A top-down estimation is a great way to abstract the problem without worrying about its nitty-gritties, and come up with a workable estimates when there is no other source of getting any better estimates. (Free Estimation Ebook) At the start of a project, when making a realistic WBS is a remote possibility, the only way one could get some estimates is by top-down estimation techniques such as
- Expert Judgment
- Wideband Delphi Method
- Analogous Estimation
- Parametric Estimation
They allow an ‘order of magnitude’ estimate to be made available to set the ball rolling. We know that such estimates suffer from cone of uncertainty, but it is better to get imperfect estimate now than to get perfect estimates much later in the project – which we still need to get – the point is that we also need to get some estimates here and now!
Another key reason why we must use such top-down techniques at the start is because if we straightaway jumped into bottom-up estimation, we might face challenges because
- Details about requirements are still emerging, and hence likely to get refined as our understanding about the problem gets better
- Requirements might keep changing, thereby changing the scope of work all the time
- Too many low-level details could introduce too many moving parts in the scope of work, thereby forcing one to skip “vital few” and start focusing on “trivial many” clearly not a prudent approach at that stage
So, we need to accept the ground conditions and accept those high-level estimates with a boulder of salt and keep moving. As we get deeper into the project, we decompose the problem better and have a more granular WBS that helps us zoom into the problem and get more refined estimates. At that point, a bottom-up approach is much more relevant, as it allows us to plan and track the work much better.
Agile methodologies clearly favor bottom-up estimation methods and the idea is to slice down every task on the project into ‘stories‘ that could be ideally done by an individual team members within a couple of days. It helps you to win daily battles but takes your sight off the big war that you must win to eventually succed. We can’t really scale up these stories for large projects even if we assume that every product scope could actually be broken down to meaningful stories of that grain size. So yes, in theory, that’s great. But the real world is not obligated to obey the theory!
A prudent approach is to adopt rolling-wave planning – start with big grain size when details are not very clear and gradually move down to small grain size as you get closer to the task. However, you don’t just abandon the overall planning simply because you are now doing task-level estimates and planning! On the contrary, it is even more important to keep track of dozens of small-task estimates becuase, to quote the great Fred Brooks in The Mythical Man-Month, it is not the tornadoes but the termites that eat up your project’s schedule! So, a delay of one day here and one day there repeated for a couple of tasks over next few sprints and you suddently are sitting with a month-long delay on your overall project. You might be hitting each timebox but that’s one problem with timeboxes – it could give you a false sense of comfort that all is well, when clearly you are being forced to jetsam things that you are not able to accomplish within a timebox. On the other hand, if you go by a task-driven schedule, you atleast have a clear idea of how much late you are to complete that task.
A frequent criticism of top-down estimates is that they are at best a wishful thinking, and at worst a highly unrealistic estimate forced upon a team against their free will. Again, we should be careful in such judgment because the reality is not always what is presented to us, but rather what we do with it! So, if you take a top-down estimate as the word of god and make it ultra-resistant to changes, than you have yourself to blame when those estimates don’t reflect the reality anymore. They should be taken as a map and should be continuously refined from the data coming from the terrain. To that end, top-down estimates and bottom-up estimates play a complmentary role and an effective project manager blends them together. A good top-down estimate covers the ground fairly well, thereby allowing realistic bottom-up estimates to happen (by allocating them sufficient, meaning as close to the actuals, time and resource available), and good bottom-up estimates at task level help reinforce assumptions made in the top-down estimates of the overall project.
So, at the end of the day, it is not about which estimation technique is superior. It is about using most appropriate tool for every type of problem.
Next time, think again before you ask the question your estimates or mine?