Top-down, bottom-up, and parametric modeling
Once the WBS is known, the PM team needs to provide estimates for time and cost. The best time and cost processes I've found are generally referred to as the top-down approach, the bottom-up approach, and parametric modeling. These estimation models enable a project team to more accurately predict and forecast task costs and timeframes from the good WBS created by the best practices of task listing noted above.
Top-down models: This generally involves the judgments and experiences of senior management based upon previous data and experience. It is then sent down to lower layers of management and staff to further define project costs. Small but important tasks in the WBS can be negated with this model, although it has been shown to be a good initial estimator of overall project costs at times.
Bottom-up modeling: As indicated, this is the reverse of the top-down approach. It begins at the lowest layers of enterprise staff, who offer insight into project costs initially, and then the model is sent up the management chain for refinement and changes. Bottom-up techniques tend to include most all of the costs that a project undertakes, and often create a more thorough estimate. Bottom-up models generally are regarded as more thorough and accurate than top-down models. One criticism of this method is that lower staff members can try to "empire build" through this model by adding too many costs and overstating individual importance.
Parametric modeling: This approach uses mathematical models, such as the construction cost model (COCOMO). COCOMO utilises lines of code (LOC) methodologies or function point analysis in software-related projects to ascertain project costs. Parametric models can lead to some very accurate results in providing project cost estimates and forecasts.
The issues associated with this approach are that parametric models may involve arbitrary assessments of costs that are not always realistic, and may in fact distort project estimates and forecasts. They also tend to be a little more difficult to understand and more time consuming to produce. Nevertheless, if your goal is to deliver truly good estimates and forecasts of a project's time and cost, and you want to spend the time and money on developing parametric models, then they can be very good estimators.
What's the best process?
It has been my experience that the best process and procedure in delivering good project estimates and forecasts is a combination of the methods. I have found the best results and most ease-of-use in project estimating from using the context/event/information model process for creating a great task list.
Once projects tasks are known, I generally use a bottom-up approach, where lower level players provide general estimates of time to the tasks. I usually ask for how many days or hours they expect a task to take. I then apply a 10- to 15-percent markup on these estimates.
I also ensure that I keep data on all project player estimates, and create a factor for each individual that addresses if they typically under, over, or accurately estimate their time efforts. Once this factor is applied, I then have good estimates for time per task. Knowing general labor costs and applying a good project management tool can easily derive estimates for overall project cost and time.
If an estimate for a project is needed before these best practices can be employed, I recommend the top-down approach that looks at empirical estimates from previous projects by high-level players.
I also must emphatically advise that an estimate is just that--an estimate. It is a statistical probability, and nothing else. Initial estimates without going through the best practices noted above will have a great degree of risk. The best method is using the three-model approach in task creation, along with the bottom-up method of scheduling, and entering that data into a good PM tool with cost parameters built in.
Nevertheless, CIOs should realise, and accept, that project estimating and forecasting is an on-going process, since projections improve as you move towards completion. Lastly, you should always compare end results to initial baselines to see if your model is working well for your particular organisation.
TechRepublic is the online community and information resource for all IT professionals, from support staff to executives. We offer in-depth technical articles written for IT professionals by IT professionals. In addition to articles on everything from Windows to e-mail to fire walls, we offer IT industry analysis, downloads, management tips, discussion forums, and e-newsletters.
©2001 TechRepublic, Inc.
