Skip to main content

Building a Bad Peer Set

Relative valuation—comparing a company's P/E, EV/EBITDA, or other multiples to a peer group—is the most widely used tool in sell-side analysis and institutional investing. It's intuitive: if Competitor A trades at 15× earnings and Competitor B at 18×, and your stock trades at 12×, maybe it's cheap. It's also mechanically simple, requiring fewer assumptions than a DCF model.

That simplicity is deceptive. A bad peer set does more damage than a bad DCF because it seems easier to defend. You can point to the names, show their multiples, and anchor your valuation to a benchmark. Nobody asks whether the peers actually compare.

The most common error is confusing a peer set with a peer group. A peer set should be a tightly curated list of companies that compete for the same customers, use similar business models, face comparable growth constraints, and operate with similar profitability profiles. A peer group is what most analysts build: everyone in the industry, plus a few adjacent names, screened for size and liquidity, then averaged together.

The difference is fatal. A bad peer set creates an anchoring trap where the "fair value" you derive is just a reprise of whatever the market decided the peer average was worth last quarter.

Quick definition: A peer set is a list of comparable companies whose valuations inform a relative valuation for the stock you're analyzing. A good peer set consists of companies with similar competitive positions, growth rates, profitability, and capital structures. A bad peer set is too large, too diverse, or too anchored to index benchmarks.

Key Takeaways​

  • Most analyst peer sets are too large (15–25 companies) and too diverse; the average obscures the true comparables and anchors valuation to consensus.
  • Screening by market cap, liquidity, or region alone creates a set of financially similar but strategically different companies that don't actually compete.
  • Peer sets often include companies at radically different growth stages (mature players, high-flyers, zombies) which have no reason to trade at the same multiple.
  • The "industry average" multiple is often the result of a margin or growth divergence between peers, not a true fair value baseline.
  • Using a bad peer set as the anchor for valuation makes it nearly impossible to identify when the peer group itself is overvalued or undervalued.

Why Peer Sets Matter, and Why They Fail​

Relative valuation is powerful because it answers the question: "Valuable relative to what?" If everyone in an industry is priced at 20× earnings, a stock trading at 15× looks cheap until you realize the industry is overvalued at 20×. Without that context, you've mistaken relative cheapness for absolute cheapness.

But that context only exists if your peer set contains true comparables. The moment you add a company that is strategically or operationally different, the average becomes contaminated. Include one hypergrowth SaaS company in a peer set of mature software vendors, and the average P/E rises to 35×. Now the mature vendors look cheap, even if they're actually expensive. The bad peer dilutes the signal.

The mechanical pressure to include too many peers comes from the need for statistical credibility. An analyst who compares one stock to three peers is vulnerable to criticism: "That's cherry-picking; you should use the whole industry." So they expand to 15 peers, then 25. The peer set becomes an index proxy. When you average across a large, diverse group, you lose the ability to identify genuine mispricings; you just echo the market's sentiment.

The Mermaid: Peer Set Contamination and Valuation Distortion​

Common Peer Set Mistakes​

1. Including Too Many Names and Averaging Them​

The classic institutional approach: get every company in the SIC code, or all stocks in the industry basket, apply a liquidity filter, and average the multiples. This yields a defensible, index-like benchmark.

But defensibility is not accuracy. By averaging, you collapse meaningful differences. A software company with 20% free cash flow margins shouldn't be traded at the same multiple as one with 8% margins, yet if both are in your peer set, you're implicitly saying they're equivalent.

Worse, large peer sets introduce a drift toward consensus. If the market has already priced the 15 largest software companies at a certain multiple, your peer average will echo that consensus. You've outsourced your valuation to the market's collective decision, which defeats the purpose of analysis.

2. Mixing Different Business Models or Revenue Streams​

A classic example: telecom companies. Some are growing fiber networks and selling broadband; others are shrinking legacy wireline businesses and returning cash; still others are mobile-focused. They compete in some areas but not others. An analyst who averages all of them to get a "fair" P/E for a new entrant is blending incomparable things.

Similarly, in cloud computing, a company with 30% subscription revenue and 70% perpetual software licenses should not be valued at the same multiple as a pure-SaaS company with 95% recurring revenue. They have different growth ceilings and cash conversion profiles. Yet peer sets routinely mix them.

3. Ignoring Growth Rate Divergence​

A $100 billion pharmaceutical company growing 3% and a $5 billion biotech growing 25% might both be in the same "pharma" peer set. They have completely different valuation profiles: the large cap deserves a lower multiple. Yet analysts often treat them as comparables and average their multiples, contaminating both valuations.

A good peer set accounts for growth. If your target grows at 10%, its peers should also grow at 10%, ±3 percentage points. If they don't, you're comparing apples to oranges, and the "average" multiple tells you nothing.

