Skip to main content

Survivor Bias in Comparables

Comparable company analysis is one of the most commonly used valuation approaches. You identify peers in your company's industry, observe the multiples at which they trade, and apply an average multiple to your company's metrics. This yields an "intrinsic value" that feels data-driven and disciplined.

The fatal flaw: the companies you observe are survivors. You compare your company to General Electric, not to dozens of industrial conglomerates that failed. You compare to Microsoft and Apple, not to the dozens of technology companies that went bankrupt. You compare to successful pharmaceutical companies, not to the ones that spent billions on failed drug pipelines and disappeared.

This is survivor bias in valuation. By comparing only to companies that succeeded, you systematically overestimate the average multiple at which companies should trade. You ignore that many companies trading at similar multiples today will fail, implying that the "average" multiple you observe is actually a premium over the true cross-sectional average of all companies (living and dead) that shared those characteristics.

Quick Definition​

Survivor bias in valuation refers to using only currently successful companies as comparables, thereby overestimating fair value multiples. Failed or defunct companies are excluded from the analysis, despite being equally relevant to understanding the range of outcomes. This creates systematic overvaluation through selection bias.

Key Takeaways​

  • Comparable multiples observed in current markets are biased upward because they exclude failed companies that traded at similar multiples before bankruptcy.
  • When you compare a risky biotech company to successful pharma firms, you're ignoring that dozens of biotech companies with similar profiles failed completely—returning 99%+ losses.
  • Survivor bias implies that if you buy a basket of companies at their "average" observed multiple, you'll underperform expectations because many will fail, dragging down returns.
  • The appropriate adjustment: apply a discount to observed multiples when the company faces meaningful failure risk, to account for the hidden loss scenarios not reflected in the peer set.
  • Historical bankruptcy rates by industry and company profile can guide the adjustment; a typical adjustment is 20–50% discount to observed multiples for risky companies.
  • The strongest protection is to build probability-weighted scenarios explicitly: "25% chance of bankruptcy, 50% chance of success at 12x earnings, 25% chance of value expansion to 18x earnings."

How Survivor Bias Distorts Multiples​

The Missing Dead Companies​

Consider the biotechnology industry in 2020. You build a comparable set of publicly traded biotech companies and observe median EV/Revenue of 4.0x (excluding ones currently trading).

But for every successful biotech company worth $2 billion, dozens of others:

  • Spent $500 million on R&D and went bankrupt
  • Merged at a loss after shareholders were diluted 90%
  • Failed clinical trials and ceased operations
  • Never made it public because the pipeline collapsed

If you included all the companies that started with biotech characteristics but failed, the true average EV/Revenue multiple across all such firms would be far lower—perhaps 0.5x or 1.0x—because most returned zero.

Yet you observe only the 4.0x multiple among survivors and assume that's fair value. You're unconsciously selecting only the positive outcomes and ignoring the negative ones.

The Narrative Smoothing​

Survivor bias is reinforced by narratives. Today's successful companies have "moats," "first-mover advantages," "superior management," or "exceptional culture." But when those companies were young, they looked identical to companies that failed. The difference in outcome is often luck, regulatory approval (for pharma), network effects (for software), or merely surviving the initial capital-constrained period.

Looking backward, we ascribe skill to survivors and bad luck to failures. This creates selection bias in the characteristics we believe predict success. A young software company with a charismatic founder and a novel algorithm has equal odds of becoming Microsoft or disappearing within 5 years. But because Microsoft is famous and its founders have written books, we unconsciously assume the profile is more positive than it actually is.

Why Survivor Bias Matters for Valuation​

The Bankruptcy Risk Is Invisible​

When you use comparables, you implicitly assume your company has a similar probability of survival to the peer set. But if your company is riskier (earlier stage, more leveraged, in a disrupted industry), the bankruptcy risk is real and should be reflected in the valuation.

Example: Two Biotech Comparables

Company A: Approved drug, generating cash flow. EV/Revenue of 3.5x Company B: In-phase-three trial, 18 months from approval/failure. EV/Revenue of 2.0x (already discounted for binary risk)

If you analyze Company B as a comparable to another phase-three biotech and apply the 3.5x multiple from Company A, you're ignoring that Company B has 40–60% failure risk. A 50% failure risk implies 50% of outcomes generate zero value. The true expected value is not the multiple times revenue but the multiple times revenue times the probability of success:

Expected Value = 2.0x Revenue × (40% success probability) versus observed peer multiple of 3.5x suggests 100% success

Your valuation is too high by 3.5x to 0.8x, nearly a 5x error.

The Time Decay of Multiples​

Survivor bias also explains why multiples change over time. In 2020, high-growth tech companies traded at 10–15x revenue. In 2024, comparable companies trade at 3–5x revenue. Was the industry suddenly recognized as less valuable? Or did investor risk perception change?

Partially the latter. In 2020, investors implicitly discounted bankruptcy risk for growth companies because the market was booming and survival seemed assured. By 2024, multiple bankruptcies and failures had occurred (crypto exchanges, unprofitable platforms, high-burn startups), and investors repriced the failure risk. The multiple contraction reflects the hidden failure risk finally being priced.

