Reading the Distribution: What P10, P50, P90 Actually Mean (And How to Use Them)

October 30, 2025 • 9 min read

You just ran Monte Carlo for your latest fund commitment.

The results are in:

Expected TVPI: 2.56x
P10: 1.82x
P25: 2.14x
P50: 2.51x
P75: 2.89x
P90: 3.76x
Std Dev: 0.62x

Now what?

Which number do you use? What do they mean? How do you make a decision?

---

Percentile Definitions (Without the Jargon)

[PercentileExplainer defaultSelected="p50"]

Click on each percentile above to see what it means and when to use it.

P10 (Pessimistic)

What it is: "Only 10% of our 1,000 simulations were worse than this."

Translation: "There's a 90% chance we'll do better than P10."

NOT the same as:

• ❌ "Worst case" (1% of outcomes are even worse)
• ❌ "Minimum outcome" (0.1% could be very bad)
• ❌ "Guaranteed floor" (nothing is guaranteed)

When to use:

• Downside risk assessment
• Conservative liquidity planning
• Stress testing

Example decision:

Fund commitment: $50M
P10 TVPI: 1.8x

Question: "Can we handle 1.8x in bad scenarios?"
- If yes → Comfortable with downside → Approve
- If no → Too risky → Pass or reduce commitment

---

P50 (Median)

What it is: "50% of simulations were above this, 50% below."

Translation: "Coin-flip probability of beating P50."

NOT the same as:

• ❌ "The average" (that's "Expected" or "Mean")
• ❌ "What will happen" (it's a probability, not a prediction)
• ❌ "Most likely outcome" (that's "Mode")

When to use:

• Expected value planning
• Base case scenarios
• Balanced risk/reward decisions

Example decision:

Portfolio projection (5 years)
P50 NAV: $685M

Use for: "We expect around $685M in typical scenarios"
Not for: "We will have $685M" (too certain)

---

P90 (Optimistic)

What it is: "Only 10% of simulations were better than this."

Translation: "There's a 10% chance we'll beat P90."

NOT the same as:

• ❌ "Best case" (1% of outcomes are even better)
• ❌ "Maximum outcome" (0.1% could be much higher)
• ❌ "Guaranteed upside" (low probability)

When to use:

• Upside potential analysis
• Stretch targets
• "What if things go really well?" scenarios

Example decision:

Fund marketing materials
P90 TVPI: 3.8x

Statement: "In strong markets, possible to achieve 3.8x"
Warning: "Only 10% probability - not typical outcome"

---

The Mean vs. Median Confusion

You have two numbers:

Mean (Expected): 2.56x
Median (P50): 2.51x

Why different?

For Symmetric Distributions (Normal)

        ___
      /     \
    /         \
  /             \

Mean = Median (they're the same)

For Right-Skewed (Lognormal - typical for TVPI)

       ___
     /     \____
    /            \__
   /                 \_

Mean > Median (fat right tail pulls mean up)

For Left-Skewed (rare, but happens)

        _______
      /         \
   __/           \
 _/               \

Mean < Median (fat left tail pulls mean down)

For TVPI distributions:

• Usually right-skewed (lognormal)
• Mean slightly > Median
• Difference tells you about tail risk

Example interpretation:

Mean: 2.56x
P50: 2.51x
Difference: 0.05x (2% of mean)

→ Mild right skew
→ Some upside potential from outliers
→ Not extreme (VC would be 10-20% difference)

---

Decision Framework by Use Case

Use Case 1: Commitment Decisions

Question: "Should we commit $50M to this fund?"

Which percentile to use: P10 (downside focus)

Decision logic:

If P10 > 1.5x:
    ✅ Good risk/reward
    Even in bad scenarios (bottom 10%), we make 1.5x+
    → Approve

If P10 = 1.2x - 1.5x:
    ⚠️ Marginal
    Downside is weak (barely above breakeven)
    → Approve with reduced commitment OR pass

If P10 < 1.2x:
    ❌ Significant loss risk
    Bottom 10% scenarios lose money or barely break even
    → Pass

Real example:

Fund A: P10=1.9x, P50=2.5x, P90=3.8x
→ Strong downside (1.9x even in bad scenarios)
→ Approved $50M

Fund B: P10=0.95x, P50=2.5x, P90=5.2x
→ Weak downside (lose money in bottom 10%)
→ Passed (too volatile)

Notice: Both have same P50, but Fund A is much safer.

