Here in the UK we’ve just had a general election, and as usual, the results differed from the conducted polls. For weeks, estimates of the number of seats and percentage of the votes put the results neck-and-neck for the two main parties. Now, the pundits and press are running around like headless chickens, and the expectations for the next few weeks have flown out of the window. Planned TV shows and editorials based on who would join with whom have now become confetti, as the Conservatives won an election that looked impossible for anyone to win.
The early exit poll predicted a win for the Conservative party. It was based on a blind sample at the polling stations and it seemed a little optimistic, but even that prediction (316 out of 650 seats) was exceeded in the final result of 331 seats. To be fair, the exit poll had the best and most up-to-date information and the margin of error was only 5%.
So, why is an IT project like an election? Like the polls, any estimate will be wrong; by definition, it’s an estimate, not reality. It’s impossible to consider everything that might happen on a project that could impact the real effort and duration, and the earlier you estimate, the less accurate the estimate will be because a lot of detail will be missing.
Like the polls, the estimate sets an expectation and people plan around it; the information is used to set the resources and schedules needed to deliver the project and to set commitments, both internally and with the client. The media made assumptions based on the early polls, and in IT project budgets, we also tend to put too much store in the early estimates, ignoring the margin of error. This is something we have to acknowledge and deal with as the project progresses and better information emerges.
The key is to use the best information available at the time to deliver the estimate and re-estimate as we know more. Reconcile multiple estimates and engage techniques such as parametric modelling and experienced-based estimating with historical data to deliver robust estimates and to plan with sufficient contingency and flexibility to deliver the project within the levels of expectation.
We know that any estimate will be wrong, but the error margin can be improved by repeating the exercise as new information is gathered. With perfect information you can get a perfect answer – in an election the perfect answer is when all the votes are counted, in an IT project, it’s at the point of delivery.
Software Sizing and Estimation Specialist