Slow Thoughts from Dale du Preez

I’ve had a few opportunities to onboard people into some internal data tools, and I find that most people are overwhelmed by the new tool(s) in addition to not being clear about where to start. I find that it’s invaluable to start by describing what you’re looking for in plain, simple language, while trying to capture user actions or attributes as simply as possible.

For me, it’s an excellent starting point because it forces me to think about what I am actually looking for, without regard for the specific data we currently have. This means it can also help to identify data gaps! More broadly, I find that taking a step back and thinking more simply has some side-effects that are very useful.

Finding the right data/events

If I can explain what I want to look at, it becomes easier to identify events or data points that reflect the thing I am looking for. It also helps me to work out whether the data already includes additional attributes that I want to use, or whether I may need to look at things slightly differently.

In some cases, I don’t have an event that captures exactly what I want. When that happens, the simple-language-first approach makes it much easier to identify that I am in this situation, because none of the data points captures what I want. More importantly, it also means that if I need to reach for data that serves as an approximation, a floor, or a ceiling, because I’m missing an attribute, I know that upfront, and can clearly describe why and how I’m using that data. This is critical for clearly communicating my eventual findings, because they need to be qualified, and the qualifications need to be explicitly documented. There are multiple reasons for this:

• Being clear about any complexities helps others to avoid doing unnecessary or repeat work
• Someone smarter than me may come along and point out a way to do this easily! They may see a way to address the gap or problem I worked around.
• You can still uncover critical findings using a ceiling or floor, and they can be more than sufficient to drive big product or business decisions

I often use approximations in early, rough explorations, as it allows me to focus on getting some sort of answer much faster, instead of digging around for much longer to get a more precise answer. For many explorations, the main thing I want to understand is a basic order of magnitude, so I am OK with consciously ignoring some edge cases.

Explaining what I did

Because I started by using simple language to describe what I was doing, I can use that language when I share and describe my findings. In turn, this makes it much easier to explain what the analysis is doing, and what your findings actually mean. It’s worth assuming that most of your audience doesn’t have deep understanding of the data, so laying out the how as well as the findings allows more people to understand and engage with your findings.

Even for simple explorations and analysis, simpler wording makes your analysis much more accessible, which in turn allows you to get participation, feedback, and reactions from a broader audience.

#data-analysis #tips