4. Using Geographic or Sector Indices as Peer Sets​

"The NASDAQ 100 is the peer set for tech stocks." This is an abdication. The NASDAQ 100 includes megacaps, growth stocks, and value stocks. A semiconductor fabless designer and a semiconductor manufacturer should have different multiples, but they're blended into the index average.

Similarly, using "all large-cap industrial companies" as peers for a specific manufacturer is too broad. Industrial equipment makers, component suppliers, and fully integrated manufacturers face different competitive dynamics and should be valued differently.

5. Anchoring to Peer Average Without Adjusting for Profitability​

You find that your peer set averages 18× P/E. Your stock trades at 16×. Looks cheap, right? But if your stock is trading at 16× because its margins are 2 percentage points below peers' due to operational issues, then 16× might actually be expensive.

An analyst who reaches for a simple "our stock is at 16× and peers are at 18×, so it's 11% undervalued" is making an implicit assumption that the peer average is fair and that your stock's discount is temporary. Often, the discount exists for a reason. A bad peer set analysis doesn't interrogate that reason; it just assumes it away.

6. Treating Outliers as Peers Rather Than Red Flags​

In a peer set of 15 companies, two might trade at half the average multiple while two trade at double. An analyst averaging them together implicitly says all five are equivalent, which is almost certainly wrong.

The outliers—the company trading at 8× when peers average 18×—isn't "cheap"; it's often cheap for a reason: lower growth, lower margins, deteriorating competitive position. Including it in the average drags down your benchmark and creates the illusion that fair value is lower than it is.

A better approach: investigate the outliers. Why is Company X trading at 25×? Is it higher growth, higher margins, or irrational exuberance? Why is Company Y at 9×? Is it deteriorating, or is it genuinely mispriced? Once you understand the outliers, you can build a more defensible core peer set that excludes the extremes or explains them.

Real-World Examples​

Streaming Video and the Peer Set Trap (2020–2023)

In 2020–2021, Netflix, Disney+, Amazon Prime Video, and Apple TV+ were all lumped into a "streaming" peer set. But they were radically different:

  • Netflix was a pure-play streamer, profitable, generating positive free cash flow.
  • Disney+ was a subsidiary of a media conglomerate, subsidized by Disney's theme parks and merchandising.
  • Prime Video was a loss-leader within Amazon's ecosystem, with no standalone profitability targets.
  • Apple TV+ was a rounding error, loss-making but bundled with Apple's ecosystem.

Analysts who averaged these companies' multiples found a "streaming P/E" that was meaningless. Netflix shouldn't have been valued like Disney+ at all. Yet the peer set averaging created an anchor that made Netflix look expensive during the 2021 peak even as it was vastly more profitable than peers.

Semiconductor Manufacturing vs Fabless Design (2022–2024)

NVIDIA (fabless) and TSMC or Samsung (foundries) are both in semiconductors, but they're different beasts. NVIDIA outsources manufacturing, allowing it to scale without massive capex. TSMC must invest 20%+ of revenue in new fabs just to keep up with Moore's Law. Their capital structures, growth profiles, and multiples bear no resemblance.

Yet industry peer sets often blend them. An analyst using an industry average P/E of 25× would massively overvalue TSMC (which might warrant 12×) and undervalue NVIDIA (which might warrant 35×). The peer set "average" is a fiction that serves neither.

Tesla and the Automotive Peer Set (2018–Present)

Tesla trades at 20+ multiple of sales despite lower margins than traditional OEMs. Ford and GM trade at 0.3–0.6× sales with much higher near-term cash generation. Is Tesla expensive?

An analyst who averages Tesla with Ford and GM finds an automotive peer set at perhaps 0.8× sales and concludes Tesla is wildly overvalued. But Tesla shouldn't be in the same peer set as legacy OEMs. It's competing on scale, margin expansion, and optionality (energy, robotics); Ford and GM are mature, capital-intensive, unionized manufacturers. Using an automotive peer average to value Tesla is a category error.

Common Mistakes Section​

Mistake 1: The Inclusivity Trap​

Including every company with an SIC code overlap to avoid accusations of cherry-picking, resulting in a peer set so large and diverse that the average is meaningless.

Mistake 2: Ignoring Profitability Divergence​

Using an industry P/E multiple without adjusting for EBITDA margins, ROE, or free cash flow yield, even when peers have dramatically different profitability.

Mistake 3: Weighting Peers Equally Regardless of Size or Relevance​

Treating a $500 million niche competitor the same as a $50 billion category leader in a 15-company peer set, giving each a 6.7% weight.

