The Best Estimating Technique

The Best Estimating Technique

When you think of estimating and planning techniques, what method do you think of first?  Do you use a parametric technique, Monte Carlo simulation, or something else?  I was over at the Implementing Scrum website and found an awesome post, asking this same question.  If you think people get crazy whether or not you should utilize an Agile approach or a traditional project management approach on a project, wait until you talk about how to estimate work.  In the traditional project management world, I’ve seen people use SLOC and PERT to arrive at an estimate.  In the Agile world, I’ve used story points and have seen others use gummy bears and t-shirt sizes.

In the end, I really don’t care what estimation technique is used and I’m pretty sure the customer won’t either.  All anyone should be worried about is if the estimates are accurate and if you get the work done.  Am I right or am I wrong?  I would love to read your input.

Thank you to Mike Vizdos for letting me use his cartoon and for the inspiration for this post.

4 Replies to “The Best Estimating Technique”

  1. Derek, I caution the use of the desire to get “accurate” estimates. That will inevitably lead you down a path of trying to get more and more accuracy for something that cannot be perfect. Kind of like forecasting in supply chain management. You need to know the basic shape and size of the work to a good enough level of detail. It also has to be acceptable to management to go under and over the estimated completion. Time buffers are very handy for this.

    1. I certainly hear you. I don’t believe actuals and estimates should be expected to match. If they do, someone is either throttling their work somewhere or they spent too much time estimating. I’m not the Amazing Kreskin and neither are my team. I don’t expect them to be dead on. I just want them in the ballpark. Without going all statistical, we should always assume there will be an acceptable variance in estimates.

      Thanks for keeping me honest. You make a really good point.

  2. Wow, if people don’t understand what is being asked to deliver how can they provide an estimate? If you ask me to deliver an apple and I think you want an orange, no estimate in the world is going bridge that gap. If we start with “deliver a fruit”, we’re at least on the right path. Impatience is a productivity killer! I wouldn’t doubt people saying “we don’t have all day…” at the beginning of the process results in a longer development cycle, due to rework later.

  3. Accuracy (removing bias) in estimates comes from modeling previous estimates against previous actuals. “Velocity” is such a model. Precision is what can’t be pursued relentlessly.

    However, a good model will also report estimating precision, aka uncertainty, via a prediction interval.

    Just the act of estimating can get a project down from a 25%-400% range to a 80%-120% range.

    That alone is worth real money, just like investing in a lower-volatility stock in the stock market.

Leave a Reply

Your email address will not be published. Required fields are marked *