Good business case analysis will not eliminate uncertainty about the results of business decisions, but it can reduce uncertainty to a minimum, measure what remains, and provide the tools for minimizing risk as the action goes forward. Risk and sensitivity analysis are key to reaching these objectives.
Senior business managers say, increasingly, that the margin of tolerance for management error is shrinking, visibly and tangibly. Decision makers today are driven by watch words like "comfortable certainty" and "accountability." Decisions to invest, enter new markets, change strategy, or launch alliances have always come with risks as well as potential rewards, but in the current business climate there is a new urgency to questions like these:
- How do we know that we're going to see the projected results?
- Are we sure this is the best business decision?
- How can I prove, later, that I'm acting responsibly now?
Most people know that business case analysis is supposed to answer questions like these, firmly and finally. That is what "decision support" is supposed to mean, after all. Nevertheless, decision makers with business case results in front of them, are still haunted by these questions. They ask: Is there any certainty to be had in the business world? This article presents an overview of risk and sensitivity analysis in the business case, the primary approach for addressing such questions directly.
The sections below further describe the rationale and approaches to business case risk reduction. For more in-depth examples, see Business Case Essentials. For guidance in building a dynamic financial model for your case for simulation-based risk and sensitivity analysis, see Financial Modeling Pro
- How can business case risk be reduced, measured, and managed?
- Risk and Sensitivity Analysis: Think elementary
- Can you trust the business case?
- What are next steps for understanding business case risk?
How can business case risk be reduced, measured, and managed?
Reducing uncertainty to a minimum and measuring what remains requires best practice cost and benefit estimates and a serious risk and sensitivity analysis. Those are unlikely to be present, however, if the business case situation fits one of these "worst case" scenarios.
What are some common worst-case approaches to business case risk?
Worst case 1: The "Fund me!" request.
You have no doubt heard business people talk about previous business case work in terms like this: "My business case was successful--I got my funding!"
People write cases all the time to support funding requests and that is as it should be. But there's an important difference between cases designed to prove that funding is a good idea, on the one hand, and cases designed to find out if funding is a good business decision on the other.
Compare the case to a scientific research project. Research scientists have their own theories and they often have a strong ego-involvement in seeing their theories proved correct. Good research experiments, however, are designed to test theories, not prove them right. Give your pet theory every chance to fail, try seriously to make it fail, and give competing theories their best chance to prevail. If your theory still stands, so much the better. That's the only way to "prove" a theory with credibility.
Similarly in business, support for funding will be credible if the business case shows convincingly that costs are not underestimated, that benefits are not overly optimistic, that important risks are in view, and that alternate actions have been compared fairly with the requestor's proposal.
Worst case 2: The crystal ball syndrome
"I have looked into the crystal ball and I see $10M net gain if we implement my proposal." The words may be slightly different each time, but that's almost the message that comes with many case results and recommendations.
Good business case analysis is not the output of a crystal ball or a "black box" predicting program, to be trusted and believed because the methodology comes with an impressive pedigree or a good track record. Instead of the crystal ball message, good business case results communicate this idea:
"I have drawn several pictures of the way the future may work out. These are detailed and concrete scenarios, based on assumptions about many factors, each of which comes with some uncertainty (future prices, resource requirements, competitor actions, market growth, business volume, government actions, inflation, currency exchange rates, and many other things that are not certain). However, if the assumptions stand, these results certainly follow."
All the uncertainty, in other words, should lie with assumptions underlying business case scenarios—not with how the case builder developed the cost/benefit estimates. And, the important assumptions should be in plain view.
Can you ever be really certain?
You can never absolutely certain about anything having to do with the future, including projected business results. However, assuming that your business case avoids both of the worst case problems above, there is a level of certainty just below "absolute" that business case builders can aim for and case users can ask for.
We can be very, very sure (say, 99% certain, or 90% certain), that an action or decision will bring business results within a given range. And, if that range is too broad for comfort, there are steps you can take to narrow the range--if you're willing to put more resources into the business case project. That range is called a confidence interval (described below).
Risk and sensitivity analysis: The underlying principles are simple
Questions about certainty in business call for good risk analysis. "Good," by the way, does not have to mean "advanced" or "inaccessible to ordinary business people."
The kind of certainty claimed here comes from the world of elementary statistics. To learn and use what follows you will need a few probability terms from the basic introductory statistics course, the one for "non-statistical people". You will not need to go beyond that. Some of the leading risk analysis software tools come with excellent user guides that do not even assume that much background.