---

Use Case 2: Liquidity Planning

Question: "How much should we reserve for capital calls?"

Which percentile to use: P10 or P5 (stress case)

Why: Liquidity shortfall is catastrophic. Better to over-reserve.

Decision logic:

Calculate net cashflow under different scenarios:
- P50 (base): Need $150M reserves
- P25 (conservative): Need $185M reserves
- P10 (stress): Need $240M reserves

Your risk tolerance:
- Risk-neutral: Use P50 ($150M)
- Risk-averse: Use P25 ($185M)
- Very risk-averse: Use P10 ($240M)

Real example:

Family office with $500M portfolio:

Cashflow simulation (next 8 quarters):
- P50 net need: $120M
- P25 net need: $165M
- P10 net need: $215M

Decision: Maintain $175M reserves (between P25-P10)
Reasoning: Family offices can't afford liquidity crisis
Cost: 3% opportunity cost on $175M = $5.25M/year
Worth it: Yes (sleep insurance)

---

Use Case 3: Performance Expectations

Question: "What should I tell the board?"

Which percentile to use: P50 + Range (P25-P75 or P10-P90)

Why: Single number encourages false confidence. Range conveys uncertainty.

How to communicate:

❌ Don't say: "We will deliver $685M NAV in 5 years." (Sounds like guarantee)

✅ Do say: "We expect median outcome around $685M, with realistic range of$620M-$750M." (Conveys uncertainty)

Even better: "We're 80% confident NAV will be between $620M-$750M, with median expectation of $685M." (Includes confidence level)

Real presentation slide:

5-Year Portfolio Projection

Expected (P50):     $685M
Realistic range:    $620M - $750M (80% confidence)
Conservative case:  $590M (P10 - stress scenario)
Optimistic case:    $780M (P90 - strong markets)

Recommendation: Plan for $685M, reserve for $590M downside

---

Use Case 4: Comparing Funds

Question: "Fund A or Fund B?"

Look at the full distribution, not just P50.

Example:

Fund A:
P10: 1.8x | P50: 2.5x | P90: 3.4x | Range: 1.6x

Fund B:
P10: 1.1x | P50: 2.5x | P90: 4.8x | Range: 3.7x

Same P50, but very different risk profiles!

Fund A: Lower risk, tighter range, better downside Fund B: Higher risk, wider range, worse downside but better upside

Which to choose?

Risk-averse investor:
→ Fund A (safer, better P10)

Risk-tolerant investor:
→ Fund B (lottery ticket, high P90)

Depends on your portfolio context:
- Need safe returns? → Fund A
- Can afford volatility? → Fund B

---

The 80% Confidence Interval Trick

Most useful metric: P10 to P90 range

Why: Captures 80% of possible outcomes (ignores extremes).

How to use:

TVPI distribution:
P10: 1.82x
P90: 3.76x
Range: 1.94x (P90 - P10)

Interpretation:
- "We're 80% confident final TVPI will be 1.82x - 3.76x"
- "Range of 1.94x indicates moderate uncertainty"
- "Not low risk (range < 1.0x) or high risk (range > 3.0x)"

Rule of thumb:

---

Common Mistakes (And How to Avoid Them)

Mistake 1: "P90 is Guaranteed Upside"

Wrong thinking: "Best case is P90 = 3.8x, so maximum upside is 3.8x"

Reality: 10% of outcomes are even better than P90.

Could be:

• P95: 4.2x
• P99: 5.1x
• Max: 8.5x (outlier)

Right thinking: "P90 is 90th percentile. 10% of outcomes exceed it."

Mistake 2: "Expected = Most Likely"

Wrong thinking: "Mean is 2.56x, so 2.56x is the most likely outcome"

Reality: For lognormal distributions, no single outcome is "most likely" (continuous distribution).

Right thinking: "Mean represents average across all scenarios, not any specific outcome."

Mistake 3: "We Have 90% Confidence Between P10-P90"

Wrong math: "90% of outcomes between P10 and P90"

Right math: "80% of outcomes between P10 and P90" (10% below P10, 10% above P90)

Correct confidence intervals:

• 50% confidence: P25 - P75
• 80% confidence: P10 - P90
• 90% confidence: P5 - P95

Mistake 4: "Monte Carlo is More Accurate"

Wrong thinking: "Monte Carlo is sophisticated, therefore more accurate than deterministic"

Reality: Garbage in, garbage out. Bad assumptions produce bad results, whether you run 1 scenario or 10,000.

