Saturday, May 12, 2007

Invisible parameters in estimations

It is a good thing to determine and estimate the plans before you actually start doing it. I was unexpectedly invited into a directors meeting to share my ideas and discussion. There was a director and an operations manager in the room. They were discussing about estimating the load on the site and resource management and at the same time the budget. It was a startup company so it was not surprising to me at all. They already have some statistics on engineers work, increasing load and so on.

From the available statistics, no of upcoming requests and the average no of requested by the engineers, the management wants to know, in order to finish the estimated tasks, how many working days it would take. And what if we keep only the engineers who are overperforming the serves per day. The problem is very simle, in deterministic terms. Let me restate this problem in my way: "We have N tasks to be finished and M workers. And determine the days count in which this has to be finished." A simple math will definitely give the solution for this.

But what acutally the big people thinking here is about the undeterministic part of it. Few tasks cannot be finisheed just like that. Say it may take an year to move it to the resolution. If we take the non transparent attributes in to consideration here, like the criticality of the task, levels of resolution stages, budgent constraints and last but not least and important thing the resource management, it crystalizes the dint of this whole discussion. Finally they concluded that they need a tool which can manage this kind of information well and reports them the answers as per their queries. If the problem given to is deterministic you can easily solve it, but if it has some inevitable intricacies, it becomes tedious and descernible problem. Many tools are now availabe in the market to answer such kind of questions. But a delphi kind of approach is allways better, thats why they could have invited me to join them in thier discussion. Finally we decided to use Spread Sheet for the time being as it has more features than what we are expecting :) So if you can figure out how to measure the invisible parameters, you can estimate and predict future happenings :)

No comments: