Skip to main content

Mean vs Median in a Comp Set

You have built your peer set. You have adjusted for size, growth, and quality. Now you have ten peer multiples, and they range from 8x to 22x EV/EBITDA. Should you use the average (mean) of 15x, the middle value (median) of 14x, or something else?

This seemingly technical question shapes your entire valuation. A single outlier can swing the mean by a full point, but the median ignores it. Neither approach is always correct. This article covers when each is appropriate, how to handle outliers, and how to communicate your aggregation choice defensibly.

Quick Definition​

The mean (average) is the sum of all multiples divided by the count. The median is the middle value when multiples are sorted. In a peer set of 11 companies, the median is the 6th value; the mean is the average of all 11.

The median is more robust to outliers; a single extreme value barely affects it. The mean is sensitive to outliers and can be distorted by one or two peers with unusual multiples.

Key Takeaways​

  • The median is typically the superior anchor for comparable company analysis because it is resistant to outliers.
  • Outliers are common in peer sets and often signal that a peer is not truly comparable; before using the mean, ask why the outlier exists.
  • A trimmed mean (excluding the highest and lowest values, or the highest and lowest 10%) is a useful compromise between median and full mean.
  • The size of your peer set matters: a set of 6 companies is more sensitive to outliers than a set of 20.
  • Documenting which aggregation method you use, and why, is as important as the number itself.

The Median vs. Mean: A Statistical Perspective​

Example: A Simple Peer Set​

Imagine a peer set of five industrial companies with EV/EBITDA multiples:

Company A: 9x Company B: 11x Company C: 13x (median value) Company D: 15x Company E: 45x (an outlier)

Mean = (9 + 11 + 13 + 15 + 45) / 5 = 93 / 5 = 18.6x

Median = The middle value when sorted = 13x

The mean is pulled up 5.6x by a single outlier. Should Company E's extreme multiple influence your valuation at all?

Before deciding, ask: Why is Company E trading at 45x when the peer set averages 13x?

Possible answers:

  1. Company E is not truly comparable. Perhaps it has a unique business model, much higher growth, or much higher ROIC. If so, exclude it from the peer set entirely.
  2. Company E is mispriced. Perhaps the market is irrationally bullish or bearish on it. If so, the median (13x) is more representative of fair value for peers, and your target should not be valued at 45x simply because one peer is.
  3. Company E is a special situation. Pending M&A, restructuring, or a unique competitive advantage. If so, note it and use the median, understanding that Company E is not a true comparable.

In most cases, a single extreme outlier reflects one of these issues, and the median is the more appropriate anchor.

When the Mean Is Appropriate​

The mean becomes more appropriate when:

  1. All peers are genuinely comparable and legitimate. If your peer set of 20 companies are all well-matched to your target (same industry, similar size, similar growth, similar quality), and the multiples range from 11x to 17x EV/EBITDA with no outliers, the mean and median will be close (roughly 13–14x), and either is defensible.

  2. You are adjusting each peer individually. If you have downweighted or excluded outliers before computing the mean, the remaining set is more compact and homogeneous. The mean of a filtered set is more robust.

  3. You want to reflect all available market data. Some argue that even an outlier represents real market opinion and should be weighted. If you believe all peer multiples have informational value, use the mean. This is weaker reasoning but can be valid for educational purposes.

When the Median Is Superior​

The median is preferable in almost all practical scenarios because:

  1. It is outlier-resistant. One or two extreme values barely affect the median. In a set of 11 peers, you could add or remove a 50x peer and the median might move from 14x to 14.2x.

  2. It reflects the "typical" peer. By definition, the median is the point where half the peers are above and half below. It represents the central tendency better than a mean distorted by outliers.

  3. It does not require assumptions about outlier causes. With the mean, you must decide whether to exclude the outlier, which is a judgment call. The median sidesteps that judgment by being naturally resistant.

  4. It is more stable across peer set updates. If you add a new peer and its multiple is in the normal range, the median is barely affected. The mean could shift significantly.

Outliers: To Include, Exclude, or Downweight?​

Before you choose between mean and median, decide what to do with outliers.

Step 1: Identify Outliers​

Use statistical tests:

  • Interquartile range (IQR) method: Compute Q1 (25th percentile) and Q3 (75th percentile). Values below Q1 - 1.5 × IQR or above Q3 + 1.5 × IQR are potential outliers.
  • Z-score method: Calculate how many standard deviations each value is from the mean. Values with |Z| > 2 or 3 are outliers.
  • Visual inspection: Sort multiples and look for gaps. A peer trading at 45x in a set of 8–15x is obviously an outlier.

