This is Markowitz (1952) applied to operating investments.
Your CFO approved a $1M automation Budget this year. You ranked all six candidate projects by ROI and funded the top two. Three months in, both are stalled - they share the same overloaded data engineering team and the same legacy order system integration. Meanwhile, the project you ranked fourth (different team, different system) was ready to ship last month, but you didn't fund it. You picked the best individual investments and ended up with the worst Portfolio.
Your Capital Investments form a Portfolio. Evaluating them individually by ROI ignores how they fail together. Apply Markowitz Portfolio Theory to your operating Budget - map shared dependencies between projects, estimate how tightly their outcomes are linked, and choose the mix that maximizes Risk-Adjusted Return, not just Expected Return.
Operating Investments is Markowitz Portfolio Theory applied to your Capital Allocation decisions inside a business.
The core idea: a Portfolio's total Variance depends not just on each investment's individual risk, but on how tightly the investments are linked through shared dependencies. The Markowitz formula captures this linkage with a term called ρ (rho) - a number between 0 and 1 that measures how much two investments depend on the same factors. When two projects share a team, a system, or a vendor, ρ is high. When they have nothing in common, ρ is near zero.
Each Capital Investment has an Expected Return (the savings or Revenue it generates) and a Standard Deviation (how uncertain that return is). When you fund multiple projects in the same cycle, they form a Portfolio. That Portfolio's risk depends on ρ between each pair of projects - how likely they are to succeed or fail for the same reasons.
The Efficient Frontier exists for your operating Budget the same way it exists for financial Portfolios. It's the set of project mixes that maximize Expected Return for each level of Volatility.
A caveat on the math. The Markowitz formula assumes returns follow a smooth, continuous distribution. Operating projects are lumpier - an automation project either delivers its savings or it doesn't. The downside is bounded at -100% (total loss of investment), the upside is often capped (you can only save what you currently spend), and the outcome is closer to works-or-doesn't than to a smooth spread. The framework still holds directionally because the key insight is about how tightly outcomes are linked through shared dependencies, not the shape of individual Return Distributions. But don't take the third decimal of your Sharpe Ratio seriously.
Your P&L doesn't report project-by-project. It reports the total. If you fund three projects and two share dependencies, a single failure mode - team attrition, vendor delay, Execution Risk materializing - can take down multiple investments at once. That's not bad luck. It's a predictable consequence of building a Portfolio where projects depend on the same factors.
For PE-Backed Operators, this matters more. PE portfolio companies typically have short Investment Horizons and high Hurdle Rates. A quarter where two projects simultaneously underdeliver doesn't just miss Budget - it erodes confidence in the Operations team and changes the Capital Allocation conversation for next year.
The upside: if you construct your Operating Investments to minimize shared dependencies (keep ρ low between projects), your expected EBITDA improvement becomes more predictable even when individual projects carry high Variance. Steady delivery compounds. Lumpy delivery gets your Budget cut.
Step 1: List candidates with Expected Return and Standard Deviation.
For each Capital Investment, estimate the annual return (Cost Reduction or Revenue growth, expressed as a percentage of the investment) and the Standard Deviation of that return.
To estimate Standard Deviation, assess three outcomes: base case, bad case, and good case. If your base case return is 40%, your bad case is 10%, and your good case is 60%, treat the bad and good cases as roughly two Standard Deviations from the base case. Your Standard Deviation is approximately (good case - bad case) / 4. In this example: (60% - 10%) / 4 = 12.5%.
Step 2: Estimate pairwise ρ by mapping shared dependencies.
You won't have precise numbers. You don't need them. Categorize:
Step 3: Calculate Portfolio Variance for each candidate mix.
For a two-project Portfolio, the Markowitz formula is:
Portfolio Variance = w₁²σ₁² + w₂²σ₂² + 2(w₁)(w₂)(ρ)(σ₁)(σ₂)
where w₁ and w₂ are each project's share of the total investment (weights that sum to 1), σ₁ and σ₂ are the Standard Deviations, and ρ is the dependency factor between the two projects.
When projects cost the same, w₁ = w₂ = 0.5. When they don't - say a $700K project and a $300K project in a $1M Budget - the weights are 0.7 and 0.3. The larger investment dominates the Portfolio's risk profile. The worked examples below include both equal-weight and unequal-weight cases.