The purpose here is not to teach risk analysis, but rather to give some sense of what it can do. (for complete introduction to applied risk and sensitivity analysis, see Business Case Essentials).
Question 1: How do we know we're going to see the projected results?
The business case for entering a new product market projects a net gain of $590,000 over the next five years. Everyone knows, however, that the real result will not be $590 thousand, exactly: it will be something more or something less. But can the business case builder say anything more about the likelihood of other results? Here is one kind of result to aim for:
Probability distribution developed with Monte Carlo simulation. The graph shows the probability of reaching or exceeding different net gain figures
This graph shows the probability of reaching or exceeding different net gain figures. It is based on a simulation exercise with the same financial model that produced the initial $590,000 estimate. Instead of showing just the most likely outcome, however, it says that the probability of realizing at least $329,000 is around 90%, while the probability of seeing $750,000 in gains, or more, is around 20%.
Moreover, the 80% confidence interval (dark blue area under the curve) says they can be 80% confident that the actual results will be between $329 and $789 thousand. This kind of answer may provide the "comfortable certainty" decision makers need, If they trust these results (see below). If the new product venture must bring gains of at least $300,000 or more in order to be worthwhile, decision makers might be quite ready to act on a 90% probability. If they must have $750,000 or more, this proposal does not look like a wise decision.
Question 2: Which is the better business decision?
Management has two competing proposals on the table, each for a different product launch. One comes with a business case projecting a net gain of $590,000 across five years. The other looks forward to a projected five-year gain of $800 thousand. The company can only afford to invest in one of these initiatives. Which one should they choose?
Decision makers will weigh the proposals against each other in several ways, considering strategic fit, the competition, economic trends, market trends, and other factors no doubt. When the overriding concern is return on investment, however, decision makers may approach "comfortably certainty" with another kind of statistical guidance based on the same simulation exercise:
|Proposal||Expected Value||80% Confidence Interval|
|Product launch A||$590,000||$329,300 - $789,500,000|
|Product launch B||$800,000||$251,000 - $1,390,000|
Proposal A has a lower expected value (most likely result), but it comes with a much narrower confidence interval. If management chooses proposal B, they have an 80% level of confidence that actual result will be somewhere between $329 and $789 thousand. The difference between these numbers is the range of "uncertainty," about $462 thousand.
By contrast, the proposal B comes with a higher expected value but a much wider range of uncertainty. Here, the 80% level of confidence stretches across $1,139 thousand of possible results (that is, $251 through $1,139 thousand).
Assuming they believe these results decision makers may take the narrower confidence interval as the comfortable certainty they need to act.
Question 3. How can I prove, later, that I'm acting responsibly now?
Accountability is becoming the watchword for decisions in project management, capital spending, policy changes, and other commitments of all kinds. Indeed, accountability angst may be the only way to explain why so many people start a financial justification business case, even though the decision is already made and the money already spent. Many ask "Can the business case really prove that I'm making a solid, responsible decision? Where does the proof come from?"
Accountability in the future comes from the same steps that establish trust and belief in case results today. We have seen above the kind of "certainty" that good business case results can claim. Results like those in Questions 2 and 3 above are useful--to those who believe them. But can anyone believe the claims? How far can you trust statistical results?
The key to making a decision on trust lies in understanding the assumptions behind the cost and benefit estimates. Remember that good business projections should not be viewed as the output of a "black box" predicting system or a high-powered analysis that very few people understand. Ideally the case builder wants to position business case results as described above. It is worth repeating the case positioning that case builders should aim for:
"I have drawn several pictures of the way the future may work out. These are detailed and concrete scenarios, based on assumptions about many factors, each of which comes with some uncertainty. However, if the assumptions stand, these results certainly follow."
Whether or not we trust the projections depends on what we know and believe about the assumptions. Below is a simple example, using a single cost item estimate and three assumptions. The approach is the same, however, when analyzing a full cost/benefit cash flow statement from the case.
Can you trust the business case?
Management in a network services company is deciding whether or not to launch a new customer service offering. They will go forward with the service only if they are confident it will be profitable. This calls for a decision support business case, with a large number of estimated cost and benefit items.
