Ideation doesn't have to be chaotic

I think learning how to develop interesting ideas is one of the hardest parts of becoming a designer, something I clearly realised during my bachelor's degree, running into a wall whenever I tried figuring out how to come up with more than the obvious solutions. Up until that point, I hadn't fully realised how to turn this into a reliable process.

Fast forward to my master's degree, an intensive one-year program learning anything machine learning and AI, while combining it with design. It was a lot, but I learned more than I could even imagine was possible in 12 months.

Structure the creative process

We all know that creativity can be in chaos, but it turns out structure actually led to better results for me. I spent much of my first semester not just understanding machine learning, but also ideation. There are dozens of methods you could use for this. I'll bring up some examples below and add examples, to show that the starter does not have to be any final result.

My workflow

I won't go into detail about the methods since there's hundreds if not thousands articles out there, but I'll highlight the process I ended up using for my master's thesis. I used the double diamond method, diverging and converging in two stages across the process, which led to a total of 136 solution areas. These were not definitive, these were just thoughts.

As you might notice in this graphic, there's an emphasis on values, scenarios and assumptions at the end. I notice that later throughout a process, I start focusing more on tailoring these ideas to the specific context of the project. Specifically, things valued by your target audience can be defined by just...talking to them!

A highlight

SCAMPER is one of my favourite ideation methods because it allows me to dump ALL my thoughts into a table, which scales really quickly if you run through all potential solution areas. In one of my projects, the SCAMPER table was driven by 'provocations', which are generally ridiculous solutions that reach far beyond just resolving the initial problem. An example could be, blocking users from entering their bank account if their data isn't correct. Super annoying to deal with as a user, but technically a very efficient way to ensure data quality. Obviously, not an idea you'd ship.

Filtering outcomes

Obviously, a gigantic grid of solutions is something you can't ship, nor something you can present to stakeholders. Using 4 categories, "Most Rational", "Most Delightful", "Darling", and "Long Shot". The long shots are solutions that aren't practical, too intrusive, or unrealistic at this moment, so those are filtered out. I ended up filtering out "Most Rational" as well, removing the obvious and easy solutions. Instead, the delightful and darling solutions are the ones that are most likely to be shipped and could have the most potential.

From there I've created a values matrix, tying in research and judging each potential solution based on values determined in the process with our stakeholders. This helped select 5 directions, which were ready for the prototyping stage. There are dozens of other methods to filter and select outcomes, I think there's no one-size-fits-all solution. Instead, we should look at what fits our existing knowledge and context, so that we can make rational decisions rather than ones fueled by what we personally like.