Scrum Smells - automated ways to improve your outcome

As an agile coach I often get the work to support projects in different phases of their work. Knowing about their goals and their personal challenges I get in personal interviews and group sessions as the planning, Stand ups and of cause the retrospective.

It also shows very helpful to have a look on their backlog. As doing this research with over a dozen projects in a large organisation there were the same signs and the same topics to go through with the team.

In a cooperation with university we defined some scrum smells and did research about an automated detection of these smells. As there is a lot about learning and retrospect the current results, we were interested, if an detection of those Scrum smells will lead to some positive effect, if only the team members knew them as early as possible.

As we focused on automated detection of those smells we started with following group of smells and structured them according to their signature.

Scrum Smell on single elements:

- tasks are too big and estimated over an duration of one or two days

- stories / Backlog items are to big so they were not finished in one sprint

- definition of stories are incomplete, no Story Point Estimation, no acceptance criteria, no description

- no tasks to larger stories and backlog items

- open tasks under closed backlog items

Scrum Smells on more than one element:

- State Changes from Backlog items from new directly to done

- Backlog Items of future sprints are all estimated

- Backlog Items are already assigned to future sprints

- number of a tasks assigned to an individual team member is too high

- review tasks is only done by a few team members and not all

- number of defects higher than double number of team members

Scrum Smells over the time

- backlog items are in the backlog without action for over more than 3 months

- velocity is not stable and varies a lot

- re estimation of backlog items is done frequently

- defects and new tasks will only come in to the backlog at the end of sprint

- tasks are not updated every day, only at end of sprint

- adding items into the current sprint

- the ordering of work does not follow the ordering of the backlog

We researched three different projects, one small one, a large one with scrum of scums and an quite new one.

Our research shows that most of the automated findings and talking about them were welcome by the team members as they could improve. At the start of the research all had the resistance of being punished - the ongoing open exchange and open communication to our relief lead to a constructive learning atmosphere, in which both could learn and improve.

We stopped the integration of the scrum smells into an automated dashboard as compared to other project issues like quality and delivery speed the effort to do so, wasn't funded and also does not fit in the strategy of the university spin off helping us with the research.

It was an interesting experiment and especially when getting to the idea of automated coaching or automated scrum masters it would be nice to continue with that work. A lot of tasks of a coach are repetitive and standard so it would be nice to focus on the more difficult items than the standards of keeping a backlog clean and organized.