Right thinking: "Monte Carlo quantifies uncertainty around assumptions, doesn't make assumptions more accurate."

---

Decision Matrix: Which Percentile When?

---

Real Portfolio Decision Walkthrough

The Setup

Fund: TechVentures Growth IV Commitment: $75M **Your portfolio:**$500M across 12 funds

Monte Carlo results:

TVPI:
P10: 1.82x | P25: 2.14x | P50: 2.51x | P75: 2.89x | P90: 3.76x

IRR:
P10: 8.5% | P25: 11.2% | P50: 13.8% | P75: 16.5% | P90: 19.8%

Step 1: Check the Downside (P10)

Question: "Can we handle 1.82x in bad scenarios?"

Math:

• Commitment: $75M
• P10 outcome: $75M × 1.82 =$136.5M
• Net profit: $61.5M (even in bottom 10% scenarios)

Answer: ✅ Yes, downside is acceptable.

Step 2: Check the Range (P90 - P10)

Question: "How uncertain is this?"

Math:

• Range: 3.76x - 1.82x = 1.94x
• As % of P50: 1.94x / 2.51x = 77%

Interpretation:

• Moderate uncertainty (not low, not extreme)
• Typical for growth equity
• More certain than VC (would be 3x+ range)

Answer: ✅ Risk level appropriate for asset class.

Step 3: Compare to Alternatives

You're also considering:

Fund B:

P10: 1.45x | P50: 2.51x | P90: 4.22x
Range: 2.77x (110% of P50)

Comparison:

Interpretation:

• Fund A: Safer, better downside, less volatile
• Fund B: Riskier, worse downside, higher upside potential

Decision depends on:

• Your risk tolerance
• Portfolio composition (already have high-risk funds?)
• Liquidity constraints

For this investor: → Chose Fund A (already had enough VC exposure, wanted safer growth equity)

---

The Probability Translation

Turn percentiles into probabilities:

P10 = 1.82x
→ "90% probability of beating 1.82x"
→ "10% probability of worse than 1.82x"

P50 = 2.51x
→ "50% probability of beating 2.51x"
→ "50% probability of worse than 2.51x"

P90 = 3.76x
→ "10% probability of beating 3.76x"
→ "90% probability of worse than 3.76x"

For any percentile Pₙ:

Probability of exceeding Pₙ = (100 - n)%
Probability of falling short of Pₙ = n%

Example questions answered:

Q: "What's probability of TVPI > 3.0x?"
A: P75 = 2.89x, P90 = 3.76x
   → Between 10-25% probability

Q: "What's probability of TVPI < 2.0x?"
A: P10 = 1.82x, P25 = 2.14x
   → Between 10-25% probability

Q: "What's probability we at least break even (1.0x)?"
A: P10 = 1.82x (well above 1.0x)
   → >90% probability (likely >99%)

---

Distribution Shapes Matter

Shape 1: Symmetric (Normal)

Characteristics:

• Mean = Median
• Upside = Downside
• (P50 - P10) ≈ (P90 - P50)

Example:

P10: 2.0x | P50: 2.5x | P90: 3.0x
Difference down: 0.5x
Difference up: 0.5x

When you see this: Balanced risk, no skew

---

Shape 2: Right-Skewed (Lognormal)

Characteristics:

• Mean > Median
• More upside than downside
• (P90 - P50) > (P50 - P10)

Example:

P10: 1.8x | P50: 2.5x | P90: 4.0x
Difference down: 0.7x
Difference up: 1.5x (larger!)

When you see this:

• Typical for VC/growth equity
• Uncapped upside, limited downside
• "Home run" potential

Interpretation: "Limited downside (worst case ~1.8x), but some scenarios deliver 4x+. Good risk/reward if you can handle 1.8x floor."

---

Shape 3: Left-Skewed (Rare)

Characteristics:

• Mean < Median
• More downside than upside
• (P50 - P10) > (P90 - P50)

Example:

P10: 0.8x | P50: 2.0x | P90: 2.5x
Difference down: 1.2x (large!)
Difference up: 0.5x

When you see this:

• Rare (usually means something's wrong)
• Capped upside, uncapped downside
• Red flag for PE/VC

Interpretation: "Limited upside but significant downside risk. Avoid."

