Reading and summarizing books on lean software development, so you dont have to. Part 2 (see Part 1).
“Lean Software Development in Action” written by Andrea Janes and Giancarlo Succi and published 2014 at Springer. The authors are scientists at the University of Bolzano, and the book clearly has a more scientific approach than the last one.
As the last – and probably every book – on the matter of lean, agile and software engineering this book starts with an introduction on each of those aspects. Again, I will not reiterate on what lean and agile are, but focus on interesting observations and perspectives exposing new angles on “known stuff”.
The first noteworthy piece is about “tame and wicked projects”. This section is referring to work by Rittel and Webber who came up with a discution between tame and wicked problems (see also). Poppendieck and Poppendieck extended this to projects, and in this book they are described and identified by following ten points:
- Wicked projects cannot provide a definitive, analytical formulation of the problem they target. Formulating the project and the solution is essentially the same task. Each time you attempt to create a solution, you get a new, hopefully better, understanding of the project.
- Wicked projects have no a stopping rule telling when the problem they target has been solved. Since you cannot define the problem, it is almost impossible to tell when it has been resolved. The problem-solving process proceeds iteratively and ends when resources are depleted and/or stakeholders lose interest in a further refinement of the currently proposed solution.
- Solutions to problems in wicked projects are not true or false, but good or bad. Since there are no unambiguous criteria for deciding if the project is resolved, getting all stakeholders to agree that a resolution is “good enough” can be a challenge.
- There is no immediate or ultimate test of a solution to the targeted problem in a wicked project. Solutions to such projects generate waves of consequences, and it is impossible to know how these waves will eventually play out.
- Each solution to the problem targeted by a wicked project has irreversible consequences. Care must be placed in managing assumed solutions. Once the website is published or the new customer service package goes live, you cannot take back what was online or revert to the former customer database.
- Wicked projects do not have a well-described, widely accepted set of potential solutions. The various stakeholders may have differing views of what are acceptable solutions. It is a matter of judgment as to when enough potential solutions have emerged and which should be pursued.
- Each wicked project is essentially unique. There are no well-defined “classes” of solutions that can be applied to a specific case. It is not easy to find analogous projects, previously solved and well documented, so that their solution could be duplicated.
- The problem targeted by a wicked project can be considered a symptom of another problem. A wicked project deals with a set of interlocking issues and constraints that change over time, embedded in a dynamic and evolving context.
- The causes of a problem targeted by a wicked project can be explained in several ways. There are several stakeholders who have various and changing ideas about what is the project, its nature, its causes, and the associated solution.
- The project must not go wrong. Mistake is not an option here. Despite the inability to express the project solution analytically, it is not allowed to fail the project.
These are some interesting observations which resonate with my experience. One might say “many of those points do not apply to our project as we have a quite clear understanding of our product delivery (like an ECU for a car, providing a certain set of functions/features for the customers)”. However I think its not that simple. Many projects of non-trivial complexity I have been involved in do not only have the goal of releasing a product to the market, but there are other, interlinked objectives which give the project in total a semi- or non-defined goal. Besides delivering a good, innnovative product these objectives may consist of e.g. financial goals (better return on investment), efficiency gains, usage of new technologies and approaches and in- or outsourcing of activities. While project I know typically start with those defined on a high-level and people are onboarded or recruited referring to those motivating goals, as soon as the project enters the death march the project’s goals are gradually getting more fuzzy, unbalanced and volatile. I don’t know silver bullets to such situations (yet), but the notion of wicked projects resonates with those and other observations. However, isn’t awareness the first step to improvement?
[…] the quality of the final product is seen as a result of the process pro-page 41
ducing them. This assumption creates a high attention for a high-quality production process according to the credo “prevention is better than healing”
This lean wisdom sounds trivial, however, I have never seen it realized. I have yet to understand why so many managers ignore the efficiency and effectiveness gains by a proper process and instead decide to continously apply brute force which cost them more money, time, energy, motivation and subsequently of course quality. And, of course, by a “proper process” I dont mean an perfect considers-everything-process which is both unreachable and undesirable.
“The role of standardization”, page 43
I like this illustration showing standardization as a wheel chock for the plan-do-study/check-act cycle. Similar to the paragraph above it strikes me how many projects and managers re-invent the wheel (haha) and then start the whole process from the bottom. Of course no one wants and says this, but this is what often happens.
