Skip to main content

Storytelling with Data

Premium

Okay, so you’ve figured out what data you’re trying to investigate and considered all the factors to finally identify a trend worth sharing with others. What next?

Know Your Audience

You'll often be presenting directly to upper management, but as project manager and collaborator across functions, you'll have to adapt. Presenting your recommendations in a way that resonates with your stakeholders is key as getting buy-in is an (often unspoken) requirement of your job.

Before each presentation, ask yourself a few key questions:

  • Who will see this and how much time do they have?
  • Are they interested in the big-picture only, or will they want details?
  • How much context do they have?
  • What will this audience take away from my presentation?
  • What are the direct outcomes?

Presenting Data

Great data-driven presentations tell a story, as humans are wired to learn through storytelling. Good stories have a beginning, a middle, and an end. Typically, your presentations will begin with the present (here's where we are), lay out the path that brought us here (likely a set of problems, if bizops is on it!) and end with a proposed solution and predictions for a better future. You can use data to reinforce any point in your story.

Note: Stay tuned for a specific lesson on presentation best practices; for now, let's keep our focus on the data.

The key to storytelling with data is to show the right point of comparison to make your point. For example, if you want to show that revenue is growing faster this quarter than last quarter, show year-over-year on a quarterly basis. Finding the right baseline for your comparison can be tricky especially because every business has its own unique set of characteristics.

You can use this nuance to your advantage in an interview scenario, as an interviewer won’t expect you to know all the nuances, so highlighting these things (out loud!) will win you brownie points.

Common ways of comparing data include:

  • Trends over time (week-over-week, year-over-year)
  • Comparing against a baseline (how does one product perform against another?)
  • Comparing deltas (eg. how much more did one product increase than another?)

Data Visualization

When considering how to share the data, data visualization is your friend. You should be super comfortable using the following forms of visualization:

Line Charts

  • When to use: Presenting continuous data, especially with many points.
  • Examples: DAU, identifying trends, comparing patterns, etc.
  • Notes: Line charts are typically not useful unless there’s a point of comparison on the same chart; choose a baseline that suits your data. Because too many lines can look messy, best practice dictates you don't add more than four lines on a single chart.

Here's a line chart showing search volume for the term "fidget spinners" from 2004 - present on Google -- an example of trend identification.

line_chart_fidget_spinners

Bar Charts

  • When to use: Comparing categories (or comparing to a baseline, if you have few points to graph.) Column charts work well when there are few categories to compare, but horizontal bar charts work better with large datasets. Stacked bar charts add another layer -- comparing how similar sub-groups contributed to totals.
  • Examples: Relative performance across marketing channels, budgeted vs. actual expenses per month, identifying over and under-performing products.
  • Notes: Bar charts are a good choice when comparing many categories, but it's still a good idea to highlight categories you want to draw attention to.

Here's a stacked (columnar) bar chart plotting Facebook MAUs over time.

facebook_meta_mau

Scatter Plots

  • When to use: Showing relationships between data across two variables.
  • Examples: Understanding relationships between delivery time and CSAT, conversion rates across ads with different CPC, etc.
  • Notes: Understanding how variables are related provides a great basis for you to start developing hypotheses about the business drivers, while not requiring the rigor of a causal analysis. Traditional scatter plots usually include a trendline; if plotting non-numerical data, try a heatmap or bubble chart (below.)

Here's a scatterplot suggesting a positive correlation between (call center) agent satisfaction and customer satisfaction.

scatterplot_satisfaction

Heatmaps and Bubble Charts

  • When to use: Scatter plot alternative; popular for showing non-numerical relationships or for exploring relationships between three variables.
  • Example: Product usage per geographic area (non-numeric), correlation between a product's rating and price with bubble size corresponding to quantity ordered (3 variables).
  • Notes: Color choice is important when working with heatmaps. Stick to a single color palette, and use darker colors to indicate higher densities / stronger relationships. Use a legend if colors map to numeric values.

Here's an example of a complicated story (enterprise cloud adoption, McKinsey 2018) conveyed cleanly using a bubble chart.

mckinsey_bubble

Recap

Storytelling with data is an art and a practicable skill. You may be asked to present a take-home, in which case you'll actually make some of these charts, but if not, be sure you know which visualizations work best for answering common business questions.

To demonstrate storytelling with data:

  • Practice triangulating common business problems with the right data visualization for the audience at hand.
  • Choosing the right baseline for comparison is key -- always pick something that makes sense for the specific prompt you're given.
  • Get comfortable working with line charts, bar charts, scatter plots, and heatmaps (at minimum.)
  • When in doubt, choose simplicity. Don't try to capture too many points with one data visualization.
Next Lesson