The romanticised version of a data analyst is usually a montage where a skinny, nerdy stereotype (with wireframe glasses for added effect) is hunched over a computer screen scouring through rows and rows of a spreadsheet before ultimately hitting that ‘Eureka!’ moment of identifying a game-changing trend. Unfortunately, the reality of generating an actionable insight is an inglorious proposition: it is an iterative process of elimination via trial and error.
The advent of big data and machine learning has accelerated the process of building data-driven strategies, but the sheer volume of numbers and endless analysis can be overwhelming! W.E.B. Du Bois, a sociologist, said, “When you have mastered numbers, you will in fact no longer be reading numbers any more than you read words when reading books. You will be reading meanings.” Here, I will attempt to create a framework to help you read meanings.
1 Envision an ‘Insight Loop’
It is easy to assume that the starting point for any insight is data, but it has to be layered with context to make it meaningful. For instance, falling GDP numbers will push economists to gauge what is driving the drop. On the other hand, sometimes an observation or a behaviour triggers a hypothesis that is later validated by data – ask Isaac Newton, whose observation of a falling apple triggered the insight more commonly known as gravity. The purpose behind the insight loop is to help generate a hypothesis or ideas that one can test. There is no set ‘starting point’ when it comes to insights; think of them more as loops where quantitative measures (data) and qualitative findings (hypotheses or observations) trigger each other. This is why successful pitches will always have a relatable story that makes one resonate with an issue – the qualitative – as well as statistics to help you appreciate the scope – the quantitative.
2 Specific Questions Get Specific Answers
As much as this may be a throwback to your third grade teacher, in analytics, there really are no stupid questions; just wrong ways to answer them. A generic question will get you generic answers that are not very actionable, but are important in helping to narrow down your hypothesis. This then allows one to ask the specific questions that help you attain particular data points that are stitched into an actionable insight. Baselines can often be obtained with simple Google searches or third party reports or even your own internal data. A digital marketer can easily identify the most visited websites via tools like Alexa to gauge which platforms should be shortlisted. The next step would be to qualify which websites have the highest probability of engaging the target audience and then quantifying them via market research. Similarly, it is important to know exactly what a KPI is measuring and whether it is helpful in answering your business questions – marketers often incorrectly assume that unaided awareness is solely a function of communication when in fact usage plays a critical role as well.
3 Leverage User Journeys to Identify Gaps
Product stakeholders in Pakistan often wonder why there is poor uptake of their product/service despite performing well in concept tests. The answer is because we fail to sufficiently empathise with the end user. A study by Acquia shows that while almost nine out of 10 marketers feel confident matching user expectations, only 50% of customers feel the same way. Similarly, a survey conducted in a hospital in the US found that when asked if most healthcare professionals provide compassionate care, 78% of the doctors said yes, while only 54% of patients replied in the affirmative. I often recommend that stakeholders map out potential gaps within the customer journey as this will allow them to gauge which issue needs to be addressed first. An app that predicts weather patterns for farmers sounds great, but it will not succeed if they only trust information from fellow farmers. Therefore, it is crucial to determine what channels of information are trustworthy with farmers prior to designing the product.
4 Smart Data Aggregation
Data is like an onion, you have to peel away the layers – and shed a few tears in this process – to get to the insights. Combining data points from different sources can help reinforce a narrative by validating it across a broader spectrum. Similarly, adding learnings across markets and/or industries is beneficial. I often search for reports aimed at South Asian and African countries as they tend to have similar market dynamics. Data aggregation is also useful, since all statistical metrics have limitations – there is no one silver bullet that gives all the answers. Therefore, analyse data from the lens of other demographic and behavioural variables to add more context to your findings. To avoid overwhelming an audience, I often add call outs to the overall results to highlight only the significant differences in demographics. Finally, always be in command of what the metric actually means so that your interpretations are accurate.
5 A Continuous Process
Any Netflix user can vouch for their seamless streaming experience. That is because every change made to the Netflix platform goes through a rigorous testing process before implementation. The homepage continuously evolves based on each user’s profile as their streaming history and preferences determine the number of rows displayed on the homepage and the suggested content within each category. It also helps to focus on trends and not the data itself during the analysis stage. The best insight comes from the direction your KPIs move compared to time ranges, such as month-on-month or quarter-on quarter. Like Netflix, it is critical to create a measurement system that provides a continuous feedback loop to help you keep improving your product/service – for each new feature added, there are thousands of tests and analysis behind it. n
Ans Khurram is an insights professional working in the telecommunication industry in Pakistan. firstname.lastname@example.org