Graduate from an Amateur Analyst: An Insider’s Perspective
In 2022, Cassie Kozyrkov, Chief Decision Scientist at Google and one of the top voices in data and analysis, wrote a series of articles on how to be a good data analyst.
In my opinion, it’s a must-read for all data analysts or those looking to enter this field. It is also more relevant than ever given the recent development of AI and tools like ChartGPT can easily perform the tasks of entry-level analysts.
Here are a few key things I learned from Cassie’s articles on how to be a professional and valuable analyst.
1. Tools are easy to learn, but adopting the analytics mindset is not.
Cassie talks about what she called “data science bias” — trusting information more when it smells of the data sciences. Adding a fancy chart to the data or building a sophisticated model from it doesn’t make it any more special.
It’s important to keep in mind that data does not inherently possess truth or wisdom. We the analysts should use it as a tool to help understand the business and not treat it as an infallible source of truth.
Data is created by humans with biases and inaccuracies. A good analyst should always maintain a healthy scepticism, carefully vetting and questioning the data — understanding the sources, limitations and assumptions that are built into it.
A good analyst also knows how important domain knowledge is. We can only make meaningful interpretations of the data when we understand the real-world context that generated the data.
Recently I’ve spent months working on a fancy Marketing Mix Model which I was so excited about. Neglecting dataset accuracy during the project led to wasting months of time and effort when senior managers questioned me during the product presentation.
2. Understand that analytics and statistics are very different fields.
While every analyst should have a grasp of fundamental statistics knowledge, the main role of an analyst is to explore the data, to inspire and to ask good questions, within the content of the given dataset, but not beyond.
On the other hand, a statistician’s job is to infer what’s beyond the data.
In short, analysts come up with good questions (hypotheses) to ask while statisticians answer the questions through vigorous testing.
There is an overlap between the two roles and they should work closely together, but they are fundamentally different.
I have to admit that I have never thought about this before, and looking back, I might, quite a few times, act as a “Data Charlatan” in the team — someone that is worse than an amateur, as Cassie argued.
Being a Charlatan is dangerous because we use data to prove what we already believe and make predictions without enough evidence.
Charlatan: a person who pretends to have skills or knowledge that they do not have. — Cambridge Dictionary
3. Knowing how to beat one of the biggest enemies of analysts — confirmation bias.
We all know what it is and we all know (hopefully) that we should resist it. But in practice, it’s not easy.
As a marketing data analyst, analysing the performance of marketing campaigns is a significant part of my job. The stakeholders of these analyses are mostly people who design and manage the campaigns. Naturally, nobody wants to hear that the campaigns ran failed to meet the expectations, so we are inclined to, intentionally or unintentionally, look for evidence from the data to support the hypothesis that the campaign is a success and generates incremental revenue. We believe that showing a “good” result will make our stakeholders happy and give us a good rating.
But if we think about it, if every campaign or product they produce is successful, why do they even need us in the first place? Our job is to help the stakeholders figure out what is working and, more importantly, what is not, so that they can take away the insights and make improvements. That will ultimately benefit the business, the stakeholders themselves and eventually the analysts.
This leads to the last and most important point: how do analysts add value?
Even today, I feel like I am seen as a second-class citizen in the organisation as I don’t do complex statistical analysis or fancy machine learning modelling.
However, as Cassie pointed out, while analysts are not decision-makers (and they shouldn’t be), we inspire them. We explore the data to “hunt for unknown unknowns worth knowing about”. We inspire the stakeholders to think from different angles. In other words, we do proactive analytics.
It’s like Apple. They knew we wanted and needed a better portable music player, so they created the iPod, that’s reactive. But they also created iPad, something we didn’t even know that we needed and inspire the app developers around the world to come up with wonderful applications that utilised the big screen. That’s proactive.
Of course, it’s not easy to be a good and professional data analyst. It might only take a couple days to learn how to write basic SQL and make nice charts in Tableau, but what differentiates a pro analyst from an amateur one would take years of deliberate practice to master.
But your effort won’t be wasted. As Cassie states in her article, analysts are the most misunderstood heroes and we are “the most-likely heir to the throne”.
The insights from Cassie Kozyrkov help me rethink my career as a data analyst and encourage me to continue this path. I hope it helps you too.
Thanks for reading!
Here you can find the original articles from Cassie Kozyrkov and If you would like to read more articles on analytics, data visualisation and AI in the future, consider following me.