If you had used 2020's multiples as "fair value" in 2024, you would have overstated intrinsic value by 50–70%.

Industries and Profiles Most Affected by Survivor Bias​

Biotechnology and Drug Development​

Biotech is the poster child for survivor bias. Approximately 90% of drug candidates fail or never generate positive returns. Yet comparable analysis of successful pharma companies yields multiples that assume much higher success rates.

When valuing a biotech company, you must explicitly model the probability of clinical success, regulatory approval, and market adoption. Using successful pharma multiples without adjustment creates massive overvaluation.

Early-Stage Software and SaaS​

Early SaaS companies fail at high rates. Churn rates increase, unit economics deteriorate, and funding dries up. Yet comparables analysis of the few successful SaaS companies (Salesforce, Atlassian, etc.) yields multiples that imply lower failure rates than actual markets experience.

A SaaS company with $5 million annual recurring revenue trading at 8x revenue might seem reasonable compared to peers at 10x. But 50% of $5 million SaaS companies go bankrupt within 5 years. The true probability-weighted valuation is far lower.

Startups and IPO-Stage Companies​

The companies that go public are the survivors of venture funding. Thousands of startups get funded, but only a tiny fraction go public. Using the public company multiples as "fair value" for recent IPOs ignores that many recent IPOs will fail.

A 2021 SPACs and recent IPO were valued using growth multiples of successful tech companies. By 2023, many had failed or been taken private at 80–90% losses. The survivor bias was extreme: only the best 5–10% of funded startups ever go public, and multiples were set based on that exceptional cohort.

Cyclical and Commodity Industries​

In boom periods, commodity and cyclical companies trade at elevated multiples because current profits are high. Comparable analysis suggests these multiples are "fair." But many companies in cyclical industries fail in downturns. Using boom-time comparables to value cyclical companies ignores the failure risk in downturns.

Steel mills, auto suppliers, construction companies, and regional banks all face bankruptcy risk concentrated in downturns. Valuations should incorporate the probability that a current peer may fail in a future downturn.

Survivor Bias Assessment Framework​

Real-World Examples​

Internet Bubble (1995-2000)​

Investors valued Internet companies using comparable multiples from the few successful ones (Yahoo, eBay, Amazon). Countless other Internet companies traded at 50x revenue, 100x earnings, or with no earnings at all.

Comparables analysis suggested these were "fairly valued" based on peers. In reality, 95% of Internet companies failed, were acquired at losses, or went bankrupt. Survivor bias created valuation absurdity.

By 2000-2002, valuations compressed 75–90% as the market belatedly recognized that most Internet companies should never have been public. The failure risk that was invisible during the bubble became obvious in hindsight.

Telecom Bubble (2000-2002)​

Telecom companies were valued on multiples of successful incumbents (AT&T, regional Bell companies). But massive overbuilding had occurred; hundreds of telecom providers had launched with similar business models.

Comparable analysis suggested reasonable valuations for many carriers. Instead, thousands went bankrupt. Survivor bias had made a category seem safer than it was.

Oil and Gas (2014-2016)​

In 2014, oil and gas companies traded at multiples reflecting $100+ oil prices. When oil collapsed to $30–$40, companies trading at those multiples faced bankruptcy. Comparable analysis of surviving oil majors had masked the failure risk in smaller, higher-cost producers.

Companies like many E&P (exploration and production) firms that seemed "cheap" at 8x revenue in 2014 went bankrupt by 2016. Survivor bias had made the category look safer than it was.

Cryptocurrency Exchanges (2017-2022)​

Cryptocurrency exchanges and platforms traded at multiples based on the survivors (Coinbase, etc.). But dozens of other exchanges existed, many of which failed spectacularly (FTX, 3Arrows Capital, and countless others).

Investors using Coinbase multiples to value FTX missed that most crypto platforms face bankruptcy risk and business model challenges. Survivor bias suggested a reasonable multiple when actual failure rates were high.

How to Adjust for Survivor Bias​

Step 1: Estimate Historical Failure Rates​

For the category of company you're valuing, research:

  • How many companies started in this space in the past 10 years?
  • What percentage went bankrupt, merged at losses, or are no longer public?
  • What's the typical duration before outcomes become clear?

For biotech: 80–90% of drug candidates fail. For SaaS: 30–50% of startups fail within 5 years. For early-stage tech: 50–70% fail or provide poor returns to later investors. For regional banks: 10–20% fail in major credit downturns.

Step 2: Model Outcome Scenarios Explicitly​

Rather than using a single comparable multiple, model outcomes:

Scenario 1 (60% probability): Success

  • Company achieves targets; multiples match peers at 12x revenue
  • Valuation: $10 million revenue × 12x = $120 million

Scenario 2 (35% probability): Moderate underperformance

  • Company achieves 70% of targets; trades at 8x revenue
  • Valuation: $10 million revenue × 8x = $80 million

Scenario 3 (5% probability): Bankruptcy

  • Company goes under; equity holders get 0
  • Valuation: $0

Probability-Weighted Value = (0.60 × $120M) + (0.35 × $80M) + (0.05 × $0) = $72 million + $28 million + $0 = $100 million

Versus the naive comparable approach that would use 10x average multiple ($100 million) or 12x peer multiple ($120 million) without accounting for failure risk.