The result is a development approach in which requirements are not “refined down to an implementation,” i.e., taken as the starting point to develop an implementation that represents those requirements, but where the business objectives are mapped to the capabilities of the technical platform to “equally consider and adjust business goals and technical aspects to come to an optimal solution corresponding to the current situationpage 63
This is another insightful depiction. It provided me a new perspective on requirements, as they help to close the gap between business goals and technical capabilities. This can be a good approach helping in situations in which a project is lacking a good “feeling” on the right amount of requirements, between over-specification and under-specification.
Referring to studies conducted by Herzberg this shows how motivators and hygiene factors influence the motivation of the staff, and how agile methods very well support/complement those.
Later the authors come to write about the “dark side of agile”, on which they also published a paper. As observed by others before, agile statements can be easily twisted and thwarted in good or bad faith to yield an abomination which leads to the opposite and extreme positions of the original intention. Citing Rakitin’s old paper the agile manifesto can be translated as
- Individuals and interactions over processes and tools: “Talking to people instead of using a process gives us the freedom to do whatever we want.”
- Working software over comprehensive documentation: “We want to spend all our time coding. Remember, real programmers don’t write documentation.”
- Customer collaboration over contract negotiation: “Haggling over the details is merely a distraction from the real work of coding. We’ll work out the details once we deliver something.”
- Responding to change over following a plan: “Following a plan implies we have to think about the problem and how we might actually solve it. Why would we want to do that when we could be coding?”
The above is one way to twist the agile manifesto in favor of what the authors a few paragraphs later call a “cowboy coder”. This reminds of the “Programming, Motherfucker” webpage (thanks, Kris). While such sentiment exists, I cant often blame the engineers on mocking the agile manifesto and similar approaches that way. Very often, such engineer’s reactions are preceeded by even more unfaithful perversions pushed by all sorts of management. Just to bring one example to the table: Who has not witnessed a manager throwing new (vague) requirements at the development team every other week claiming this is what agile is about and, of course, everyone has to be faster in reacting. Because agile is faster. I could go on.
In chapter 6 the authors start to synthesise and bring lean and software development together. They start with citing the seminal book of Poppendieck and Poppendieck “Implementing Lean Software Development From Concept to Cash”. I couldnt read it yet as it wasnt available in print when I tried to get a hold of it. So they provide 7 principles for lean software development:
- Eliminate waste;
- Build quality—we used the terms “autonomation” and “standardization”;
- Create knowledge;
- Defer commitment—we used the term “just-in-time”;
- Deliver fast—get frequent feedback from the customer and increase learning through frequent deployments;
- Respect people—we used the term “worker involvement”;
- Optimize the whole—we used the term “constant improvement.”
Nowadays, those principles may sound obvious. However, the Poppendieck book was published in 2006 and I think at that time many if not all of those principles where both not clear nor any best practices and tooling was available back then to realize them.
In a break-out box a comparison between lean and agile is given
- Agile Methods aim to achieve Agility, i.e., the ability to adapt to the needs of the stakeholders.
- Lean production aims to achieve efficiency, i.e., the ability to produce what the stakeholders need with the least amount of resources possible.
After this section which gives some more references to earlier work the book enters its less interesting but extensive part. Janes and Succi present three methods which shall support lean software development. They all have their merits, but I have to admit I dont catch fire for any of them.
Let me sketch those methods in short: First they introduce the “Goal Question Metric (plus)”, short GQM+. GQM+ is based on GQM, which is a methodology to derive crisp business goals. While I find some of the leading questions worthwhile, the overall concept strikes me as overly complex and hard to grasp.
After this, the authors present the “Experience Factory”. This is essentially an extension of the classic plan-do-study/check-act cycle with additional steps and a “sub-cycle”. Its a semi-interesting read, but doesnt convince me in its current form.
Finally, the concept of “Non-invasive Measurement” is laid out. The goal of this approach is to collect data without distracting the engineers. While such non-invasiveness is indeed desirable, the proposal seems overly complex to me. I mean, there are so many ways of analyzing process flows, code quality, efficiency, etc. Why do the authors describe a database scheme for a concrete solution.
All-in-all the book “Lean Software Development in Action” didn’t convince me. Its best part is where lean and agile are described, and the book offers a few interesting new perspectives on them to an already somewhat informed reader. Those aspects I have mostly covered above. The part where the authors bring in their ideas for methodologies which augment existing known approaches is rather weak, probably because its about academic ideas with little (not none!) exposure to real project life.