Knowledge

Many junior developers tend to think that software development is just typing code, and when they become more experienced in typing, get to know more IDE shortcuts, and know frameworks and libraries by heart, they will be ninja developers, able to write something like Instagram in a couple of days.

Well, the reality is harshly different. In fact, after getting some experience and after deliberately spending months and maybe years in death-marches towards impossible deadlines and unrealistic goals, people usually slow down. They begin to understand that writing code immediately after receiving a specification might not be the best idea. The reason for this might already be apparent to you if you have read all the previous sections. Being obsessed with solutions instead of understanding the problem, ignoring essential complexity and conforming to biases—all these factors influence us when we are developing software. As soon as we get more experience and learn from our own mistakes and, preferably, from the errors of others, we will realize that the most crucial part of writing useful, valuable software is the knowledge about the problem space, for which we are building a solution.

Domain knowledge

Not all knowledge is equally useful when building a software system. Knowing about writing Java code in the financial domain might not be very beneficial when you start creating an iOS app for real-estate management. Of course, principles such as clean code, DRY, and so on are helpful no matter what programming language you use. But knowledge of one domain might be vastly different from what you need for another domain.

That is where we encounter the concept of domain knowledge. Domain knowledge is knowledge about the domain in which you are going to operate with your software. If you are building a trading system, your domain is financial trading, and you need to gain some knowledge about trading to understand what your users are talking about and what they want.

This all comes to getting into the problem space. If you are not able to at least understand the terminology of the problem space, it would be hard (if not impossible) to even speak to your future users. If you lack domain knowledge, the only source of information for you would be the specification. When you do have at least some domain knowledge, conversations with your users become more fruitful since you can understand what they are talking about. One of the consequences of this is building trust between the customer and the developer. Such trust is hard to overestimate. A trusted person gets more insight and mistakes are forgiven more easily. By speaking the domain language to domain experts (your users and customers), you also gain credibility, and they see you and your colleagues as more competent people.

Obtaining domain knowledge is not an easy task. People specialize in their domains for years, they become experts in their domains, and they do this kind of work for a living. Software developers and business analysts do something else, and that particular problem domain might be little known or completely unknown when they need to obtain domain knowledge.

The art of obtaining domain knowledge is through effective collaboration. Domain experts are the source of ultimate truth (at least, we want to treat them like this). However, they might not be. Some organizations have fragmented knowledge; some might just be wrong. Knowledge crunching in such environments is even harder, but there might be bits and pieces of information waiting to be found at the desks of some low-level clerks, and your task is to see it.

The general advice here is to talk to many different people from inside the domain, from the management of the whole organization, and from adjacent domains. There are several ways to obtain domain knowledge, and here are some of them:

• Conversations are the most popular method, formalized as meetings. However, conversations often turn into a mess without any visible outcome. Still, some value is there, but you need to listen carefully and ask many questions to get valuable information.
• Observation is a very powerful technique. Software people need to fight their introversion, leave the ivory tower and go to a trading floor, to a warehouse, to a hotel, to a place where business runs, and then talk to people and see how they work. Jeff Patton gave many good examples in his talk at the DDD Exchange 2017 (https://skillsmatter.com/skillscasts/10127-empathy-driven-design).
• Domain Storytelling, a technique proposed by Stefan Hofer and his colleagues from Hamburg University (http://domainstorytelling.org/), advocates using pictograms, arrows, and a little bit of text, plus numbering actions sequentially, to describe different interactions inside the domain. This technique is easy to use, and typically there is not much to explain to people participating in such a workshop before they start using it to deliver the knowledge.
• EventStorming was invented by Alberto Brandolini. He explains the method in his book Introducing EventStorming (2017, Leanpub), and we will also go into more detail later in this book when we start analyzing our sample domain. EventStorming uses post-it notes and a paper roll to model all kinds of activities in a straightforward fashion. Workshop participants write facts from the past (events) on post-its and put them on the wall, trying to make a timeline. It allows the discovery of activities, workflows, business processes, and more. Very often, it also uncovers ambiguities, assumptions, implicit terminology, confusion, and sometimes conflicts and anger. In short—everything that the domain knowledge consists of.

Avoiding ignorance

Back in 2000, Philip Armour published an article called Five Orders of Ignorance (Communications of the ACM, Volume 43 Issue 10, Oct. 2000), with the subtitle Viewing software development as knowledge acquisition and ignorance reduction. This message very much correlates with Alberto's quote from the previous section, although it is somewhat less catchy but by no means less powerful. The article argues that increasing domain knowledge and decreasing ignorance are two keys to creating software that delivers value.

The article concentrates on ignorance and identifies five levels of it:

• The zero ignorance level, which authors call the lack of ignorance, is the lowest. On this level, you have no ignorance because you have most of the knowledge and know what to do and how to do it.
• The first level is the lack of knowledge. It is when you don't know something, but you realize and accept this fact. You want to get more knowledge and decrease your ignorance to level zero, so you have channels to obtain the knowledge.
• The second level also called the lack of awareness, is when you don't know that you don't know. Most commonly, this occurs when you get a specification that describes a solution without specifying which problem this solution is trying to solve. This level can also be observed when people pretend to have competence they do not possess, and at the same time are ignorant of it. Such people might be lacking both business and technical knowledge. A lot of wrong decisions are made at this level of ignorance.
• The third level is the lack of process. On this level, you don't even know how to find out about your lack of awareness. Literally, you don't have a way to figure out you don't know that you don't know, which sounds like inception, but that's exactly what it is. It is tough to do anything on this level since apparently there is no way to access end users, even to ask if you understand their problem or not, in order to get down to level two. Essentially, with the lack of process, it is nearly impossible to find out if the problem you're trying to solve even exists. Building a system might be the only choice in this case, since it will be the only way to get any feedback.
• The fourth and last level of ignorance is meta-ignorance. It is when you don't know about the five degrees of ignorance.

As you can see, ignorance is the opposite of knowledge. The only way to decrease ignorance is to increase understanding. A high level of ignorance, conscious or subconscious, leads to a lack of knowledge and a misinterpretation of the problem, and therefore increases the chance of building the wrong solution:

Ignorance is highest at the earliest stages

Eric Evans, the father of DDD, describes the upfront design as locking in our ignorance. The issue with the upfront design is that we do it at the beginning of a project, and this is when we have the least knowledge and the most ignorance. It has become the norm to make most of the important decisions about the design and architecture of the software at the very beginning of a project when there is virtually nothing to base such decisions on. This practice is quite obviously not optimal.

In the article Introducing Deliberate Discovery (https://dannorth.net/2010/08/30/introducing-deliberate-discovery/), Dan North suggests that we realize our position of being on at least the second level of ignorance when we start any project. In particular, the following three risks need to be taken into account:

• A few unpredictable bad things will happen during the project.
• Being unpredictable, these things are unknown in advance.
• Being bad, these things will negatively impact the project.

To mitigate these risks, Dan recommends using INTRODUCING DELIBERATE DISCOVERY, that is, seeking knowledge from the start. Since not all knowledge is equally important, we need to try to identify those sensitive areas where ignorance is creating the most impediments. By raising knowledge levels in these areas, we enable progress. At the same time, we need to keep an eye on new troublesome areas and resolve them too; and this process is continuous and iterative.