Dr. Mario's 2nd 🧠

Reducing Data Questions Deluge

2 min read

rw-book-cover

Metadata

Highlights

most questions can be reduced to the following “meta questions” which cover 80% or more of business questions. What’s happening with X? Are things working well?

  1. If yes, how can I improve it? What are the drivers?
  2. If no, what happened? Why did it happen? How can we fix it? (View Highlight)

The beauty of these questions is their fractal nature. They apply the same way on multiple levels of abstraction in the business. Whether you’re talking about a single metric, a business process, a business function (e.g. sales) or the entire business, you can keep asking the same exact questions. (View Highlight)

New highlights added 2024-06-15

Next you need to standardize ways of assessing metric performance. Are you comparing against a target? Are you comparing to a previous time period? Are you looking at trends over time? Are you looking at changes in variability? Amazon for example uses trailing 6 week and trailing 12 month (View Highlight)

With that out of the way let’s talk about self-service analytics that works. The ideal tool accomplishes very specific goals:

  1. It shows which key metrics are doing well vs which are not
  2. It shows what the key drivers of the metrics are helping you debug the business
  3. It focuses your attention on what matters instead of scattering it (View Highlight)

Of course this means you must have a metric tree already set up with all the drivers to all the key metrics already identified. This kind of work can be done upfront. All you need afterwards is to maintain accuracy and find new metrics/drivers which is essentially research work. (View Highlight)

you want a quick heuristic about where to begin, start with sales. Take the company’s P&L statement and turn it into a metrics tree. Enrich with any other metrics you have available and start looking for causal relationships and input metrics. (View Highlight)

Graph View