Peeking into a Data Analyst’s Mind
How to handle anxieties beyond numbers and statistics
Last week, a friend of mine invited me to share my experience as a data analyst to his team in eFishery, an Indonesian aquaculture start-up. During the session, I talked about becoming a business-minded data analyst. In summary, I introduce an alternative perspective where we have to put ourselves in the business user’s shoes. Since the content is similar to my post back in 2017, I am not going to reiterate the idea again.
What I want to discuss is what happened by the end. As I try to understand the current condition, I decided to survey the audience on this scenario:
Imagine that you have just finished an analytics project. Here comes the time when you have to share it to the business team. You are standing in front of the meeting room door. What worries cross your mind?
(Feel free to pause and think about your answer before continuing)
Surprisingly, the answers still resonate with my experience from a couple of years ago. It seems the situation has not improved much. In the following paragraphs, I offer my methods on handling those worries. I gathered twelve responses in total, which I grouped into three categories:
1. Lack of confidence
Answers:
“I think the data is wrong”
“I am not confident with my own opinion on the results”
“Is this analysis correct? What if I am wrong?”
“Have I used the correct business context?”
It is impossible not to have any doubt when you are about to present the results. In fact, those doubts are necessary. Based on those doubts, we can do our best to minimize the issue. The mistake occurs when we become too obsessed or choose to do nothing about it. If we are not confident with our own analysis, why should the users?
Any data analysis, prima facie, is never 100% flawless: incomplete data and transformations involved. That is why, on all occasions, we should remind the users on the downside of our analysis. Think about the limitations and potential biases. If we believe the collected data or the method is inaccurate, explain why do we think it is still acceptable to use them. The ideal approach would be to quantify our doubt, for instance, using the margin of error. Nevertheless, not all important things are quantifiable.
I understand that sometimes we are not aware of our mistake until it is too late. Based on my experience, the best resolution is to come clean and do not forget to reflect, so it will less likely to happen in the future. Trust the users to consider the plus and minus of our analysis. Do not try to hide it. They will notice.
2. Lack of understanding
Answers:
“The users do not understand what is the important takeaways from the analysis.”
“The users have initial assumption when I have not done explaining the results. The discussion follows the assumption, but that was not what I intended to bring up.”
“The users do not actually understand what I am explaining.”
“The method that the data team used is too complex for the users.”
Recently, data storytelling emerges as one of the key skills that a data analyst need. I can imagine why it has taken a while until people realize the importance of this skill. No matter how mind-blowing your result is, if you do not present it well, the effect shrinks. Funnily, it also works the other way around: if you can explain it well, even small insight can initiate big action.
There are plenty of resources you can find on the internet to learn about storytelling, so I will not pour salt into the ocean. I just want to state one thing: if the users do not understand the results, it is our fault. When a method is too complex to describe, we do not understand it well enough. Just use a simpler approach. The analysis sounds complicated? Break it down and make it is easier to chew. It is part of our responsibility.
Data analysis is like a journey. You started with a destination in mind, but often encounter interesting stuff that you want to carry with you. Based on my experience, it is better to be ruthless on our path. Focus on what is relevant, and you will have an easier time explaining it to the users. Tell them about your journey too.
3. Lack of impact
Answers:
“The conclusion is dull, just something good to know.”
“I am afraid that the analysis is not impactful despite the hard work that I have invested in it. The users respond ‘that is interesting’ but no follow-up.”
“No feedback, just agreement.”
“Under/over-expectation from the business team. It is painful when we found something interesting but the response is just ‘okay’. HOWEVER, it is also dangerous if we oversell the results.”
I have seen many data analysts work in isolation until they feel like they have found something precious. They expect to have a eureka moment and then share it. The problem is, no one can guarantee you will encounter such a moment. When you realize you are going nowhere, you have invested too much time and efforts. Worst case, we are tempted to sugarcoat our analysis to make it sound worthy.
Most of the time, my initial findings are next to worthless, and I have learned to accept that. My approach is to include the users early in the process. Ask for their inputs, keep in touch. The regular updates should help to manage their expectation. Additionally, when both parties sense that the project does not seem to work out, just move on to the next inquiry. Do not forget to write what you have learned, no matter how trivial it seems. It can be useful in the future.
Easier said than done
Of course, merely stating those suggestions is easy. I know the difficult part comes when real pressure exists. I am guilty of those worries too, even until now. However, seeing how common they still are today concerns me. It is time to start discussing issues beyond numbers and statistics. What are your worries? Let’s discuss and learn from each other.
Special thanks to the eFishery data team for having me and allowing me to share a screenshot of the session!
