Computer-assisted schedule generation--tapping the experience and intuition of the project team
The generation of project plans and the schedules for the implementation of those plans is at the heart of project management. All of the other components of the PMBOK® will be seriously impacted if the schedule is flawed—and an alarming percentage of schedules are. Many of these flaws are recognized only after their damage has been done, and many do their damage without ever being recognized as the culprit. Wickwire and Ockman (1999) skillfully described and illustrated many of these flaws as they are manifested in the use of construction scheduling software. This paper addresses a methodology that uses the power of the computer to make it possible for all members of the project team to contribute in a graphical, interactive, manner to the design of their project's schedule. In so doing, they will produce a schedule which will be significantly more error-free and much more realistic than normal.
The Conventional Process for Producing Flawed Schedules
The generation of a project schedule is generally led by or supported by an individual known as “the scheduler.” It is the scheduler's responsibility to establish time frames for the conduct of all components of the project. Sometimes, schedule flaws result from the scheduler's having insufficient knowledge of the processes being employed in the project. Many more of these scheduling flaws, however, are a result of the scheduling practices, which have become common to the use of scheduling software of the Primavera® or Microsoft Project® type. When using such software, the most often-used approach involves naming as scheduler someone who has obtained reasonable competency in the use of the software and has some basic understanding of the technology of the project. The scheduler talks individually with some, but rarely all, of the key members of the team. In these conversations, the scheduler solicits activity descriptions, durations, and precedence information to serve as a model for that team member's portion of the project schedule. Activity direct costs and/or resource requirements may be obtained but are not required for the software to produce a schedule.
The scheduler creates an activity numbering scheme and inputs the data into the program, often adding activities for known additional components for which there is no identified or readily available expert on the project team. The scheduler may adjust durations or precedence relationships to reflect his or her experience or to reflect known pressure to achieve a particular schedule length. Precedence information may be added to integrate the input of the team members interviewed. The scheduler inputs to the program the anticipated start date and the length of the workweek. The program is then run and program-detected errors, such as a logic loop or activities without precedence tie-ins, are eliminated until a “clean” run is obtained. If the resulting project length varies significantly from the target, the scheduler may further adjust durations and/or precedence relationships of activities on the critical path(s). This may or may not involve discussions with the original sources of that information.
When the desired project length is obtained, the activity list is printed out, in tabular or barchart form. The barchart form will usually be a graphic representation of the Early Starts schedule. Early and Late Starts and Completions and activity Float are included for each activity and the “schedule” has been created. It may be distributed to the project team for review and comment, and final adjustments made to produce the “baseline” schedule.
So What's Wrong With This Approach?
A lot, that's what! Among the flaws are the following:
No Real Schedule, Only Two Schedule Extremes
The CPM software has calculated the Early Starts and Late Starts schedules. Neither of these is likely to be the best project schedule. When pressed on this issue, most project managers hesitate, and then declare the Early Starts schedule to be the project schedule. When shown this Early Starts schedule in the form of a Time-Scaled Precedence Diagram (TSPD), they invariably start explaining how that isn't really how they plan to conduct the project. Exhibit 1 is a TSPD Early Starts schedule. The Early Starts schedule has the apparent virtue of minimizing the risk that an originally non-critical activity will become critical through activity slippage or through its taking longer than the duration estimate. The Late Starts schedule (Exhibit 2) has the seemingly desirable property of delaying activity expenditures until the last possible moment, thereby reducing the cost of capital to complete the project. However, many more concerns enter into the determination of what a good schedule is.
Exhibit 1. Early Start TSPD
Exhibit 2. Late Start TSPD
Allocation of Management Attention
Early Starts schedules tend to have a simultaneous start for many activities sharing a common predecessor. When these activities have varying levels of criticality, some of them should probably be started later than others, to relieve the management crunch of having to oversee multiple simultaneous activity kick-offs. This desirable staggering of activity starts (and a possible staggering of activity completions), within the available slack, should be come part of the schedule.
Exhibit 3. Leveled TSPD
If there is a need to account for the limited availability of resources among certain groups of activities, or if there is a desire to level or smooth the utilization of resources, this will require the moving of some activities away from their Early Starts position. These new Start positions need to become part of the schedule. Exhibit 3 is a TSPD schedule that is somewhere between an Early Starts schedule and a Late Starts schedule and has slack-path activities relocated so as to reduce the peak resource requirement shown for both of the previous schedules. Some schedulers try to solve these problems by adding precedence requirements that are not technologically based, but are there only to represent a preferred flow of resources from one activity to another. Sometimes the need to flow resources in a particular manner is obvious, and this approach is satisfactory. However, the CPM software treats these precedence relationships just the same as true technological precedence relationships. The users of the schedule have no way to detect the difference between the two, and therefore they have a restricted ability to examine the true flexibility of the schedule. A TSPD on which the resource-related constraints show up differently from the true technological constraints and the strong preferential constraints are the most helpful tool for examination of the effects of resource constraints and for exploration of alternative approaches.
