Loading ...
Sorry, an error occurred while loading the content.
 

care about estimates

Expand Messages
  • Piergiuliano Bossi
    I don t mean to be too provocative, but here it is: who cares about estimates? Well, we do. Who thinks that there should be a high correlation between
    Message 1 of 1 , Nov 5, 2002
      I don't mean to be too provocative, but here it is: who cares about
      estimates? Well, we do.
      Who thinks that there should be a high correlation between estimates and
      actuals? Again, we do and we think that every "XP shop" should seriously
      think about it.

      We have experienced during our project a gradual reduction in estimation

      error, down to 0.61% in last three months. The estimation that I'm
      talking about is the one that we have done at the beginning of the third
      renewal of the
      contract, planning roughly more than 3 months of work. At the end of the

      contract it turned out that we have done something less than expected,
      but we have also implemented some new unpredicted functionality.
      Overall,
      during those 3 months, we have therefore experienced that the estimation

      error calculated against first estimate was gradually reducing down to
      virtually zero (coming from about 30% in the 7 months before).

      Point is that our initial estimates where not doubled, nor scaled
      artificially in any means. We have just said:
      *) everything that we think it's a matter of few days (less than 5) has
      to be estimated in days
      *) everything that is more than a week has to be estimated in weeks
      *) no fractions are allowed
      That is to say: if you know less, try assuming less precision and see
      what happens. It turned out in our case that we have experienced a
      significant precision, isn't it?

      - Does it mean that we were learning to estimate? Yes, I think so, but
      there's no way for me to prove it, nor I'm really interested in proving
      it.
      - Does it mean that we will never underestimate again? I don't think so,
      every project has a story itself, in this case we had data coming from 7
      months before that helped us.
      - Does it mean that we have estimated **only** at the beginning of the 3
      months and never done it again? Absolutely NOT! We have reestimated at
      the end of every iteration, that is every 1 or 2 weeks. The final
      estimation error is just a result that we have measured as time goes by.

      - Does it mean that we are selling projects easier? No, this is a
      completely different issue. Our customers are happy to trust in our
      estimates but this is often not enough to catch a new project. Anyway,
      it may help running the project, for example when our account needs to
      manage customer's expectations, especially when dealing with allocated
      budget.

      We may discuss a lot which are the conditions and the intellectual tools
      that permit to reach reliable estimates, but before that I'd prefer to
      hear from the community what they feel about these issues, because
      thread "Estimation stuff" has made me think that there is a lack of
      attention for things like these ones.

      Don't know, maybe I'm wrong. What do you think about it?
      Cheers,
      Giuliano

      PS: as I already said in the past, in our case estimation is done in
      terms of pomodori, see June thread
      http://groups.yahoo.com/group/extremeprogramming/messages/52851?threaded=1
      for more details. 1 day = 10 pomodori per pair, 1 pomodoro = 30 minutes
      of uninterrupted work.
    Your message has been successfully submitted and would be delivered to recipients shortly.