Step 4: Compute the Sharpe Ratio for each mix.
Use your Hurdle Rate as the baseline. The Portfolio with the highest Sharpe Ratio gives you the best Expected Return per unit of Volatility.
Step 5: Choose based on your Risk Tolerance.
If you're optimizing for predictable EBITDA delivery (most PE-Backed Operations teams), favor the higher Sharpe Ratio mix even if it has a lower absolute Expected Return. If you can absorb Variance - say you have excess capacity and a long Investment Horizon - you can move up the Efficient Frontier toward higher-return, higher-Variance mixes.
Use this framework when:
Skip it when:
Practical signal: If you can draw a line between two funded projects and a shared dependency (person, system, vendor, or customer segment), you have a ρ worth estimating.
Budget: $1M. Each project costs $500K, so you fund exactly two.
ρ estimates based on shared dependencies:
Portfolio A+B (the two highest individual returns):
Expected Return = (40% + 35%) / 2 = 37.5%
Variance = (0.5²)(25²) + (0.5²)(20²) + 2(0.5)(0.5)(0.80)(25)(20) = 156.25 + 100 + 200 = 456.25
Standard Deviation = √456.25 = 21.4%
Portfolio A+C (highest return paired with the low-ρ option):
Expected Return = (40% + 30%) / 2 = 35.0%
Variance = (0.5²)(25²) + (0.5²)(15²) + 2(0.5)(0.5)(0.10)(25)(15) = 156.25 + 56.25 + 18.75 = 231.25
Standard Deviation = √231.25 = 15.2%
Compare Sharpe Ratios using a 10% Hurdle Rate:
Portfolio A+B: (37.5 - 10) / 21.4 = 1.29
Portfolio A+C: (35.0 - 10) / 15.2 = 1.64
Portfolio A+C wins. You give up 2.5 percentage points of Expected Return but cut Volatility by 6.2 percentage points. The Risk-Adjusted Return is meaningfully better.
Insight: Picking the two highest-ROI projects gave you the worst Portfolio. The shared order system dependency between A and B meant their risks compounded instead of offsetting. Project C, ranked third individually, was the better portfolio addition because it brought independent risk. The ranking of projects changes when you evaluate them as a Portfolio instead of a list.
You're planning next year's Capital Investments. Five candidates on the board, each with different teams and systems involved. Before running any Markowitz math, you need to estimate which projects move together.
Count shared dependencies for each pair. D and E share two dependencies (robotics team + warehouse management system). D and G share one (warehouse management system). D and F share zero. D and H share zero.
Assign ρ brackets. Two shared dependencies = high (estimate 0.75). One shared dependency = moderate (estimate 0.35). Zero = low (estimate 0.05).
Build the dependency map:
Identify the risky combination. Funding both D and E in the same year creates a concentrated bet on the robotics team and warehouse management system. If either dependency falters, you lose two investments, not one. Spreading your Budget across D and F, or E and H, gives you lower Portfolio Variance at a modest cost in Expected Return.
Insight: You don't need spreadsheet-precise ρ numbers. The operational question is: 'If one of these fails, does the other fail for the same reason?' If yes, ρ is high, and funding both in the same cycle is a concentration bet you should make deliberately, not accidentally.
Budget: $1M. Two projects with different investment requirements.
Shared dependencies: none (different teams, different systems). ρ ≈ 0.05.
Calculate weights from investment size. w_P = 700K / 1M = 0.7, w_Q = 300K / 1M = 0.3. These replace the 0.5/0.5 split used in equal-cost Portfolios.
Portfolio Expected Return = 0.7(35%) + 0.3(45%) = 24.5% + 13.5% = 38.0%.
Note: this is no longer a simple average. The larger project pulls the Portfolio Expected Return toward its own return.
Portfolio Variance = (0.7²)(20²) + (0.3²)(28²) + 2(0.7)(0.3)(0.05)(20)(28)
= 0.49(400) + 0.09(784) + 11.76
= 196 + 70.56 + 11.76 = 278.32
Standard Deviation = √278.32 = 16.7%
Sharpe Ratio (Hurdle Rate 10%) = (38.0 - 10) / 16.7 = 1.68
Insight: Project P contributes 196 to Portfolio Variance while Project Q contributes only 70.56 - even though Q has higher individual Standard Deviation (28% vs 20%). The weight appears squared in the formula: 0.7² = 0.49 vs 0.3² = 0.09, so the larger investment dominates Portfolio risk regardless of the smaller project's Volatility. When project costs differ, the largest project's risk profile drives the Portfolio. This is why Operators sometimes prefer phasing a large Capital Investment into smaller increments - it changes the weight structure and can reduce Portfolio Variance even without changing ρ.