A resource allocation procedure within a CPM software package follows a computational scheme to place activities in locations other than Early Start in order to improve the resource utilization. Often, such computational schemes will create activity schedules that contain undesirable characteristics, such as having a resource make an awkward locational transfer from the completion of one activity to the start of another. Intelligent team members would not schedule such a move, but they may find it quite difficult to detect in the schedule before it becomes imminent in the conduct of the project. Another problem with some schemes for leveling resources is that they recognize no difference between a 3-4-3-4-3-4 period-by-period resource profile and a 3-3-3-4-4-4 profile. If a team member sees such a profile as the first, he is hard-pressed to modify it without insight into the anatomy of the schedule through the use of the TSPD.
Conventional CPM scheduling has no capability of coping with a constraint that says, “Either of these two activities can precede the other, but they cannot take place simultaneously.” Yet this can be a very real constraint, such as between Priming Wood Trim (with a Primer Which Gives off Noxious Fumes) and Installing Bathroom Wall Tile. The schedule must place one before the first but either could be first. The project team can best make such a decision after seeing the TSPD and understanding the differing effect of the two options. Of course, one could run the CPM program twice, with the sequence reversed between the two runs. If there were three activities in the non-simultaneity set and two such sets, one would need to run 36 iterations of the CPM program to select the best set, and that selection would be made using only project length as the basis for selection.
Effective Review of the Schedule and its Underlying Data is Very Difficult
The form of the output makes review of the schedule so tedious that it rarely gets done. When it is done, it is difficult for even the most conscientious team member to do effectively. Activity reports and precedence reports in tabular format are almost worthless for review. They are primarily restricted to conveying final activity timing, without providing significant context to help the reviewer or user see how an activity fits into the overall project plan and why that is its scheduled location.
Bar chart reports are more useful than tabular reports but are seriously limited by their “one activity per line” structure. These are usually printed on a series of 8 ? x 11 pages, which further complicates the review. Precedence information is often not shown, and when it is, the “spaghetti” effect of including all precedence connectors, regardless of criticality and location of the two activities being connected, limits the reviewer's ability to truly understand the underlying logic of the schedule. Lag factors are shown in a way that further obscures the meaning of the input data.
Problems Resulting From Difficulties With The Review Process
Bad Logic, Deeply Buried
Verbal descriptions of predecessors and successors obtained in the initial data collection interviews are subject to substantial error. Incorrect precedence information is often incorporated as a result of the team members' misunderstanding of, or erroneous assumptions about, the descriptions of other activities with which their activities must interface. A valid predecessor for an activity is omitted so frequently that the scheduler learns to press for these missing links. Even so, some precedence omissions will be part of every first pass at generation of a schedule. Redundant precedence information will also be generated in these data collection interviews. While creating no computational errors, these redundant precedence connections further clutter the barchart when it is generated with precedences included. Lag factors are either not put in when they would best describe activity overlapping possibilities, or are described erroneously. Frequently, precedence relationships obtained from one party reflect a strategy that is at odds with that of another party.
But the team can't see these logic errors and they remain hidden in the data files, waiting to inflict future damage.
Many important activities fail to get identified by anyone during the data collection process. They may be outside the responsibility of any team member, such as regulatory activities. The lines of responsibility may be vague enough that each party thinks that some other party will provide the information. And it is rare that any party will, on the first pass, think of all the activities that are indisputably their responsibility. Without an effective team review process, these missing activities will remain missing until a scheduled activity start is stymied by the unidentified and therefore uncompleted predecessor activity.
Duration estimates given by team members are subject to many sources of error. Lack of experience with conduct of a given activity, estimate padding as a protective step, failure to include the start-up and shut-down components of an activity, and failure to pair resource levels with activity durations are some of the common sources of error. Only when the individual providing the estimate and the rest of the team see graphically the duration, in the context of the surrounding activities and precedence relationships, will the review be more than superficial at best.
So What Can Be Done to Minimize These Problems?
We can reverse the modeling process. The Time-Scaled Precedence Diagram can be the initial venue for generating and refining the schedule. Through the use of a CAD program, activities and precedence relationships can be developed and displayed on a calendar or work-period grid. Color and line patterns can be used to classify activities into various categories. Displaying the process using the computer projector provides the opportunity for large-group participation. This Time-Scaled Precedence Diagramming/CAD/Projector (T/C/P) process eliminates a large portion of the errors that would otherwise fine their way into the published schedule.
