8.33x | 54.17 | 40.17 |
| Harmonic mean | 8.69x | 56.46 | 42.46 |
| Third quartile | 9.00x | 58.50 | 44.50 |
| Mean | 9.14x | 59.41 | 45.41 |
The mean puts enterprise value 6.11m, or 11.5 per cent, above the median. Leverage magnifies it: the equity gap is 15.5 per cent, because net debt is a fixed deduction from both. The mean sits above the third quartile, which is a clear sign that a single company is setting it. The defensible statement is "8.20x, within an interquartile range of 7.80x to 9.00x", which puts enterprise value between 50.70m and 58.50m.
| Peer removed | Mean | Median | EV on median |
|---|---|---|---|
| Peer A | 9.65x | 8.60x | 55.90 |
| Peer B | 9.47x | 8.60x | 55.90 |
| Peer C | 9.38x | 8.40x | 54.60 |
| Peer D | 9.17x | 8.00x | 52.00 |
| Peer E | 8.03x | 8.00x | 52.00 |
The drop-one test is the honest measure of a small set's fragility. The mean ranges from 8.03x to 9.65x, a spread of 20.2 per cent; the median from 8.00x to 8.60x, 7.5 per cent. Remove the outlier and mean and median almost coincide, at 8.03x and 8.00x, which confirms that one name was doing all the work.
The median is robust, not immune. With four peers left, it becomes the average of the middle two, so removing any name moves it. Add a sixth peer at 10.5x and the median rises from 8.20x to 8.60x, lifting enterprise value 4.9 per cent to 55.90m, from a company that changes nothing about the subject. With five or six names, the choice of who is in the set matters more than the choice of statistic.
Report the drop-one range with the answer. A reviewer who sees 52.00m to 55.90m on the median knows how much weight the selection of peers carries before arguing about any single one.
Adjusting the chosen multiple for those differences, and normalising the EBITDA it is applied to, is worked in how to value a private company with an EBITDA multiple. Why a transaction multiple then lands above the trading median is the subject of how to calculate a control premium.
Lead with the median, 8.20x here, report the interquartile range and the drop-one range, and use the mean only as a test: when it sits above the third quartile, an outlier is setting it. The free workbook for this book carries a peer set and a transaction set with their medians, quartiles and the effect of dropping any one name.
Because multiples are skewed and a single outlier pulls the mean. In the illustrative set one peer at 13.6x lifts the mean to 9.14x, above the 9.00x third quartile, while the median stays at 8.20x. Removing that peer moves the mean to 8.03x and the median only to 8.00x.
When averaging price-based ratios such as EV/EBITDA or P/E, because it averages the earnings yields, which is the same as putting an equal amount of value into each company. In the illustrative set the average yield is 11.51 per cent, a harmonic mean of 8.69x, between the 8.20x median and the 9.14x mean.
Enough that removing one does not move the answer much, which five rarely achieves. In the illustrative set the drop-one median runs from 8.00x to 8.60x, and adding a sixth peer at 10.5x raises the median to 8.60x and enterprise value by 4.9 per cent.
This article is one calculation from Business Valuation. The book takes the same case from first principles to the decision, chapter by chapter, and every figure it prints is a live formula in the free companion workbooks.
Get the book on Amazon →Free companion files
Also on Amazon UK · Amazon Germany · Amazon France · Amazon Canada
Reading guide: corporate finance, valuation and markets → · All 1340 articles →
If this book helped, or didn’t, a few lines on Amazon are worth more than they look: they are what the next reader goes on. Write a review. The workbook stays free either way.