Becoming agile

Gantt Fetish

“A fetish (derived from the French fétiche; which comes from the Portuguese feitiço; and this in turn from Latin facticius, “artificial” and facere, “to make”) is an object believed to have supernatural powers, or in particular, a man-made object that has power over others. Essentially, fetishism is the emic attribution of inherent value or powers to an object.”
Source: Wikipedia, http://en.wikipedia.org/wiki/Fetishism

Engineering projects, and software projects are no exception, tend to become complex and challenging. Naturally, the bigger and the more individuals are involved in making it, the projects becomes more complex.

Also naturally, when organizations make projects, they wish to simplify them. To make them more visible and understood, so that they are more controllable and manageable. One aspect of this complexity is to understand the dependencies between steps towards completing the project.

Luckily there is a widespread tool to achieve visibility of such dependencies within the project: The Gantt Chart.

“A Gantt chart is a type of bar chart, developed by Henry Gantt in the 1910s, that illustrates a project schedule. Gantt charts illustrate the start and finish dates

of the terminal elements and summary elements of a project. Terminal elements and summary elements comprise the work breakdown structure of the project. Some Gantt charts also show the dependency (i.e. precedence network) relationships between activities. Gantt charts can be used to show current schedule status using percent-complete shadings and a vertical “TODAY” line...”
Source: Wikipedia, http://en.wikipedia.org/wiki/Gantt_chart Emphasis added.

A typical Gantt chart might look like this:

A Gantt Chart

Source: de.wikipedia.org, http://commons.wikimedia.org/wiki/File:OpenProj-screenshot.jpg¹

The advantages of such a chart include that it can quite clearly show the project’s critical path, its progress (how much was done, and how much is left to do), when to expect its completion, and more.

How does this relate to fetishism?

Let’s describe a plausible scenario of creating a Gannt chart with dependencies in a project. Our project consists of changes in three integrative products and ten teams of engineers across these products. In the beginning, product managers in each of the products create a list of requirements, and lay them
down on a Gantt chart, and then they work together to combine a composite Gannt chart to visualize the expected work on the entire project, including all three products.

So far there is some waste in this approach, yet it could be attributed to a necessary waste².

At this point I wish to draw your attention to a feature of Gannt charts: the milestones, typically denoted as a diamond symbol (‘♦‘), as in the chart above.

In order to meet a milestone we need to know whether there are dependencies between tasks, and in our plausible story, the VP development has asked all teams to provide all dependencies in the project: which team/person you depend upon, and which team/person depends on you. “Please provide this
information by end of the week”, said the manager.

Of course, none of this can occur in your organization. This is only hypothetical. A hypothetical plausible scenario.

The result is that all teams and all their managers are now engaged in reviewing all the activities to be completed by the end of the project, attempting to provide dependencies.

This is likely the complete opposite of principle #7 of the agile manifesto: “Working software is the primary measure of progress”.

And yet, all these people are now working towards a common goal: provide all dependencies by EOW.

Recall that fetish “is an object believed to have supernatural powers, or in particular, a man-made object that has power over others”.

Seemingly this is good: Everyone, or at least many, are now working together towards a common goal. However, the object of this goal is not the real thing (making working software). Rather, all are working towards an object that is believed to provide powers that will enable managers to better monitor and
control the execution of the project. Nothing to do with working software. Nothing a sensible customer will be willing to pay for.

Moreover, among the top reasons for project failures are changing requirements (11.8%), unrealistic expectations (5.9%) and unrealistic timeframes (4.3%), according to the 2011 Standish Group Chaos Report. According to State of Agile 2012 report, the second most important reason for going Agile is changing requirements, by 29% of respondents.

It goes without saying that investing days of work in getting all known dependencies in a Gantt chart (or any other media, for the sake of the argument) is a largely futile exercise.

In his 1988 book, The Workplace Within, Larry Hirschhorn refers (to management training techniques) as follows: “[These]… will fail, of course, if the technique functions as a fetish. Fetishes are not transitional objects. Like sexual fetishes, they block relationships between people and are used inflexibly. Social-scientific methodologies that are used indiscriminately function as fetishes.”

I do truly that Gantt charts are a good thing – as long as they are not being used indiscriminately. I have worked with a Scrum team that produced a Gantt chart during every Sprint Planning ceremony. At first I felt opposed to the idea. I talked to the Scrum Master about it, and he said it’s working for them, and I said nothing more. The truth was that it was working for them – it helped the team deliver more responsibly and predictably.
A few months later, they stopped using it – the tool was not required anymore. They were not using it indiscriminately.

At the end of the day, the object of our work is working software. Visualizing dependencies or milestones or critical paths is important if, and only if, it effectively contributes to the working software of the project it relates to.

Update – 2013.10.15:
In response to referring to his book, Larry Hirschhorn points out Eli Goldratt’s Theory of Constraints as an alternative to mapping dependencies (Thank You!).
My interpretation is – rather than spending time on identifying dependencies, look for bottlenecks (“Buffers”, in Eli Goldratt’s terms) and attend to them, until new bottlenecks are found, and repeat forever. Set the pace of the process according to the observed pace of the slowest bottleneck (“Drum”), and let all other elements in the process observe this pace (“Rope”).