Mistake 4: Anchoring to Peer Average Without Explaining Divergence​

Concluding "our stock is cheap relative to peers" without explaining whether the discount is due to temporary factors (valuation reset, temporary margin pressure) or structural weakness.

Mistake 5: Using Outdated Peer Multiples​

Pulling peer multiples from an old analyst report, updating them mechanically for current prices, without reassessing whether the peer set itself has evolved (acquisitions, bankruptcies, business model shifts).

Mistake 6: Confusing Correlation with Comparability​

Saying two companies are "peers" because their stock prices move together or they're in the same index, rather than because they're strategically similar.

FAQ​

Q: How many companies should be in a peer set?

A: The smallest number that adequately represents the competitive set without allowing outliers to dominate. For a narrowly defined business (a single-product software company), 3–5 true peers might be sufficient. For a diversified industrial, 8–12 might be needed. Above 15, you're usually compromising on homogeneity to gain statistical credibility.

Q: Should I use market-cap weighted or equal-weighted peer multiples?

A: Depends on your analysis goal. If you want to know what the market is paying on average, market-cap weight reflects the consensus of the largest investors. If you want to identify structural multiples for peers at similar competitive stages, equal-weight avoids bias toward the largest players. Neither is objectively "right"; know which you're using and why.

Q: What if my target company has no true peers—it's too unique?

A: That's a signal that relative valuation may not be appropriate. Build a peer set of companies with similar growth rates, capital intensity, and profitability, even if they operate in different industries. Or acknowledge that you're using a loose peer set and lean on DCF valuation instead. Forcing a peer set where none exists is worse than admitting you don't have one.

Q: Should I include international peers in a domestic peer set?

A: Yes, if they're true competitors. A German chemicals company competes with American ones for global customers. But account for currency, tax, and macro differences. Don't assume a German-listed peer trades at the same multiple as a US peer just because they're in the same business.

Q: How do I adjust for growth differences between peers?

A: The most common method is the PEG ratio (P/E divided by expected growth rate). If Peer A has a P/E of 20× and 15% growth, its PEG is 1.33. Peer B has a P/E of 18× and 10% growth, so PEG is 1.8. On a PEG basis, Peer B is more expensive. But be careful: PEG assumes growth differences are permanent. If Peer B is in a low-growth phase but will accelerate, the current PEG is misleading.

Q: Can I use a peer set if some companies are about to undergo major transitions (spin-offs, divestitures)?

A: Proceed carefully. A company that's about to spin off a division has a different capital structure and profitability going forward. Using its current multiples as a comparable is anchoring to an outdated picture. Adjust either the peer's multiples forward or note the transition in your analysis.

  • Selection Bias and Peer Set Composition — The peers you include define the valuation range. Starting with a different universe (all listed companies vs top 100 by market cap) produces different peer sets and different "fair values."
  • PEG Ratio and Growth Adjustment — The Price-to-Earnings-to-Growth ratio is a mechanical way to adjust for growth differences, but it assumes growth differences are permanent and equally important across sectors.
  • Earnings Quality and Multiple Divergence — If peers have different earnings quality (one is high-margin and growing sustainably; another is margin-diluted or subject to one-timers), their multiples should differ, and that difference should be explained, not averaged away.
  • Market Sentiment and Consensus Multiples — A bad peer set is often one that simply echoes the market consensus. Using it doesn't provide a check on whether consensus is rational.
  • Precedent Transactions vs Trading Comparables — M&A transactions (precedent transactions) provide a different peer universe than trading comps, often at higher multiples because of control premium.

Summary​

A good peer set is small, tightly curated, and intellectually defensible. It answers the question: "If I had to explain to a skeptic why these five companies are comparable to my target, what would I say?" If you find yourself listing 15 companies just to seem objective, you've built a bad peer set.

The mechanics of peer analysis are simple—pull multiples, average them, compare to target. The hard part is defining the peer set itself. That requires judgment: understanding competitive dynamics, recognizing when business models differ, and having the confidence to exclude companies that don't belong.

Most sell-side analysis builds large peer sets to avoid criticism for cherry-picking. That creates an index-like benchmark that echoes whatever the market has priced in. You've traded the hard work of valuation for the false comfort of consensus. That's not analysis; it's capitulation.

Start with a small, core peer set of true comparables. Use it to derive a fair multiple range. Then expand to a broader group to understand where your core sits relative to the full market. The expanded group informs context, not valuation. Your core peers drive the valuation; the rest provide perspective on whether the market is rational.

Next​

Read Ignoring management incentives to understand how poorly structured executive compensation can corrupt both the business and the analyst's confidence in management.

Statistic: Research on analyst peer sets shows that expanding a peer set from 5 companies to 15 increases the standard deviation of average multiples by 30–50%, while adding almost no new information about what the truly comparable companies are worth.