Uncertainty About Commitments and Targets

An important reason for quantifying uncertainty is that it forces management to articulate beliefs about uncertainty and related assumptions. In addition, it forces an appreciation of the significance of differences between target, expected values, and commitments, with respect to costs, durations, and other performance measures. This in turn forces management to clarify the distinction between provisions and contingency allowances.

In cost terms, expected values are our best estimate of what costs should be realized on average. Setting aside a contingency fund to meet costs that may arise in excess of the expected cost defines a "level of commitment" (probability of being able to meet the commitment). The contingency allowance provides an uplift from the expected value, which is not required on average if it is properly determined. Determining this level of commitment should involve an assessment of perceived threats and the extent to which these may be covered by a contingency fund, together with an assessment of the implications of both over and underachievement in relation to the commitment.

Targets set at a level below expected cost, with provisions accounting for the difference, need to reflect the opportunity aspect of risk. Targets need to be realistic to be credible, but they also need to be lean to stretch people.

Sometimes differences between targets, expectations, and commitments are kept confidential, or left implicit. We argue that they need to be explicit, and a clear rationale for the difference needs to be understood by all, leading to an effective process of managing the evolution from targets to realized values. The ability to manage the gaps between targets, expected values, and contingency levels, and setting those values appropriately in the first place, is a central concern of project risk management.

This approach to quantifying uncertainty is useful if there is concern with aggregate performance on a single criterion such as cost or time. However, this approach is less helpful if applied to quantifying uncertainty about each activity individually in a chain or collection of activities. For example, setting commitment levels for the duration of each task in a chain of tasks may be counter productive. In Critical Chain, Goldratt (1997) describes the problem in the following terms: "we are accustomed to believing that the only way to protect the whole is through protecting the completion date of each step", as a result "we pad each step with a lot of safety time." Goldratt argues that this threat protection perspective induces three behaviors that, when combined, waste most of the safety time:

1. The student syndrome (leaving things to the last minute).

2. Multitasking (chopping and changing between different jobs).

3. Delays accumulate; advances do not (good luck is not passed on).

The challenge, and opportunity, is to manage the uncertainty about performance for each link in the chain in a way that avoids these effects and ensures that good luck is not only captured, but also shared for the benefit of the whole project. Joint management of the good luck, efficiency, and effectiveness of each project-related activity is needed. This implies some form of incentive agreement between the project manager and those carrying out project activities that encourages the generation and delivery of good luck, efficiency, and effectiveness, with the minimum of uncertainty. This agreement needs to recognize interdependency between performance measures. Duration, cost, quality, and time are not independent, and all four are functions of the motivation and priorities of those involved. For example, uncertainty about the duration of an activity is usually driven by ambiguity about quality, cost, and inefficient working practices. This needs to be managed, to reduce uncertainty and to capture the benefits of managing good luck, in the sense of the scope for low duration, high quality, and low cost.

To illustrate these ideas briefly, consider an extension of the earlier example of pipeline construction. If the activity of "pipeline design" is followed by "order pipe", "deliver the pipe", and so on, and if the internal design department is involved, do we need to set a commitment date for design completion? The simple answer is we do not. Rather we need a target duration and an expected duration for design that becomes firm as early as possible, in order to manage the good luck as well as the bad luck associated with variations in the duration of the design activity. We also need an agreement with the design department that recognizes that the design department can make and share their luck to a significant extent if they are motivated to do so. As the design department is part of the project-owning organization, a legal contract is not appropriate. However, a "contract" is still needed in the form of a "memorandum of understanding", to formalize the agreement. Failure to formalize an "internal contract" in a context like this implies "psychological contracts" between project parties that are unlikely to be effective. The ambiguity inherent in such contracts can only generate uncertainty, which is highly divisive and quite unnecessary.

Most internal design departments have a "cost-per-design hour" rate based on an historic accounting cost. A "design hours" estimate multiplied by this rate yields a "design cost" estimate. Design actual cost is based on realized design hours. The internal contract is "cost plus." The duration agreed to by the design department is a "commitment" date with a low chance of excedence, bearing in mind all the risks noted earlier. To address the problems that this arrangement induces, some form of incentive agreement is required which recognizes the potential for tradeoffs between different measures of performance. What is needed is a fixed "nominal cost" based on the appropriate expected number of design hours, with a premium payment scale for completion earlier than an appropriate "trigger duration", and a penalty deduction scale for later completion. Additionally, a premium could be introduced for the correct prediction of the design completion date to facilitate more efficient preparation of following activities. The trigger duration might be something like an 80 percentile value, comparable to a "commitment duration." The target should be very ambitious, reflecting a plausible date if all goes as well as possible. Other performance objectives could be treated in the same way, with premium and penalty payments relative to cost and quality level triggers as possible options. However premiums and penalties need to be designed to ensure that appropriate trade-offs are encouraged.