9. The Importance—and Limitations—of Storytelling

For in much wisdom is much grief: and he that increaseth knowledge increaseth sorrow.

—Ecclesiastes 1:18¹

It’s almost trite to suggest that any good presentation of analysis needs to tell a story, but of course it’s true. We’re wired for stories. Our primitive brains evolved to detect patterns so we could perceive anomalies that might mean danger. This didn’t just mean distinguishing the tips of a lion’s ears from a field of tall brown grass—it also meant story arcs with causes and effects. Our experiences create expectations, and stories have historically been powerful means for how we pass along our knowledge about these expectations. And so, effective storytelling has been carefully studied and refined since classical times (and likely longer). If this seems a bit musty a reference for a modern business book, have a peek at how modern practitioners are applying Aristotle’s ethos, pathos, and logos modes of rhetoric.²

1. King James Bible, Cambridge edition.

2. Kate Verrill, “Ethos, Pathos, Logos,” Gamestorming, April 5, 2011, http://www.gogamestorm.com/?p=634.

Translated into the modern business world, where on a daily basis we need not only to communicate, but to persuade to action in a data-rich (and thus potentially confusing) context, well-told stories are not only guides but also powerful noise filters. And here is the source of tension for today’s analyst. You appreciate nuance. You’ve read Nassim Taleb’s The Black Swan³ and Daniel Kahneman’s Thinking, Fast And Slow (which we discuss in the next chapter). So Taleb’s concept of the “narrative fallacy” and Kahneman’s cautions about our inherent biases for seeing strong causes where they are weak or non-existent, are front and center for your thinking. And so, you proceed logically and methodically, up from the data, clearly describing what variation you can and can’t explain in your models, always cautioning: “Remember, correlation’s not causality.”

3. Nassim Taleb, The Black Swan, (Random House, 2007).

Your audience, on the other hand, doesn’t like nuance. Even in a high-functioning organization, they want you to “boil it down,” to give them the “bottom line.” (In less well-functioning organizations, they may already have the answers they want in mind, but that’s a different problem.) At least in my experience, this will tend to be even more strongly the case in sales organizations. Senior sales executives have alpha personalities that tend to impatience and distractibility inside the walls of the firm, perhaps as a relief from the careful listening and relationship cultivation skills they have had to cultivate to be effective in selling themselves and managing others who do. On the marketing side, many executives will have come from a communications background, and so even if they are more patient about getting to the punch line, their focus may often be on how well the story is told rather than necessarily on its substance. The modern “analytic marketer” is still more the exception than the rule. In any event, getting your message through isn’t just about your signal’s accuracy but about its strength as well—and packaging is a signal booster.

You have to strike a balance between a core story signal conveying cause and effect to predict outcomes and inform recommendations, and conveying enough about the noise—the distributions around your “mean” message—that you have responsibly framed the story. Probably the most common way of doing this is to select and assign probabilities to scenarios for different predictions you make based on your analysis, which then support your recommendations (if you’re going as far as making them). Gartner, the technology analyst firm, has probably done one of the most memorable jobs of branding its use of scenarios; for example, it explicitly speaks of “p=0.7” in describing the dominant scenario in its forecasts.

But don’t stop at scenarios. The next step should be to evaluate the stakes associated with a decision. This includes the magnitude of costs and benefits at hand. It also covers the “optionality” associated with pursuing them. Optionality means whether and how you can split up your implementation to gauge whether or not outcomes are tracking predictions. For example, digital advertising via banner ads has much higher optionality than a television ad campaign. The cost of production and media is lower per-unit, so you have the flexibility to test and adjust creative executions, and the ads can be targeted much more closely—than with broadcast television—in terms of who sees them and when. So if the decision at hand is whether to proceed with a television campaign based on predictions about sales it might generate, if the predicted scenarios that would green light the campaign only sum to a 70% likelihood, and your audience can accept no less than 85% certainty, this is worth calling out explicitly. Also, you can use this “stakes-framing” proactively to scope your analysis. By asking decision makers about the necessary burden of proof or confidence level in advance of your work, you can plan schedules and resources more effectively for your investigation.

Summing up, here is one simple outline for an effective sales and marketing analytic presentation that balances the need for a core story with the need for nuance and context:

• “Thank you for your time today...”

• “We’re here today because you/we have noticed metric x has declined by y and we need to improve it to z...”

• (This will be more powerful if it’s framed in the context of a broader analytic map, like the one we introduced in Chapter 7, “Practical Analytics—Knowing When to Say When.”)

• “This analysis considers the following possible levers and options within the scope of what we could use, in the next n months, to close that performance gap...”

• (Many presentations are critiqued as sounding academic and forcing the executives consuming the analysis to make this linkage. If you do it for them, you get two benefits: you ease their cognitive workload and you also confirm the object of your analysis. For example, you might ask, before proceeding from this section, “I’d just like to check—does this scope still apply? Are there any changes, drops, or additions we should factor in as we talk today and afterward?”)

• “So, here are the questions we would need to answer, and the outcomes we’d have to predict, to decide what to do...”

• (Leave plenty of room here. Ask, “These seemed to us to balance being focused with being comprehensive—any comments on these, or other things you’d like us to address?”)

• “We’ll start with a base-case story about who our target customer is and how they behave that is foundational to understanding his/her reaction to options for sales and marketing approaches...”

• (Here again is where the Venus-oriented customer experience approach is a good place to start. As you talk about how the target customer engages with and is influenced by your marketing and sales touch points, it also introduces the opportunity to a) note places where you may be over- or under-invested, and b) where attribution analysis might suggest some indirect effects between one channel and another—“Person X sees TV ad, and searches for us via Google on his or her iPad while watching the show.”)

• “We’ll flesh that story out by describing the data underneath it, and some information about our general logic and specific models based on that data—in particular how much variation in those behaviors we can explain...”

• (Remember here that a picture is worth a thousand words. Use graphs aggressively. You might even go so far as to use visualization tools like Tableau or Spotfire to bring your data to life. These in particular make it easier to illustrate how your data is distributed when you dimensionalize it through important customer or product attributes.)

• “Then we’ll re-frame our basic story as a principal scenario, along with a few others that also have meaningful probabilities...”

• (It’s generally best to have a main case (60-70% probability), a better case (15-20%), and a less good case (15-20%) in mind, linked explicitly to the few main variables that influence them, plus maybe a fourth case (0-10%) that addresses what happens if some harder-to-predict variables break one way or another.)

• “Finally, we’ll examine the stakes and optionality associated with our options...”

• (You might examine the minimum sums and lead time requirements of the options you outlined at the start.)

• “With that, we’ll come to a recommendation for what we think you can do now, and what we think we should examine further...”

• (Bring it home! If you are uncomfortable pushing to a specific recommendation, try suggesting some tests. The more you can link your work to near-term actions that stem from it, the more credible you and your team will be.)

Some discussion questions:

• What’s the most effective analytic presentation you’ve seen recently?

• Let’s judge that objectively: What recent presentation generated the swiftest drive to action?

• What made the presentation effective?

• How could it have been improved?

• What’s unique to the circumstances of the subject matter, and what can be documented as a best practice, to be repeated in your shop?

• Have you recorded, or can you record (screencast with voice-over, live video), the presentation?