This month’s focus is additional follow-up from my previous article titled "A Strong Business Case Gets Results." I made the assertion that cash flow was king and that the single point estimate was the primary problem behind obtaining support for key business case components. Part 2 will explore how to expand beyond the single point estimate by exploring the different levels of sophistication behind estimating.
Estimates Have Multiple Levels of Sophistication
In Part 1, we reviewed key components of a well-constructed business case. Three components that involve estimates are benefits, costs and value creation.
I indicated that each of these estimates should be provided in terms of ranges and should indicate the associated probabilities of various outcomes. Understanding the different sophistication levels of estimates will clarify this statement.
Single Point Estimates Fail to Provide Buy-In
Estimating is used extensively when trying to identify the costs, benefits and value associated with a given investment. In many cases, the estimates are based upon a single value which associated stakeholders are expected to support. It is very difficult to obtain buy-in with this method because the likelihood of the estimated value being right on target is slim. In fact, the chance of the value changing is high. Given this, the common response is to not support the estimate, making it extremely difficult to create a compelling business case.
Figure 1: Multiple Levels of Estimates
Level A is the point estimate and is the most familiar. For example, we request a bid for the cost of hardware and receive a quote for $200,000. We can comfortably say that the cost for hardware will be $200,000 and the likelihood of that changing is negligible. This is an example where the point estimate is acceptable. However, when a range of possibilities exist, more sophisticated estimating is necessary as shown in level B.
Level B is the range estimate. Two of the more popular methods, the uniform and the triangular distribution, are shown for level B within Figure 1. The reason for their popularity is their simplicity and better accuracy which helps develop stakeholder confidence. For example, if the number of hours estimated to build a solution is between 1,000 and 1,200 with an equal chance of the estimate falling within that range, then the uniform distribution should be selected. Similarly, if the number of hours estimated to build a solution is between 1,000 and 1,500 hours with the most likely outcome equal to 1,200 hours then the triangular distribution will more accurately reflect the estimate.
Level C in Figure 1 is an extension of the range of values estimate found in level B. It provides more flexibility for representing the behavior of a particular estimate. Allowances are made for situations when all values within the estimate range are possible (solid fill) and when only finite integer values (solid bars) exist. Determining which distribution to use is one of the most challenging aspects of this approach. Factors that influence which distribution to use are availability of information, time constraints and complexity. These factors will vary depending on the type of estimate under consideration.
Situation for both estimate examples: The labor rate is held constant at $90/hr and the estimated hours based on historical results from similar activities are: 1,000, 1,100, 1,150 and 1,600 hours.
1) Single Point Estimate. Using the single point estimate, one could take the average of the above noted hours, add 100 hours of contingency based on the fact that one of the projects was as high as 1,600 hours and use the result for the estimated cost. However, confidence in this estimate would be low based on the fact that the probability of movement up or down is not considered.
Figure 2: Understanding the possible variations of an estimate is essential for accurate broadcasting.
If costs are estimated too high and this cost is used to calculate projected cash flows for the project, the estimated project value will be underestimated. Given the competitive nature of projects today, allocated funds only go to those projects that provide the most value.
2) Range Estimate. After reviewing the historical project results one could conclude that the minimum number of hours needed is equal to 1,000 and the maximum number of hours is equal to 1,600 with the most likely number of hours needed on the low end of the range. The best estimate for this example will be 1,200 hours.
Figure 3: Range of Values for Estimated Hours
The results of this forecast will be easier to support since it provides a clearer perspective for the estimated cost by using a range of values for the estimated hours. The cost is represented as a distribution and one can quickly discern the most likely cost that will be encountered along with the probabilities associated with how the cost vary.
Utilizing these techniques for estimating the costs, benefits and value of potential business intelligence (BI) solutions is essential. Modeling certain inputs and forecasts as a probability distribution helps increase the accuracy as well as reflect real- world scenarios compelling stakeholders to support the results. In the next column, I will take this approach one step further by building a generic forecast model for a data warehouse solution.
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access