Question about project duration probability distribution

Hi,

This is a question I already had in my mind when I read the Hyper book and it came back to me when reading the latest Tame Your Work Flow.

The following Beta probability distribution is used in both books to illustrate the probability of projects duration:

Where is this distribution coming from? Is it a mental model or is it indeed factual?

As a side note, when I collected flow time data from one team few months ago and plotted the values in a histogram I was surprised to see the distribution did indeed match roughly the one above. So I’m really curious to find more about it.

Thanks in advance,
Daniel.

It is the most common distribution in project settings. In the Kanban community there’s been a lot of research around the topic, and many seem to prefer a Weibull distribution.

In the pragmatic style of TameFlow, I would say: the actual kind of distribution doesn’t matter a lot. Just measure your own Flow Time distribution. As long as you can (1) use it to do probabilistic forecasting (preferably in an automated fashion); and (2) use it to establish buffer thresholds and/or ageing signals; then you’re all good.

2 Likes

Hi Daniel,

Steve’s comment is absolutely right. Note however that the Weibull distribution that is ‘cherished’ by the Kanban community is severely impacted by classes of service and queue discipline (first in, first served) which are also pushed by the Kanban community.

Daniel Vacanti makes a poignant testimonial of this in this video:

So one has to wonder why fostering certain habits leads to entanglements with the Weibull distribution.

These details have not eluded Tameflow!

1 Like

Thanks for the pointers Steve, Daniel.

That makes sense, one has to work with the actual distribution even if it is not the ideal one depicted by Vacanti in the video.

50th and 80th percentiles are your friends :slight_smile:

1 Like