Writers like to talk about the tyranny of the blank page. Despite the volume of ideas they might have and dictionaries of words at their disposal, putting pen to paper can be insurmountably challenging. Analytics can engender a similar sentiment in businesses. There are all these data, all these tools, all these machine learning frameworks to choose from. How to know where to begin?

As I have covered, structure only gets you so far: you need to have a goal defined. Given analytics is the art of making better decisions with data, the easiest way to start is to ask yourself this: What questions keeps you up at night? What questions, if you had the answers, would allow you to take confident action? These do not have to be technical. The best questions are those that are understood across the company. What do our customers have in common? Is it worth opening a new factory? What food should we serve in our cafe?

Not all questions are created equal. Photo by Bob Jagendorf.

Next, get specific. For example, if you are considering optimising a cafe menu, what are you optimising for? Do you want to make customers as happy as possible by serving their favourite food, or are you looking to save costs? Do you want to reduce your carbon footprint during this process? All of these are valid questions and what matters is not what you choose, but that you choose. Many analytics projects fail to launch because of a lack of definition up front.

Once you’ve settled on a question, it helps to lead with a hypothesis. What do you think is going on? In its simplest form, what statement are you looking to prove?* This will help you understand what data you need to analyse in order to answer this question. Back to our example, if your hypothesis is that dumplings make your customers happiest, you ideally have some data on when you sold dumplings, as well as data on the happiness of your patrons.

Just because a page has lines does not make it any less intimidating. Photo by Mary Cullen.

Now you have a worthwhile, well-defined question, and you have some idea of how to approach it. This is where you can start working your way up the analytics value chain, which is really just a robust way of getting to an answer. What this looks like will depend on the resources at your disposal and can take any amount of time depending on the complexity of your question. Most importantly though, you have made a start — the first page has been written.

— Ryan

* Or inversely, what null hypothesis are you looking to disprove?