How do I know in advance whether a feature, a product or a project is valuable?
The short answer is: You can’t know. However, it is possible to make more or less good estimates, depending on how much your assumptions correspond to reality. Practical experience in software development shows that decision-makers are often far off the mark. 97% of all large size projects that work according to the waterfall model (analysis before implementation) fail or exceed the budget, time frame or requirements. So if neither analysis nor intuition helps – what does?
Learning by doing
The basic idea of Scrum and evidence-based management is based on the empirical method: Inspect and adapt. It is somewhat reminiscent of the principle of “learning by doing”. In a complex system, value cannot be sufficiently determined in advance. So instead of investing a lot of resources in an analysis and design phase, meaningful hypotheses and assumptions are made and verified as quickly as possible through releases. From the knowledge gained, new hypotheses are formed and implemented – resulting in the fastest possible value maximisation. We like to say: “If you draw twice per round when playing chess, you will beat even a World Chess Champion”.
According to the Lean Startup community, it is not too rare for the original idea to be completely discarded and considerably more efficient ideas to be implemented through a better understanding of the market. The origins of this scientific approach can be found primarily in Lean Manufacturing, which was further developed by Toyota in the 1930s, based on the scientific management approach of Frederick Taylor and Henry Ford. In his book “Lean Startup”, Eric Ries brought together the approaches of Lean Manufacturing with those of modern software development and highlighted the “build, measure and learn” approach as a core concept of product development. With this doer mentality, projects are implemented more successfully.
Dealing with uncertainty
We now live in a complex world in which various factors influence the success of a project or product. Since the end of the Cold War, the acronym “VUCA” has often been used for this. It stands for volatility, uncertainty, complexity and ambiguity. In software development, foresight is usually limited. What happens next? To which event, which changes in legislation, which changes in the market, which new technology do I have to react next? It is like a fog that lies in the valley. To see what is in the valley, you have to go inside.
You can also imagine this uncertainty like the Cone of Uncertainty from meteorology. Due to the complex meteorological interactions, an exact course of a hurricane is currently not calculable. It is possible to calculate probabilities for the path of the cyclone, which are subject to increasing uncertainty over a longer time horizon.
In order to make more accurate predictions, the probability range must be narrowed down. In Lean Startup, this is achieved by reducing the batch size. So how much work is needed to implement a task? The smaller the tasks are cut, the faster they are completed. This theory is supported by Little’s Theorem, which was mathematically proven in 1961. To put it simply, it says
Work in progress = throughput * lead time (on average)
As we want to validate our hypotheses as quickly as possible, we work on as few elements as possible at the same time. It has also been shown that parallelism often leads to hidden wasted time. “I don’t really get to work until the evening” – a sentence that is heard more often. A typical sign of too little focus, triggered by parallelism or too many areas of responsibility.
Both Lean Startup and evidence-based management show that in a complex world, value can be created faster if you learn faster. Often within a few weeks and with a small budget, assumptions can be validated by first releases, minimising project risks. The value-added curve increases more in product development than with traditional approaches such as the waterfall model.