What are some of the common pitfalls and biases that can affect your data interpretation and storytelling?

Data quality issues can reflect the sophistication of an analytics organization. Regardless of your org's sophistication, there's a key aspect: your analysis must link back to finance. In doing so, various reports and metrics can aid in providing a broader context. Your analysis doesn't need to mirror financial figures precisely. It should fall within the 'ballpark', though. This fosters trust and coherence in the interpretation. Whether analyzing customer behavior, market trends, or efficiency, results should relate to financial impacts like cost savings, revenue increases, or better investment decisions. By linking analysis to finance, you offer insights that guide decisions, enhancing the relevancy and impact of your data story.

With all the cleaning in the world, it’s probably safe to assume that no real-world business has completely reliable, valid and consistent data. In fact, most businesses combine unreliable data with unreliable narrators, but act like the opposite is true. With that in mind, a more realistic approach to analysis (for the majority of businesses that aren’t sending probes to deep space) is as a hypothesis for further validation and measurement, versus the definitive answer to every decision. Approaching data as something to be informed by, amongst other factors, makes it more useful in application, and over time improves the reliability of this data as a decision-making tool.

In smaller companies, the collection, cleaning, processing, and analysis of data is usually completed by the same person or team. However, in larger companies, these different parts of the process can be handled by entirely different teams. When analyzing data always run the trends and insights your find in the data by the people who collected and cleaned the data. Their additional context could help you avoid a misinterpretation of the data.

Confirmation bias is a common pitfall that can affect data interpretation and decision-making. Remaining open-minded and actively seeking alternative explanations and diverse perspectives can help mitigate its impact. In my experience, consciously challenging my existing beliefs has led to more objective and well-rounded decision-making.

The easiest way to fall into the confirmation bias trap is to use a top down approach. By that I mean starting with the larger industry trends and then analyzing the data through that lens. Taking a bottom up approach by starting with your own data and identifying key trends and insights allows you to come to the data with a clear mind. Then, you can look at the wider industry trends and see how your company is either fitting into them or going against them, as well as what the outcome of that is.

This is where the saying “correlation does not imply causation” comes into play. Be careful with how you word your insights. Makes sure your observations are measurable and consistent. Never speculate beyond your data without some kind of third party research to back it up.

1 ice-cream yesterday 2 today Equals 100% increase 0 isn't really a number, similar to infinity which also isn't a number. 0 is the absence of numbers, no-number: infinity is the fullness of numbers which can be added to. You can create strange problems with both: how many odd numbers are there? How many even numbers? How many numbers overall? Does all numbers / even numbers = 1/2 No, that is to misunderstand infinity. Divide any number by 0, the answer is always 0. How about 0/0? Either 0 or a no-number (=0).

An example of this might be if I said that I had 100% more ice cream sundaes than yesterday. On the surface, that seems like a lot. You may be concerned for my health. But if I had zero ice cream sundaes yesterday, then that means I only had one ice cream sundae today. The way a number is represented can have a big impact on how we perceive the situation. Always try and transform your numbers to understand their true value. For instance, in our ice cream sundae example, transforming the 100% into the integer 1 revealed the truth of the situation.