Step 2: Investigate the Outlier​

Do not automatically exclude it. Understand why it is an outlier:

  • Does the outlier have higher growth or ROIC than the peer set? (If so, it may deserve to be priced differently; document this and either exclude it or adjust your target's multiple accordingly.)
  • Is it a different stage (early-stage growth vs. mature)? (If so, it is not a true comparable; exclude it.)
  • Is it known to be in a buyout discussion or restructuring? (If so, its multiple is not representative of normal trading; exclude it.)
  • Does it have drastically different margins or capital intensity? (If so, it is not comparable; exclude it.)

Step 3: Make a Decision​

Exclude: If the outlier is not truly comparable (different industry segment, different business model, different stage), exclude it from the peer set entirely. Do not include it in mean or median.

Downweight: If the outlier is comparable but unusual (e.g., temporarily depressed due to cyclical downturn, or temporarily elevated due to special situation), downweight it. Assign it 50% of normal weight when computing the median or mean.

Include at full weight: If the outlier is a legitimate peer but just happens to trade differently, include it. Use the median to reduce its influence.

A Mermaid Decision Tree: Aggregating Peer Multiples​

Practical Approaches: Beyond Simple Mean or Median​

Approach 1: The Trimmed Mean​

Compute the mean but exclude the highest and lowest values (or the highest and lowest 10%). This combines the simplicity of the mean with outlier resistance.

Example: Eight companies with EV/EBITDA of 9x, 10x, 12x, 13x, 14x, 15x, 16x, 45x.

  • Full mean: (9+10+12+13+14+15+16+45) / 8 = 16.75x
  • Median: 13.5x
  • Trimmed mean (exclude high and low): (10+12+13+14+15+16) / 6 = 13.33x

The trimmed mean (13.33x) is very close to the median and filters out the 45x outlier without requiring exclusion logic.

Approach 2: Weighted Mean by Peer Strength​

Assign weights to each peer based on how directly comparable it is:

  • Direct comparables (same size, growth, quality): 100% weight
  • Good comparables (some minor differences): 75% weight
  • Partial comparables (meaningful differences but in same industry): 50% weight
  • Weak comparables or outliers: 25% weight or exclude entirely

Compute a weighted mean using these weights. This is more sophisticated and requires clear documentation but can feel more rigorous.

Approach 3: Median + Range​

Instead of picking a single number, report the median and also the range (e.g., 25th to 75th percentile, or the range excluding the top and bottom 10%). This communicates both the central point and the uncertainty.

Example: "Based on 12 peers, the median EV/EBITDA is 13x, with a range of 11–15x (interquartile range). Given our target's superior profitability, we apply 13x to conservative estimates and 15x to base case estimates."

This approach is honest about uncertainty and gives the user context for how the multiple sits in the distribution.

Real-World Examples: Mean vs. Median Choices​

Example 1: Semiconductor Equipment Peers​

You are valuing a semiconductor equipment manufacturer. Your peer set of 10 companies has EV/EBITDA multiples:

8x, 9x, 10x, 11x, 12x, 13x, 13x, 14x, 15x, 38x

The 38x outlier is a company (let's call it "HighFlyer") that has won a major new contract and is expected to grow at 30% for the next 3 years. Its competitors grow at 8–12%.

The right approach:

  1. Recognize that HighFlyer is not a true comparable because its growth profile is materially different.
  2. Either exclude HighFlyer entirely, or include it but note that its multiple is not representative of the base peer set.
  3. Use the median of the remaining 9 (12.5x) or the median of all 10 (12.5x, since the 38x is far from the center).
  4. For your target, if it has similar growth to the 8–15x peers, apply a multiple in that range. If your target is closer to HighFlyer's growth profile, consider applying a higher multiple, but justify it separately (not by averaging it into the peer set).

The mean of 12.4x would be skewed by HighFlyer. The median of 12.5x is more representative.

Example 2: Retail Peer Set with One Outlier in Turnaround​

You are valuing a mid-market retailer. Your peer set of 8 companies has EV/Sales multiples:

0.4x, 0.5x, 0.6x, 0.7x, 0.8x, 0.9x, 1.0x, 2.5x

The 2.5x outlier is a competitor undergoing a turnaround and currently trading at an inflated multiple pending strategic clarity.

