In the beginning (of any analytic) is a question

There was a question. The question showed a gap. Theoretically answered it would bridge the gap allowing the hero of the story to successfully navigate the terrain, cross the canyon, save the person in danger, slay the dragon, and so on.

But in the meantime, there’s still the question. part of the problem is that sometimes you’re not really sure if it’s the right question. In fact, you’re pretty much 100% sure that it’s not the right question, and you’re stabbing in the dark, hoping not to cut yourself in the process. It always makes a mess.

So quite often the first question is the first of many. It’s a small, delicate question intended to be answered quickly and simply put a stake in the ground, saying to the effect of “Here we are, the lay of the land is thus, north is that away, and the sun is directly overhead.”

Who knows, eventually that question may grow up to be a serious KPI that answers all of the CEO’s deepest fears and leads to that amazing bonus that will let you buy the ultra lightweight bicycle.

More than likely - it’s a simple question. And that’s a really good thing. Keep in mind that your mind only has so much capacity. If the question can’t be boiled down to one or two brief, short, sharp words, then you’ve probably got two or more questions.

For example,

Can you please tell me the sales of the medium sized locations over the past six quarters, compared to the market size, with a rolling average and high/medium/low bars to show context?

… is more than any one person can quote in an elevator and still have an impact.

Let’s break this down into a simpler way of thinking:

• Store
  • Market Size
  • Medium
• Time Frame - Six Quarters
• Measures
  • Sales
  • Rolling Average (over what timeframe?)
  • Max/Min/Median Numbers Stores

Is a single unit of knowledge. How big they are, where they’re located, what kind of customers they serve, are all contextual and typically don’t change much. For the sake of the analysis, we’ll assume that they do not.

 Timeframe 

A couple of things have been hammered out already:

• Grain - quarterly - pretty broad but maybe there’s noise at the monthly level?
• Last Six - we don’t have to go too far back in time

Measures

The rubber is hitting the road here. We want to aggegate sales up to the quarterly level per store. To add a little more window dressing, we want to include the max and min, and throw in the median for seasoning.

Boil the ocean

How far can I compress this question?

Last six quarters of sales by store with market size

That’s five unique concepts squashed into one, but it’s getting better.

What if?

We break out the various components by what they are?

Market size is useful but could easily be done in a separate bar chart including the number of stores.

The max/min/median can be reference lines in the main chart, but that introduces more information that the reader has to process.

Rolling averages are great at reducing noise, but they also can be hard to digest. You have to be up front about it.

Bringing it home

The elevator pitch was given to me in an elevator. By my business client. Using my own data. Badly. The torrent of numbers flew out of him in an exhausting overheated mess, and the worst of it was I knew what he was talking about. I can only imagine what a naive recipient would think.

Keep it tight. Boil it down to a simple question with a clear answer. That answer is going to guide across the ocean, through the woods and to strategic grandmother’s house you.