Traditional investment analysis evaluates decisions by their expected return: projected cash flows discounted to present value, compared to the cost of capital. If the NPV is positive, the investment creates value. The framework is correct but incomplete. It treats every positive-NPV investment as worthy of approval, regardless of the consequences of being wrong.
Two projects may have identical NPVs. Project A, if successful, generates Rs 50 crore of value. If it fails, the company loses Rs 20 crore and moves on. Project B, if successful, generates Rs 50 crore of value. If it fails, the company loses Rs 200 crore, breaches its debt covenants and faces a liquidity crisis.
Same expected return. Radically different cost of being wrong. The board that evaluates both using the same framework is making a governance error that the NPV calculation does not prevent.
The Three-Variable Error Assessment
Probability of being wrong x magnitude of downside x reversibility = cost of decision error
Probability of being wrong: How uncertain is the outcome? An investment in a proven market with demonstrated demand has a lower error probability than an investment in an unproven market with projected demand. The error probability should be estimated explicitly, not embedded in the discount rate where it becomes invisible.
Magnitude of downside: If the decision proves wrong, how much does the company lose? Not just the capital invested, but the full chain of consequential losses: management time, customer disruption, reputational damage, covenant breach, opportunity cost and strategic option foreclosure.
Reversibility: Can the decision be undone, and at what cost? A reversible decision has a bounded downside because correction is available. An irreversible decision has an unbounded downside because the capital, the option and the strategic position are permanently consumed.
The product of these three variables is the cost of being wrong. It should be presented alongside the expected return for every major investment decision.
The Governance Application
In Northrop Management governance advisory work, we recommend that every major investment proposal include a dual presentation.
The expected case: NPV, IRR, payback, incremental ROIC. This is the standard analysis that supports the investment.
The error case: probability of underperformance, magnitude of loss under the downside scenario, reversibility of the commitment, consequential impact on covenants and liquidity, and the total cost of being wrong.
The board then evaluates not just whether the expected case justifies the investment, but whether the error case is survivable. An investment with an attractive expected return and a catastrophic error case requires different governance treatment from one with the same expected return and a modest error case.
This framework connects directly to the irreversibility test: irreversible decisions with large error costs deserve the deepest governance scrutiny, regardless of their expected returns.
Ashish Chaudhary, frames the principle directly: “A board that prices only the expected return is pricing half the decision. The other half is the cost of being wrong. And for irreversible decisions with asymmetric downside, the cost of being wrong may be more consequential than the expected return.”
Questions for the Boardroom
- For our last major investment decision, did we model the downside scenario with the same rigour we applied to the base case?
- What is the single largest risk to the company if this investment underperforms, and is that risk survivable?
- If this decision proves wrong, can we reverse it, and at what cost?
- Are we presenting expected returns and error costs separately to the board, or blending them into a single NPV that obscures the downside?
- Have we rejected any investment in the last three years not because the expected return was low, but because the cost of being wrong was too high?
Closing Implication
The expected return tells the board what happens if the decision is right. The cost of being wrong tells the board what happens if it is not. A complete decision framework requires both.
