The Data Storytelling Stack

A stackey tour of using data analysis, engineering, and marketing to grow a business

I’ve updated my job description.

The data storytelling stack identifies and builds adjustments that drive a high ROI on business metrics. Data Storytelling involves collecting and arranging information, theorizing about available evidence, and using theories to create and exercise real options through software engineering and UX.

I’m toying with the stack format as a way for describing the different skills a data storyteller uses without falling into preexisting identity traps like “I’m a marketer” or “I’m an analyst”.

Instrumentation

Data represents the potential energy for all the work done in the data storytelling stack. Collecting that data requires instrumentation, and this is where the stack begins. Services like Mixpanel are the classic instance of instrumentation, but instrumentation can be broadly thought of as anything that reveals new information. Think marketing surveys or customer calls.

Arranging Facts

In raw form though, data isn’t useful. A business analyst might use a lot of SQL to transform data into insights. They might also rope in previous analyses or other sources of processed data. In the data storytelling stack, I call this the “Arranging Facts” layer. This is a closed-loop system. We’re not injecting new information from users or customers.

Theorizing

Theorizing, on the other hand, re-opens the system to pseudodata. Data from the previous layers is data that comes from a primary source. Pseudodata is data that we are making up; it is not real. Lots of hypothesizing happens in the theorizing layer of the data storytelling stack. We’re taking the same dataset that any business unit might start with and thinking about other possible bits of insight we’d need to explain patterns in our dataset.

Create Real Options

We verify if this pseudodata is useful by creating real options. Borrowed from finance, a real option is an asset that we have the option to use, but aren’t obligated to use a fully-formed version of it. This layer uses a mix of software engineering and UX to build a minimum viable experiment. Nothing fancy or scalable needs to be done here, but it should be as lightweight as possible to hold the “option” part of the bargain. Volatility is another necessary constraint if we want to maximize option value. When we’re wrong, we’re wrong. But when we’re right, we’re really right.

Exercise Options

And when we find a valuable option that grows the business, we exercise it. We pay off any debts that we incurred while creating the option and we ship it. This is the kinetic energy end of the stack.