The best individual investments don't always make the best Portfolio. Shared dependencies create high ρ that amplifies your downside - your worst quarter gets worse, not just your average.
You can estimate ρ between Operating Investments by mapping shared dependencies: same team, same technology, same vendor, same customer segment. Two shared dependencies = high ρ. One = moderate. Zero = low. That's enough to shift the decision.
The Efficient Frontier applies to your operating Budget the same way it applies to a financial Portfolio. Optimizing for Sharpe Ratio - Expected Return per unit of Volatility - usually beats optimizing for raw ROI, especially when your Hurdle Rate demands predictable EBITDA delivery.
Ranking projects by individual ROI and funding from the top. This is the default Capital Budgeting process at most companies, and it systematically ignores shared dependencies between projects. It produces the highest Expected Return Portfolio, but often the worst Risk-Adjusted Return Portfolio. The top-ROI projects frequently share dependencies (they're high-return because they tap the same high-value system or team), which means they have high ρ.
Treating ρ estimation as too imprecise to bother with. You don't need decimal precision. The difference between 'these projects share a team and a system' (high ρ) and 'these projects have nothing in common' (low ρ) is enough to change which Portfolio sits on the Efficient Frontier. Rough is better than ignored.
You have three Capital Investments under consideration. Project X requires your platform engineering team and your payment processor integration. Project Y requires your platform engineering team and your shipping vendor integration. Project Z requires your predictive modeling team and your customer database integration. Rank the three pairwise ρ values (X-Y, X-Z, Y-Z) from highest to lowest and explain your reasoning.
Hint: Count how many dependencies each pair shares. Same team = one shared dependency. Same system = another.
X-Y: Highest. They share the platform engineering team. If that team is overloaded, delayed by production incidents, or loses a key engineer, both projects stall. One shared dependency = moderate to high ρ, estimate ~0.5.
Y-Z: Lowest (tied with X-Z). Zero shared dependencies - different teams, different systems. Estimate ~0.05.
X-Z: Lowest (tied with Y-Z). Zero shared dependencies. Estimate ~0.05.
Ranking: X-Y >> X-Z ≈ Y-Z. The practical implication: funding both X and Y in the same cycle concentrates risk on the platform engineering team.
Two Capital Investments, each costs $400K. You fund both ($800K total Budget). Project M has Expected Return 45% and Standard Deviation 30%. Project N has Expected Return 25% and Standard Deviation 12%. You estimate ρ = 0.70. Calculate the Portfolio's Expected Return, Standard Deviation, and Sharpe Ratio using a 10% Hurdle Rate. Then recalculate assuming ρ drops to 0.15 and compare.
Hint: Use the two-asset Markowitz formula with equal weights (0.5 each). Portfolio Variance = w₁²σ₁² + w₂²σ₂² + 2(w₁)(w₂)(ρ)(σ₁)(σ₂). Then take the square root for Standard Deviation.
High ρ (0.70):
Expected Return = (45 + 25) / 2 = 35%
Variance = (0.25)(900) + (0.25)(144) + 2(0.25)(0.70)(30)(12) = 225 + 36 + 126 = 387
Standard Deviation = √387 = 19.7%
Sharpe = (35 - 10) / 19.7 = 1.27
Low ρ (0.15):
Expected Return = 35% (unchanged - ρ doesn't affect Expected Return)
Variance = 225 + 36 + 2(0.25)(0.15)(30)(12) = 225 + 36 + 27 = 288
Standard Deviation = √288 = 17.0%
Sharpe = (35 - 10) / 17.0 = 1.47
Same projects, same Expected Return, but the Portfolio with low ρ has a 16% better Sharpe Ratio. The only thing that changed was how tightly the projects depend on the same factors.