---

Your Personal Risk Tolerance

Map percentiles to your risk profile:

Conservative Investor

Focus: P10 and P25

Questions:

• "Can I handle P10 scenario?"
• "Is P25 acceptable minimum?"
• "What's my true downside?"

Example decision: "P10 is 1.8x. I can handle that. Approved."

---

Balanced Investor

Focus: P50 and P25-P75 range

Questions:

• "What's the median outcome?"
• "What's the realistic range?"
• "Am I comfortable with that range?"

Example decision: "Median is 2.5x, range is 2.1x-2.9x. Comfortable. Approved."

---

Aggressive Investor

Focus: P75, P90, and upside potential

Questions:

• "What's the upside if things go well?"
• "Can I afford the downside for that upside?"
• "Is this a potential home run?"

Example decision: "P90 is 3.8x, and I can tolerate P10 of 1.8x. Worth the shot. Approved."

---

The 5-Second Decision Test

After running Monte Carlo, ask yourself:

Question 1: "What's P10?"

If P10 is unacceptable → Pass, don't dig deeper

Question 2: "What's P50?"

If P50 is attractive AND P10 is acceptable → Deep dive

Question 3: "What's the range?"

If range is too wide for comfort → Reduce commitment or Pass

Question 4: "How does this compare to alternatives?"

If this has better P10 than alternatives with similar P50 → Prefer this

Question 5: "Can we handle the downside?"

This is the only question that matters.

---

Practical Workflow

Step 1: Run Monte Carlo

(2 minutes)

Step 2: Look at P10 First

If P10 is acceptable: → Continue analysis

If P10 is unacceptable: → Stop, pass on opportunity, move on

Step 3: Check P50 and Range

• P50: Expected outcome
• Range: Uncertainty measure

Step 4: Compare to Your Criteria

Your investment criteria:
- Minimum P10: 1.5x
- Target P50: 2.2x+
- Maximum acceptable range: 2.5x

This fund:
- P10: 1.82x ✅ (above 1.5x)
- P50: 2.51x ✅ (above 2.2x)
- Range: 1.94x ✅ (below 2.5x)

Decision: Passes all criteria → Approve

Step 5: Make Decision

Based on:

• P10 downside
• P50 expected
• Range uncertainty
• Your risk tolerance
• Portfolio context

Not based on:

• Gut feel
• Single scenario
• GP's promises
• Historical performance alone

---

The Confidence Trick

Instead of: "The model projects $685M" **Say:** "We're 80$620M-$750M, with$685M as median"

Instead of: "Expected TVPI is 2.5x" Say: "Median TVPI is 2.5x, with 80% confidence interval of 1.8x-3.8x"

Why this matters:

• Conveys uncertainty
• Sets appropriate expectations
• Reduces anchoring on single number
• More honest communication

---

Summary: Your Percentile Cheat Sheet

Quick reference:

Decision framework:

1. Check P10 (can you handle downside?)
2. Check P50 (is expected attractive?)
3. Check range (is uncertainty acceptable?)
4. Compare to alternatives
5. Decide based on risk tolerance

---

Try It Yourself

Want to see how percentiles work with your data?

[PercentileExplainer]

Adjust the parameters and see:

• How P10/P50/P90 change
• What each percentile means
• When to use which one

See it with your actual portfolio: Request a demo →

---

Related Reading

• Three Questions Monte Carlo Answers →
• How We Validate Our Monte Carlo Engine →
• Monte Carlo vs Deterministic Modeling →

---

The Bottom Line

Percentiles are tools, not answers.

P10 tells you downside risk. P50 tells you expected outcome. P90 tells you upside potential.

Your job: Combine these insights with your judgment to make better decisions.

The shift:

From: "What will happen?" (unknowable) To: "What's the range of possibilities, and am I comfortable with that?" (actionable)

---

Want help interpreting your Monte Carlo results?

Request a demo → We'll walk through your actual portfolio data.

---

Published: October 30, 2025 Category: Portfolio Management Tags: Monte Carlo, Risk Management, Decision Making, Portfolio Planning

Part 2 of "Practical Monte Carlo" series for fund managers.