Global Head of Analytics at JOKR.

In my latest experience, beginning the data environment from zero in a startup, I understood the definition of good analytics was shifting over time and from person to person.  For an ultra-fast company that changes and grows from month to month, setting the precise definition and expectations of the analytics is critical.

We tend to get lost quickly in trend topics like big data analytics, AI, and data literacy. When trying to produce results. Ending up with complex solutions that take more work to explain and maintain. So I strongly recommend a regular "step back" on the work and ask yourself the pillars below:

What does it mean to deliver good analytics? It seems a trivial question but one that we rarely do ourselves. Directing our work with this clear in our heads may save time and drive significant results.

Good analytics is composed of 3 main pillars: the best possible information for the right people at the right time.

The best possible information is a delicate balance between precision, data availability, and media (the tool you choose to display the result). Assessing the gains of precision, and applying complex models should be the first thing we do. Most of the time, our internal customers will be satisfied with more straightforward and uncomplicated ways to explain models than with advanced results.

Over complex analysis may, and almost always, produce marginal gains. Mining new data to expand the precision of your work can be time, and money, consuming also. Most of the time, a good enough answer with the data available will suffice for your internal customer. More than that it can buy you time, and confidence, to develop more deep analytics in a second phase. The media you choose to pass it also will impact the analytic capabilities. It is a dashboard? It is a table? Or it is an e-mail? All media can deliver a message, but not all messages can be delivered by any media. Choose your media fitting the messages, and always try to simplify the results and the analysis explanations. In a Twitter era of everything fitting 150 words is better to tell a simple story on how the results are built.

 The best possible information is a delicate balance between precision, data availability, and the media you choose to display the result. 

The right time can mean different things if you can set your internal customer expectations. Always be clear on what is possible to deliver and when. Split your deliverables to create a constant flow of deliveries. We have all heard the agile concept but we hardly apply its basic concept. Better to deliver every week and evolve than have a faraway deadline.

There will be a time for advanced analysis on your path, but it will be sparse and only effective if you first meet the pillars above. With the literacy necessary, the good enough data, and the correct expectation, your internal customer will be able to fully appreciate the scope of an AI or a natural language model.