Why Stage a Strategic Commitment?
A large strategic decision is often not a single irreversible choice. An organization may fund a pilot, observe its results, and then decide whether to expand, revise, or stop. The pilot limits the initial commitment, but its purpose is not merely to make the first expenditure smaller. Its value comes from preserving a meaningful choice after uncertainty has partly resolved.
For an insurer considering a new claims platform, for example, a limited deployment might reveal whether the system performs adequately before a wider rollout. The pilot itself can fail to answer the important question, cost more than expected, or delay benefits. Staging is therefore not automatically preferable to committing at once. It should be assessed by comparing the available strategies, including their timing, costs, information, and opportunities to stop.
To compare strategies, specify the cash flows and the information available at each decision point. In the examples below, values are hypothetical, the effective annual interest rate is \(i=10\%\), and cash flows occur at the stated year-ends or decision times. Expected present values are used as a decision criterion. This criterion treats expected dollar value as sufficient for comparison; an organization may also need to consider capital limits, risk appetite, strategic objectives, or other constraints.
A Practical Decision-Gate Method
A gate is useful only when it connects new information to a feasible decision. Before launching a pilot or phase, define what will be observed, when it will be observed, which actions remain possible, and what conditions would support each action. A vague promise to “review results later” does not establish a disciplined staged strategy.
List the initial outlay, later phase costs, possible benefits, and available actions at each gate. Include a genuine stop or redesign option if one is claimed as a benefit of staging.
Describe the uncertain outcomes and their probabilities, and state what the pilot or first phase can reveal. If information is imperfect, distinguish the observed signal from the underlying outcome.
At each gate, compare the value of continuing with the value of stopping or another available action, using what is known at that gate.
Discount later cash flows consistently and include the cost of the pilot or earlier phases. Compare the resulting value with the relevant alternatives, including making the full commitment immediately or not proceeding.
The decision rule at a gate follows from choosing the best available action. If stopping has value zero relative to the gate and continuation has conditional net value \(C\), the decision value at that gate is \(\max(0,C)\). If stopping instead provides salvage value \(R\), the relevant value is \(\max(R,C)\). Taking the expectation of this choice over possible information outcomes gives the expected value of the staged strategy at that gate. The initial cost is then included when evaluating the strategy from its start.
This expression explains why a flexible strategy can be better than a forced commitment: unfavorable outcomes can lead to stopping while favorable outcomes can lead to continuing. The maximum is applied after the information is observed, not to an average calculated before the information arrives. The strategy creates no such advantage if the gate does not produce useful information or if the organization cannot actually change course.
Comparing a Pilot with Immediate Commitment
Worked Example: A Pilot That Reveals Whether Expansion Will Succeed
A hypothetical insurer can either invest $100,000 immediately in a claims automation project or run a pilot. With probability 0.40 the project is in a successful state; otherwise it is unsuccessful. The immediate investment produces a $180,000 benefit at the end of year 2 only in the successful state. The pilot costs $20,000 now and reveals the state perfectly at the end of year 1. If the state is successful, the insurer can invest $100,000 at year 1 and receive $180,000 at year 2; otherwise it stops. Assume no other cash flows, no salvage value, and \(i=10\%\). Compare expected present values at time 0, including the option not to invest.
Method. Discount all future cash flows to time 0. For the immediate strategy, the expected year-2 benefit is \(0.40(180{,}000)=72{,}000\). Thus:
For the pilot strategy, the successful state occurs with probability 0.40. Conditional on success, the value at year 1 of expanding is \(-100{,}000+180{,}000/1.1=63{,}636.36\). Its time-0 value is \(63{,}636.36/1.1=57{,}851.24\). If the state is unsuccessful, the insurer stops and incurs no expansion cost. Therefore:
The insurer should prefer the pilot to immediate commitment under these assumptions, and the pilot has positive expected value relative to doing nothing. This result depends on the pilot’s perfect information and the ability to stop without further cost. Common mistake: subtracting the full $100,000 expansion cost at time 0 in every state. The expansion cost is incurred at year 1 only if the pilot identifies success. The calculation above discounts it at that time and assigns it only to the successful state.
Use the Signal Observed at the Gate
A pilot often provides an imperfect signal rather than revealing the true outcome. The decision should then use the probabilities conditional on the signal. An unconditional average can hide the fact that a favorable signal supports expansion while an unfavorable signal does not.
Worked Example: Expanding Only After a Favorable Signal
A hypothetical service operation has a high-demand or low-demand state, each with prior probability 0.50. A pilot costs $5,000 now and produces a favorable or unfavorable signal at the end of year 1. The probability of a favorable signal is 0.80 in the high-demand state and 0.20 in the low-demand state. At year 1, the operation may expand for $60,000; expansion pays $100,000 at year 2 in the high-demand state and $20,000 in the low-demand state. Stopping has value zero. Assume \(i=10\%\) and no other cash flows. Find the optimal gate decisions and the pilot strategy’s expected present value at time 0.
Method. First calculate the probability of each signal. The probability of a favorable signal is \(0.50(0.80)+0.50(0.20)=0.50\); the unfavorable signal also has probability 0.50. Bayes’ rule gives \(P(\text{high}\mid\text{favorable})=0.80\) and \(P(\text{high}\mid\text{unfavorable})=0.20\).
After a favorable signal, expected year-2 benefit is \(0.80(100{,}000)+0.20(20{,}000)=84{,}000\). Its value at year 1, net of the expansion cost, is \(84{,}000/1.1-60{,}000=16{,}363.64\). This exceeds the value of stopping, so the operation expands.
