7 July 2013

My take on #NoEstimates

When speaking with Neil he very quickly moves past "Forecast with data, be honest about your guesses" to the following points, which I think are pretty valid.

The word "Estimating" has changed it's meaning in the software business to mean "Commitments" so we need to challenge people's expectations about how we work, and open up the typical sponsor conversation to more meaningful ones.

Schedule and budget are able to be forecast when you apply good practices around capturing past data and unpacking work requests into small more easily understood parts, what we need to do is give people a access to these idea that isn't scary or threatening. The agile movement knows that people perform better when they build up their own systems rather than follow a best practice model.

If we don't have that much trouble with time and money forecasting, why do we spend all our time on these topics when talking to clients/sponsors?  Shouldn't we be taking this time to check we are optimising the flow of value? And so shouldn't our higher order conversations be about what we are doing and course correcting on that front rather than talking about percent of budget to forecast used.

In this context Neil uses the idea of keeping your options open, and your backlog is a list of options.  There is also the idea that the further you plan into the future you are more likely to make mistakes. (And I add that your likelihood to make an error here increases if you load up one product manager/owner with too much single point accountability.)

Lastly, people need to rediscover and reinvent ideas in their own contexts, so challenging the status quo, and putting up the #NoEstimates meme provides a place for people to interact and learn from each other. That has to be good for the overall community of practitioners. (This and the other blog posts being proof points of that.)

I think Neil could probably correct me and clarify on his meaning on several points here, but in the several beers and conversations that I have had with him, this is the essence of what I hear.

I buy these as legitimate ideas. I wonder what you think?

(Neil, how did I go?)

1 comment:

  1. Craig,

    If estimates are taken as commitments without variances, you've made two failures. You've failed the statistics class and you've failed the management class. No you personally of course.

    If that is the case, and it is sometimes, find a better project to work.
    And end the discussion of why estimates can or cannot be used in specific domains.

    Assuming you get past that, the formal "basis of estimating" process used in our domain MANDATES comprehensive bottom up EAC. So that puts that to bed.

    Finally comes the planned horizon problem. You can on estimates what you can control. Rolling Waves are controllable, because they are baselined. Beyond the Rolling Wave is called "Planning Packages," not estimates. They have budget and are on baseline but they are not commitments.

    This is all very normal for us.