Step 3: Adjust Comparable Multiples Downward​

A simpler approach: take the observed comparable multiples and discount them based on failure probability.

Observed comparable multiple: 10x revenue Estimated bankruptcy risk: 25% Implied adjustment: 25% × 50% loss in bankruptcy = 12.5% discount Adjusted multiple: 10x × 0.875 = 8.75x

(The assumption is that bankruptcy results in a 50% loss to remaining equity; adjust based on expected liquidation value.)

Step 4: Distinguish Between Currently Dead and Future Dead Companies​

Some companies in your comparable set are currently thriving. Others are struggling. Apply higher failure probability to strugglers and lower to thriving peers. Weight the comparable multiples accordingly rather than assuming all peers have identical risk profiles.

Common Mistakes​

1. Using Only Current Public Companies as Comparables

Public companies are survivors. To get a realistic sense of multiples, you need to include failed companies (those delisted, bankrupted, or acquired at major losses) in your analysis. This is harder but more accurate.

2. Assuming Historical Survivors Will Survive in the Future

A company that survived 2008 may not survive the next cycle. Using pre-2008 financial metrics of companies that barely made it through as benchmarks for current valuations is a mistake.

3. Ignoring Leverage When Assessing Bankruptcy Risk

A highly leveraged company in a cyclical industry has far higher bankruptcy risk than a low-leverage company. Adjust comparable multiples higher for low leverage and lower for high leverage.

4. Using Peer Multiples from Different Economic Environments

Multiples in a low-rate, growth environment (2010-2020) are not comparable to multiples in a high-rate environment (2023-2025). Adjust for the interest rate regime and growth expectations of the original comparable period.

5. Cherry-Picking Peers to Get a Desired Valuation

If you include only the highest-multiple peers and exclude struggling ones, you create upward bias. Always include a representative mix of peers, including the struggling ones.

FAQ​

Q: How much should I discount for survivor bias?

The adjustment depends on historical failure rates and your company's characteristics. Mature, profitable companies in stable industries: 0–10% discount. Growth-stage companies in newer industries: 20–50% discount. Pre-revenue startups: 50–80% discount or use scenario analysis instead.

Q: Is survivor bias relevant if I'm analyzing a large-cap company?

Less so, but not zero. Large-cap companies face lower bankruptcy risk, but they're not immune. GE, Lehman Brothers, and Sears were once large-cap. Always account for some failure probability, even if small.

Q: Should I discount for survivor bias if using medians instead of means?

Yes. Medians are less skewed by outliers but still reflect only surviving companies. The failed companies would shift the median downward. Apply a smaller adjustment (10–30% instead of 20–50%) but still adjust.

Q: How do I value a company in a category where failure is extremely likely?

Don't use comparables. Build an explicit scenario analysis with probabilities of different outcomes: bankruptcy, acquisition at a loss, modest success, or major success. Weight those scenarios and calculate expected value.

Q: If I build scenarios for failure, should I use comparable multiples for the success cases?

Use comparable multiples for success cases, but adjust downward from observed multiples to account for the fact that the success cases in your scenario analysis are higher-probability than the average comparable company. Be consistent in your probability assumptions.

Q: Doesn't my margin of safety protect me against survivor bias?

Partially. If you demand a 40% margin of safety and survivor bias creates a 30% overvaluation, you're partially protected. But if survivor bias is 60% and your margin is 40%, you're still taking on hidden risk. Always try to quantify survivor bias separately.

  • Comparable Company Analysis — Learn how to build and use peer comparables correctly while adjusting for bias.
  • Scenario Analysis in Valuation — Explore how to model multiple outcomes and probabilities explicitly.
  • Selection Bias in Data — Understand the broader category of selection bias in investing.
  • Risk Factors and Bankruptcy Probability — Learn how to estimate and model bankruptcy risk in valuations.

Summary​

Survivor bias in valuation is invisible but consequential. By comparing only to companies that succeeded, you systematically overstimate fair-value multiples and create two problems: (1) you overpay for risky companies that look reasonably valued on comparables, and (2) you underestimate the risk of losses when a significant fraction of companies in the category ultimately fail.

The solution is not to abandon comparable analysis but to augment it with explicit failure-probability modeling. Estimate the historical failure rate in the company's category. Model scenarios with different success outcomes and their probabilities. Discount observed multiples accordingly. Build your valuation to reflect the reality that many companies in a peer set will fail, pulling down the true average multiple across all companies in that category (dead and alive).

Professional investors who have navigated multiple market cycles understand this intuitively. They discount multiples for risk factors invisible in the current comparable set. Retail investors often don't, leading them to pay "reasonable" comparable-based multiples for companies in categories facing 50%+ failure rates. By addressing survivor bias explicitly, you protect yourself against this systematic source of overvaluation.

Next​

Continue to Using Valuation for Market Timing to explore how attempting to time markets based on valuation metrics systematically underperforms and why most investors should avoid this trap.