Monday, March 30, 2015

Agile planning and the story points approach - Part 01

We begin with this post a series dedicated to the exploration of a well-known agile planning techniques. The story points estimation method.
In this post, we will introduce some fundamental concepts with the aid of a very simple example. In subsequent posts, we will dive a little more into depth, discussing why this approach works, exposing common pitfalls and, hopefully, answering some doubts.

Let us try a little experiment.

In the middle of a field, there are three swimming pools
  • Swimming pool #1    2 meters long, 1 meter wide and 1 meter deep.
  • Swimming pool #2    4 meters long, 1 meter wide and 1 meter deep.
  • Swimming pool #3    3 meters long, 2 meters wide and 1 meter deep.

Imagine you have to empty all the three pools, one after the other, starting with the smaller pool and finishing with the bigger one, using just a bucket. Every time you pull a bucket of water out of a pool, you must pour it in a well, at one corner of the field. The bucket is known to bear a small defect that will slow down the tasks’ execution, without impeding it. You do not have any clue about what the defect is.

Could you give me an estimate of the time you will need to perform the three tasks?


No. You cannot.

You cannot, because I have not been fair to you. I deliberately have hidden some information that you would need, in order to construct a reliable estimation.
  • You do not know how big the bucket is. It may be the size of a mug or as big as a gasoline tank.
  • You do not know which the bucket’s defect is and how this will affect the tasks’ execution.
  • You do not know how far the well is. You know that the well is at one corner of the field, but you ignore how big is the area. The field could be a 10x10 meters square, or it may be as big as a football court.
A guess about the tasks’ execution time would probably be so much erratic, or its standard deviation of such a magnitude, that the estimation itself would end up to be utterly useless.

Points instead of time
Let us work with what we have. 
We know that swimming pool #2 is twice the size of swimming pool #1, and that swimming pool #3 is three times the size of swimming pool #1. 
Defined X the time required to empty Swimming pool #1, we can state as 2X the time required to empty Swimming pool #2 and 3X the time required to empty Swimming pool #3.

Since we do not have any clue about the value of X, we drop it in our estimations. 

We define that 
  • Emptying Swimming pool #1 (activity #1) is 1 a point activity.
  • Emptying Swimming pool #2 (activity #2) is a 2 points activity.
  • Emptying Swimming pool #3 (activity #3) is a 3 points activity.

Ok, but how much is a point? Velocity is the key
At this point, we can start emptying Swimming pool #1. I have decided to help you in the job. You do not have to thank me for this; I love emptying swimming pools with a broken bucket.

After a very hard work, we realize that the activity has taken us 1 day, or, that is equivalent, we have a velocity of 1 points/days. We can now estimate, using the concept of velocity, activity #1 to be a 2 days activity and activity #3 to be a 3 days activity.

Propagating and refining the estimation
The next day you start emptying Swimming pool #2. Yes, you. 
I feel a little sick and tired of buckets and water. After three days of hard work, you finish this activity too. It has taken longer than expected since I did not help you till the morning of the third day and that, probably, you felt a little tired from the previous days of hard work.

How long will take finish activity #3?
It had been initially estimated as a 3 days activity (3 points with a velocity of 1 points/days), but our velocity has changed. Using the last activity velocity (0.67 points/days), we can now re-estimate activity #3 as a 4.5 days activity. 

This approach may work, but I think it is not a good idea. I will help you in activity #3 and do not forget that tomorrow is Friday. We will have two days to rest. 
Let us estimate our velocity as the mean of the velocities achieved in the last two activities, obtaining a velocity of 0.83 points/days and an estimation of activity #3 ‘s duration of 3.5 days.

Conclusions
In a context of high uncertainty, estimate an activity’s duration could be a daunting task. We could do our best but sometimes, as Christopher Cross would say, it is not enough. An estimation would be so much erratic, or its standard deviation so wide, that the estimation itself would be completely useless.
Instead of guessing and correcting, we could rely on an agile project management concept. 

The concept is to estimate activities’ relative efforts, instead of activities’ absolute durations. 
Then, as the project progresses, we will be able to transform points into time, using performances achieved in previously accomplished activities.

The more we proceed in a project, the more the estimations of activities' durations become accurate, since averaging velocities smooth out particular events, which could have affected, positively or negatively, our performances.

Doubts
In this post, we have presented a widely used agile planning technique, the estimation of activities’ efforts by (story) points. We have introduced the approach by mean of a simple example, and we have investigated how it works.

In the next post of this series, we will talk about why this kind of approach works. We will go a little more in depth, removing some simplifications that I have introduced for the sake of simplicity. We will give evidence to some common pitfalls, and we will try to solve doubts that typically emerge when first approaching this technique.






Licenza Creative Commons
Quest' opera รจ distribuita con licenza Creative Commons Attribuzione - Non commerciale - Non opere derivate 3.0 Unported.