You're an Operator at a PE-Backed retailer. Budget: $1.2M. Four candidate Capital Investments, each costs $400K (you fund exactly three). Equal project costs simplify the Portfolio math - the unequal-cost case uses different weights but the same formula.
| Project | Expected Return | Std Dev | Dependencies |
|---|---|---|---|
| Alpha | 50% | 30% | Platform team, order system |
| Beta | 40% | 22% | Platform team, warehouse system |
| Gamma | 35% | 18% | Predictive modeling team, customer database |
| Delta | 28% | 10% | Finance ops team, order system |
Estimate pairwise ρ values, calculate the Expected Return and Standard Deviation of all four possible three-project Portfolios (equal weights of 1/3 each), and recommend which three projects to fund. Your Hurdle Rate is 12%.
Hint: First estimate ρ from shared dependencies. People dependencies (shared team) tend to produce higher ρ than system dependencies (shared platform) because you can't easily parallelize human capacity. For a three-asset equal-weight portfolio: Variance = (1/9)(σ₁² + σ₂² + σ₃²) + (2/9)(ρ₁₂σ₁σ₂ + ρ₁₃σ₁σ₃ + ρ₂₃σ₂σ₃). There are four possible three-project combinations: Alpha-Beta-Gamma, Alpha-Beta-Delta, Alpha-Gamma-Delta, Beta-Gamma-Delta.
ρ estimates from dependencies:
Portfolio 1: Alpha + Beta + Gamma
E[R] = (50+40+35)/3 = 41.7%
Var = (1/9)(900+484+324) + (2/9)[(0.55)(30)(22) + (0.05)(30)(18) + (0.05)(22)(18)]
= 189.8 + (2/9)[363 + 27 + 19.8] = 189.8 + 91.1 = 280.9
σ = 16.8%, Sharpe = (41.7-12)/16.8 = 1.77
Portfolio 2: Alpha + Beta + Delta
E[R] = (50+40+28)/3 = 39.3%
Var = (1/9)(900+484+100) + (2/9)[(0.55)(30)(22) + (0.35)(30)(10) + (0.05)(22)(10)]
= 164.9 + (2/9)[363 + 105 + 11] = 164.9 + 106.4 = 271.3
σ = 16.5%, Sharpe = (39.3-12)/16.5 = 1.65
Portfolio 3: Alpha + Gamma + Delta
E[R] = (50+35+28)/3 = 37.7%
Var = (1/9)(900+324+100) + (2/9)[(0.05)(30)(18) + (0.35)(30)(10) + (0.05)(18)(10)]
= 147.1 + (2/9)[27 + 105 + 9] = 147.1 + 31.3 = 178.4
σ = 13.4%, Sharpe = (37.7-12)/13.4 = 1.92
Portfolio 4: Beta + Gamma + Delta
E[R] = (40+35+28)/3 = 34.3%
Var = (1/9)(484+324+100) + (2/9)[(0.05)(22)(18) + (0.05)(22)(10) + (0.05)(18)(10)]
= 100.9 + (2/9)[19.8 + 11 + 9] = 100.9 + 8.8 = 109.7
σ = 10.5%, Sharpe = (34.3-12)/10.5 = 2.12
Recommendation depends on Risk Tolerance. Portfolio 4 (Beta+Gamma+Delta) has the best Sharpe Ratio at 2.12 but drops Alpha - your highest-return project. Portfolio 3 (Alpha+Gamma+Delta) is second at 1.92 and keeps Alpha while avoiding the Alpha-Beta platform team concentration. If maximizing Risk-Adjusted Return, choose Portfolio 4. If you need Alpha's upside to hit an aggressive EBITDA target, choose Portfolio 3. Either way, don't pick Portfolio 1 or 2 - both load up on shared dependencies.
Operating Investments connects forward to Capital Allocation at the P&L level - how a CFO or Allocator decides how much total Budget each Operator gets - and to Portfolio Construction - the discipline of building a mix of investments that delivers target returns within a Volatility constraint. It also connects to Sensitivity Analysis: once you've chosen a Portfolio, the next question is 'how wrong can my ρ estimates be before the ranking changes?' That tells you where to invest in better estimation versus where rough is good enough.
Disclaimer: This content is for educational and informational purposes only and does not constitute financial, investment, tax, or legal advice. It is not a recommendation to buy, sell, or hold any security or financial product. You should consult a qualified financial advisor, tax professional, or attorney before making financial decisions. Past performance is not indicative of future results. The author is not a registered investment advisor, broker-dealer, or financial planner.