The Scheduling Conference With the T/C/P Process
With or without prior data collection, the project team is assembled in a typical conference room. Using a high-resolution computer projector, the calendar or work-period grid is displayed on a large screen. A facilitator/scheduler leads the discussion and a CAD specialist operates the computer/projector system. The individuals provide activity descriptions and durations, or call up previously defined activities. Within seconds, the CAD operator provides the time-scaled activity representation and places it at a tentative location identified on the screen by the participant's use of a laser pointer. If predecessors or successors are already in the display, the participant points to them and the CAD operator adds them graphically, as precedence arrows, incorporating lag factor information if appropriate. If such precedence information causes a need to position the activity differently in time, the activity can be quickly move to the desired new location, keeping the precedence connections intact.
The conference can be conducted productively without prior data collection, but having some activities identified in advance makes it easier to get the process underway initially. The participants do not need any scheduling training to participate effectively and assure that their concerns are being addressed. For many team members, this will be the first time they truly understand what the scheduling process is all about. They will frequently feel empowered in a way that enhances the effectiveness and enthusiasm of the team.
Key Personnel in the T/C/P Process
A facilitator, other than the project manager, and a CAD operator who is not the facilitator, are important to the success of this approach. The facilitator should be skilled in asking the right questions to get the participants to convert their ideas into the activities that can be used to illustrate those ideas. His effectiveness will be reduced and the process slowed down if he also operates the CAD program. The project manager needs to be free to examine the emerging plan/schedule, give general direction to the process, comment or question as appropriate, and be the final word when the team cannot reach consensus on an issue.
The Group Effect
The group will, without exception, make changes to the display as more and more context is developed. Seeing the concurrency of activities in a TSPD schedule causes concerns that typical software approaches never uncover. The need for additional activities will become obvious in a way that conventional CPM scheduling will never identify. Review and modification of duration and precedence inputs is a very natural and easily implemented component of this CAD-based TSPD approach. Indeed, the ease with which changes can be made in the T/C/P process removes a barrier that inhibits change in conventional CPM scheduling, even when errors are identified.
The interaction of the members of the team as they examine the emerging schedule on the screen provides a level of enhancement of the schedule that is virtually impossible to achieve with conventional scheduling. The experience of others can add to the experience of a given input provider, giving more options to examine. Playing “What If” frequently becomes a phase of the process, with better approaches being incorporated into the schedule on a real-time basis. The experience and concerns of others also provides valuable checks on durations, precedences, and missing activities.
Sometimes a team member suggests a change in an activity's location in the schedule without being able to give CPM-type reason for the change. Such intuitive adjustments in the emerging schedule may reflect important criteria or experience that will contribute to a more workable schedule with improved buy-in of the participants. In this process, the schedule is not complete until all participants have had all their ideas examined and either incorporated or omitted because of the results of testing them in context. It is usually the case that most parties have to make some concessions to achieve a schedule which all can support. The goal of the process is sometimes stated as “getting everyone equally unhappy with what has to be done to create a good schedule.”
The T/C/P process allows interim drafts to be plotted and distributed during breaks. This hands-on copy provides increased opportunity for the ongoing review to be thorough and effective. When the conduct of the conference appears to be reaching the point of cost/value neutrality, it is time to adjourn, plot review copies, and plan for a future continuation of the schedule development, or a fine-tuning session if the total project has been preliminarily covered. Rarely can a team reach the final version in one session. The nature of the process leads to the identification of many questions that will require research. These are questions that would never have been identified in a conventional scheduling approach.
T/C/P in Updating
The TSPD/CAD/Projector approach works equally well in regular project updates, for all the same reasons that it works so well in initial project scheduling. The ability of the group, in the update meeting, to work out a process of recovery from slippages or to take advantage of better-than-scheduled performance gives a new character to the updating process. Instead of being primarily oriented to documentation of status, the update process is now a proactive, real-time process.
The Use of T/C/P With Conventional CPM Software
If processing in a conventional CPM software format is a hard and fast requirement of your situation, the T/C/P approach is still the most effective way to work out the input to the CPM program. The quality of the input data will be significantly enhanced over that obtained by the conventional approach. The schedule plot from the T/C/P approach can easily be input by a technician into the CPM software files. After a few minutes instruction, a technician can transfer the data about as fast as his keyboard skills permit. Future development is aimed at automatically converting the CAD display into a spreadsheet that can become direct input into other software for project schedule maintenance.
The T/C/P approach is in its infancy. There are no restrictions on anyone's development of their version of the process. The industry needs to have such an enhanced process for schedule generation. It is the author's hope that this paper will encourage others to address the scheduling problems described herein and elsewhere and develop similar or other improvements that will improve the quality of scheduling throughout project management.
Proceedings of the Project Management Institute Annual Seminars & Symposium
October 3–10, 2002 • San Antonio, Texas, USA