The right approach:

  1. Recognize that the 2.5x outlier is in a special situation and is not a normal trading comparable.
  2. Downweight or exclude it. If you exclude it, the median of the remaining 7 is 0.7x. If you downweight it to 50%, it has less influence but is still in the set.
  3. Use the median of the "normal" retailers (0.7x) as your anchor.
  4. Note that if your target is also in a turnaround, you might apply a higher multiple, but again, justify it separately.

The mean of all 8 (1.19x) is heavily skewed by the 2.5x outlier and is not representative of normal retail trading multiples.

Example 3: A Tight Peer Set with No Outliers​

You are valuing a large-cap consumer staples company. Your peer set of 15 companies has P/E multiples:

18x, 19x, 19x, 20x, 20x, 20x, 21x, 21x, 21x, 22x, 22x, 23x, 23x, 24x, 24x

The right approach:

There are no outliers. Mean (20.93x) and median (21x) are nearly identical. Either is defensible. Use the median (21x) as your primary anchor, note the range (18–24x), and apply professional judgment about where your target sits in that range based on relative growth and quality.

Common Mistakes with Mean vs. Median​

Mistake 1: Using the Mean Without Questioning Outliers​

Computing a mean on a peer set with one extreme outlier and using it as gospel is sloppy. Always inspect the mean and median; if they differ by more than 10%, investigate why.

Mistake 2: Arbitrarily Excluding Outliers​

The opposite extreme is to exclude peers simply because they do not fit your narrative. Exclude or downweight outliers only if you have a sound reason (genuinely not comparable, special situation, different business model). Document your reasoning.

Mistake 3: Using the Mean on a Small Peer Set​

A peer set of 5–7 companies is inherently unstable; a single outlier has huge influence. Use the median for small sets. For large sets of 15+, the mean and median converge and both are reasonable.

Mistake 4: Not Acknowledging Uncertainty​

Whether you use mean or median, the multiple is not a precise number. It is a point on a range. Communicate that range (25th to 75th percentile, or high and low multiples), not just the central point.

Mistake 5: Forgetting That the Median is Not the Target's Multiple​

The peer median is an anchor for your target's multiple, not the answer itself. If your target has higher quality (ROIC) or growth than the typical peer, apply a multiple above the median. If it is lower quality, apply a multiple below the median. The median is the starting point, not the ending point.

FAQ​

Q: Is there a standard rule for which aggregation method to use?

A: No universal rule, but the median is the default choice for most comparable company analysis. Use the mean only if you have a clear reason (e.g., deliberately weighting all available market data, or a peer set with no outliers).

Q: What if my peer set has only 5 companies?

A: With 5 companies, the median is the 3rd value, and a single outlier can shift it by 20%. Ensure all 5 are truly comparable. If there is any doubt about one peer, exclude it and use a set of 4. Or, use a trimmed mean (exclude the highest and lowest) to get a more robust central point.

Q: Should I report both mean and median, or just one?

A: Reporting both is transparent. It shows the reader that you are aware of the aggregation choice and gives them context. Note: "Median EV/EBITDA of 12x (mean 13x, range 10–16x)."

Q: Can I use different aggregation methods for different multiples?

A: Yes, if justified. For example, you might use the median EV/EBITDA (because outliers are present) and the mean P/E (because the peer set for P/E is tighter). Document this clearly.

Q: What if the median and mean are far apart?

A: That signals outliers. Investigate why. Identify the extreme peers and decide whether to exclude them. If the mean and median differ by more than 10–15%, something is off.

Summary​

The median is the preferred aggregation method for peer multiples in most cases because it is resistant to outliers and represents the central tendency of the peer set. The mean is useful only when all peers are directly comparable and there are no extreme outliers, or when you deliberately want to weight all available data.

Before choosing between mean and median, identify outliers and investigate their causes. Exclude peers that are not truly comparable; downweight peers that are comparable but in special situations; include at full weight peers that are legitimate comparables but happen to trade differently.

For small peer sets (5–10 companies), use the median. For large peer sets (15+), mean and median converge and both are defensible. Always report the range along with the central point, document your aggregation method, and remember that the peer median is a starting point, not your target's final multiple—adjust up or down based on relative quality and growth.

Next​

We now move to Precedent transaction multiples, where we explore how to use M&A transaction data to value companies and why precedent multiples often tell a different story than trading multiples.