After an unfavorable signal, expected year-2 benefit is \(0.20(100{,}000)+0.80(20{,}000)=36{,}000\). The year-1 value of expansion is \(36{,}000/1.1-60{,}000=-27{,}272.73\), so the operation stops.
The expected value at year 1 of following these gate decisions is \(0.50(16{,}363.64)+0.50(0)=8{,}181.82\). Discounting this amount and including the pilot cost gives:
For comparison, immediate expansion for $60,000 would have expected year-2 benefit \(0.50(100{,}000)+0.50(20{,}000)=60{,}000\), giving NPV \(-60{,}000+60{,}000/1.21=-10{,}413.22\). The pilot strategy is better than immediate expansion and doing nothing under the stated assumptions. Common mistake: using the unconditional $60,000 expected benefit as if it were the expected benefit after a favorable signal. The favorable signal changes the conditional expectation to $84,000, which is why expansion is worthwhile in that case.
Stopping Decisions, Salvage, and Sunk Costs
At a gate, the organization compares the future value of continuing with the value of stopping. An asset or capability may have a salvage value if the project stops. That value belongs in the stopping alternative, but should not also be added to the continuation value unless it is actually received in that scenario.
A pilot cost is different: once incurred, it is sunk at the gate. It matters when comparing strategies from the original decision date, but it is not a reason to continue if the future benefits of continuing are inferior to the available stopping alternative. The next example separates these comparisons explicitly.
Worked Example: Applying a Gate with a Salvage Alternative
A hypothetical insurer spends $10,000 on a pilot now. At the end of year 1, the pilot reveals the state perfectly: high performance has probability 0.40, and low performance has probability 0.60. At the gate, the insurer can stop and recover $4,000 immediately, or spend $50,000 to expand and receive a benefit at the end of year 2: $90,000 in the high state or $40,000 in the low state. Assume \(i=10\%\), no other cash flows, and that salvage is received only if the insurer stops.
Method. Compare the alternatives at year 1, when the information is known. In the high state, expansion is worth \(-50{,}000+90{,}000/1.1=31{,}818.18\), which exceeds the $4,000 salvage value. Continue. In the low state, expansion is worth \(-50{,}000+40{,}000/1.1=-13{,}636.36\), which is less than the $4,000 salvage value. Stop.
The expected value at the gate is \(0.40(31{,}818.18)+0.60(4{,}000)=12{,}727.27+2{,}400=15{,}127.27\). Its present value at time 0 is \(15{,}127.27/1.1=13{,}752.07\). Subtract the original pilot cost:
If the insurer were forced to expand in both states, the gate value would be \(0.40(31{,}818.18)+0.60(-13{,}636.36)=4{,}545.45\). Its time-0 present value would be \(4{,}545.45/1.1=4{,}132.23\), and the NPV after the pilot cost would be \(-5{,}867.77\). Flexibility changes the low-state action from expansion to stopping. Common mistake: including the $10,000 pilot cost in the year-1 comparison. It has already been spent, so it cannot make expansion preferable to salvage. It is correctly included when calculating the strategy’s overall NPV.
When Staging Helps—and When It Does Not
Staging is most useful when several conditions hold together: uncertainty is material, a phase can reveal information relevant to the decision, later commitments are substantial, and the organization can act on what it learns. A small pilot that measures an irrelevant outcome may add expense without improving the decision. Likewise, a formal gate has little practical value if contracts, regulatory obligations, or operational dependencies make stopping infeasible.
Staging also has costs. It can delay benefits, require duplicated setup, increase unit costs, consume management attention, or cause an opportunity to expire. A complete comparison includes those effects as well as the costs of each phase. The comparison must use consistent time units: a cost at year 1 is not a time-0 cost, and a benefit received at year 2 must be discounted accordingly.
A useful governance practice is to establish decision criteria before observing the pilot results. Criteria might address performance, implementation cost, customer impact, or operational readiness. Predetermining them reduces the risk that decision-makers reinterpret a disappointing result to justify continuing. The criteria should still allow appropriate judgment when the data expose an unanticipated issue; they are a discipline for decision-making, not a substitute for it.
Check Your Understanding
For each hypothetical situation, use \(i=10\%\) where a rate is needed. Treat stopping as value zero unless a salvage value is stated. Amounts are in thousands of dollars.
- A pilot costs 6 now and reveals success perfectly at year 1, with probability 0.30. If successful, expansion costs 40 at year 1 and pays 60 at year 2; otherwise the organization stops. Find the pilot strategy’s expected NPV and decide whether to undertake it rather than do nothing. (Answer: NPV \(=-2.03\); do not undertake the pilot.)
- At a gate, the posterior probability of high demand is 0.60. Expansion costs 50 at year 1 and pays 66 at year 2. Find its conditional net value at the gate and the decision, assuming the payment occurs in either demand state. (Answer: \(66/1.1-50=10\); expand.)
- At a gate, expansion costs 80 at year 1 and pays 110 in the high state or 55 in the low state at year 2. If the conditional probability of the high state is \(q\), find the threshold for expansion rather than stopping. (Answer: expand if \(q>0.60\); be indifferent at \(q=0.60\).)
- A perfect-information pilot costs 4 now. At year 1, it reveals a favorable state with probability 0.25; expansion then costs 30 at year 1 and pays 50 at year 2. In the unfavorable state, the organization stops. Find the strategy NPV and decide whether it beats doing nothing. (Answer: NPV \(=-0.49\); do not undertake the pilot.)
- At a year-1 gate, stopping returns salvage of 5. Continuing costs 40 at year 1 and produces a certain benefit of 55 at year 2. Find the gate value of continuing and choose between continuing and stopping. (Answer: continuation value \(=-40+55/1.1=10\); continue rather than take the salvage value of 5.)