Here is their approach for estimating one cost item, annual labor costs for service delivery. The estimated is based on just three assumptions.
|Assumption||Least likely value||Average||Highest Likely|
|No. of service calls per month||70||100||135|
|Onsite time required per call (hrs)||1.00||4.00||7.00|
|Labor rate plus overhead ($ / hr)||60.00||80.00||100.00|
The average values are the most likely values in the eyes of case builders using the expected averages:
Annual labor cost estimate
= onsite time * Labor rate
= 12 * 100 * 4.00 * $80
The road from assumption to cost estimate is out in the open for all to see and evaluate. If the assumed values turn out exactly at the averages given, no doubt about it, the annual cost of labor will be $384,000.
Should management trust that estimate? Remember, these decision makers are highly risk averse. They will go forward only if they have confidence in the projections. If the cost of labor really turns out to be more than $500,000 the service becomes unprofitable. How likely is that?
Run the future 10,000 times
Risk analysis addresses such questions by applying Monte Carlo simulation1 to a financial model. In this case, the "financial model" is just the formula above.
Notice first the ranges of possible values for each assumption: the case builder expects an average 100 calls for month, but in fact the number could be more or less. However, the case builder is very certain that it will not be less than 70 or more than 135. The other assumptions also can--and probably will--differ from the expected value, but the case builder is very sure they will fall within the ranges given.
With the assumptions viewed as ranges of possibilities, and with a few more assumptions about the likelihood of different assumption values2 we can use Monte Carlo simulation to "run the future" 10,000 times and summarize the different cost estimates that appear. Here is what the case builders found with their first-pass simulation run3:
Cost results risk probabilities from Monte Carlo simulation. The graph shows the probability of reaching or exceeding total cost figures.
There is a 50% probability of seeing an actual cost of $384,000 or more. That's no surprise: we already knew the expected average. But these risk-averse decision makers are more than a little uncomfortable with a 20% chance of having actual costs hit $500,000 or more. Moreover, the 90% confidence interval for the cost estimate is very wide: $240,000 through $530,000.
Can they do anything to build confidence in seeing that labor costs stay low enough to make the service profitable? Some very clear guidance comes from another result of the simulation exercise: sensitivity analysis:
Cost results risk probabilities from Monte Carlo simulation. The graph shows the probability of reaching or exceeding total cost figures.
While performing the risk analysis, the simulation program was also keeping track of the correlation between each assumption and the forecast result (here, the cost figure). This chart tells us that the dominant assumption, by far, in controlling different cost results, is the assumption about time spent per service call.
If case builders can reduce the uncertainty in the "time per call" assumption, they can reduce the uncertainty in the projected cost. Here, it turns out that more discussions with the service product manager involved, and with the service personnel training manager, led the business case builders to narrow the range of "near certain" average onsite time requirements from 1-7 hours to 3-5 hours (very occasionally there may be a much longer or much shorter onsite call time, but they are very sure now that the average onsite time will be in the narrower range, 3-5 hours).
Case builders also reduced the range of estimates for "average labor rate" from $60-$100/Hr to a much narrower range of $75-$85 per hour. Very sure now that the assumptions will fall in the narrower ranges, the case builders re-ran the simulation program and produced this result:
The most likely result is still $384,000, but now the probability of an actual cost at $500,000 or more is virtually 0, if you can believe the new assumptions about the assumptions!
Furthermore, management can be 90% confident that the actual cost figure will fall between $344,000 and $422,000. If they trust these results, they are ready to implement the service.
Should you Trust These Results?
The question of trust all comes down to what you believe about your assumptions. If the (now narrower) ranges of possibilities for the assumptions stand, then the probabilities in the final summary graph can be trusted.
What are next steps for understanding business case risk?
For a practical introduction to applied risk and sensitivity analysis, as shown on this page, see Business Case Essentials. For guidance in building a dynamic financial model for your case for simulation-based risk and sensitivity analysis, see Financial Modeling Pro.
1. This approach is called Monte Carlo simulation because it uses probabilities the same rules that operate on the gaming tables in the Grand Casino in Monaco. The illustrations in the example are produced with a leading Monte Carlo simulation program, "Crystal Ball"® from Oracle, Inc. (//www.crystalball.com/. Crystal Ball is a registered trademark of Oracle).
2. The likelihood that assumptions take on different values is not shown in this issue, but the approach is illustrated in Business Case Essentials and our other books, as well as the documentation and online help that comes with the leading Monte Carlo simulation programs, such as Crystal Ball.
3. In a nutshell, the simulation program chooses a new set of assumption values, based entirely on what we have just told it about the assumptions, looks at the cost estimate that results, and then repeats the process thousands of times.