Читать книгу Decision Intelligence For Dummies - Pamela Baker - Страница 50

Math is the cornerstone of data analytics in particular and of decision-making in general. However, the right math can deliver the right answer to a wrong question, which leads to nothing good in the way of making a sound decision.

How does that happen? It’s the result (mostly, but not always) of communication errors and mismatched assumptions between people or groups. That’s right — the problem has nothing to do with the math. The math is right and the answer is right, yet it’s all wrong because the question was wrong for the result folks were looking for.

For example, it’s quite common for a data scientist or a data analyst to query the data based on a question asked by a business manager or an executive. But managers and executives often pose questions from assumptions rising from their own (limited) perspective, often using imprecise language. Data scientists and data analysts, on the other hand, think and speak in the precise terms and statistical assumptions that are the norm in their crafts. The two seldom meet on the same train of thought.

If you ever want to develop a firsthand appreciation of what I think of as the great data divide, learn any programming language (provided you don’t know one already). The first thing you notice is how profoundly it changes the way you think, the assumptions you make, the way you approach logic, and the expectations you have of machine performance.

The truth is, people fall into patterns of thinking and often can’t imagine that any other pattern exists. Imagine a cake baker asking a bridge engineer to go outdoors and bring back some fruit flies. Never mind that what the cake baker really wants is a set of edible creations that look like flies made of fruit for that entomologist’s birthday cake that’s on order; the cake baker said “fruit flies,” which the more technical-thinking bridge engineer took to mean those nasty little fruit flies that bedevil your fruit bowl. The bridge engineer may work diligently and for endless hours to collect fruit flies and deliver them to the cake baker who will then see this result as disastrous to their own efforts and squash the lot. That, in a nutshell, is why so many data queries end up delivering so little in terms of business value.

Real-world examples that aren’t quite so fanciful are plentiful. As a science-and-technology journalist, I see news publications regularly derailed by their addiction to following the answers to the wrong questions. For example, it’s typical for news media to “run the numbers” to see which articles attract the most audience eyeballs, clicks, likes, and shares. Whatever that outcome is becomes the next list of assignments for staff and freelance journalists. Sounds like a good plan, yes? Well, it is, but only for as far as it goes.

There’s a problem with diminishing returns. Think for a moment: How many times can the same article be written in different variations before readers lose interest and the publications pay for articles that readers won’t read? Those dead-on-arrival articles also impact other metrics, including the ones advertisers consider before buying ads or sponsored content with that publication.

In response, the publications run the numbers again to see which articles are trending now in order to repeat the cycle until it again ends in diminishing returns — all because of a wrong question, and even worse, a wrong question repeated endlessly.

The right question would be one that would put the publication in the lead position of trending articles rather than following the leader at the midsection or tail of moving trends. When a publication can figure out “what readers want to read about” instead of looking just for “what readers are reading now,” they move to the top position in the competitive pyramid. Further, they stand a real chance of commanding more (and higher) ad dollars as well as greater industry respect and brand loyalty.

In this example, one could use the wrong question as a supplementary question or as input in the algorithm, in support of the correct question: “What do readers want to read?” But ultimately this formula falls short as well because the answer is based on the question instead of the destination. How so? Well, consider if I pose this same query and the answer is that readers don’t want to read any of the tech- or science-related topics I write about — they want to read about the newest reality show instead. That’s simply an output — an answer — I can’t use, and neither can any science or tech publication. Magazines that focus on TV entertainment would give this answer a collective shrug, as in, “Tell us something we didn’t already know.” It’s just not that useful of an insight for the work and cost involved for either genre.

Examples of the right questions for this scenario might include:

 Which descriptive words appear most frequently across topics in the most read articles over the past year and how do they correlate with the number of likes and shares on this publication’s articles across social media? (What I’m looking for here are reader triggers and themes of recurring interests.)

 What are the top ten shared memes or social media post issues in my audience demographic and how do they correlate with current or breaking science or tech news? (What I’m looking for here are emerging or sustained interests that I can tap into as popular culture or high-interest angles for articles.)

 How much did writer style and word choices vary between the top performing articles (in terms of eyeballs, clicks, or social media shares and likes) and where are the commonalties. (I’m looking for what kinds of story-telling readers prefer so I can change writer guidelines to improve readability of articles across the board.)

 What is the impact of SEO keywords on article readership? (Here I’m looking to see if incorporating SEO keywords in the text and headline actually helped or hurt readership and to what extent, so I can adjust how stories are written accordingly.)

 What is the overall pattern across all top performing articles over the past six months? (Here I’m looking to see what bells and whistles readers may be responding to, even if subconsciously.)

 What are my competitors top performing articles according to readership numbers and social media shares and what are their commonalities? (Here I’m looking to see if my reader patterns match my competitors and where they diverge so I can consider topic options based on patterns my publication may not have previously considered.)

Now it’s your turn. What do you think the right questions would be for a publication to increase its readership by taking the lead instead of following the crowd?

In decision intelligence, you decide first where you want to go or what you want to achieve and then figure out which tools, queries, data, and other resources you need to get there. Think of it as marking a destination and mapping the course to get there before you take the trip or take an action. In other words, decision intelligence asks you to regroup your decisioning processes so that they focus on specific goals — rather than formulate queries that may prove of little business consequence.

The problem doesn’t lie in the math or the data queries. Rather, organizations have a problem because they lack a clear definition of the desired business outcome, resulting in a lack of direction at the outset of the decision-making process.

Let the business outcome you seek define the queries you ask of data to ensure that your decisions lead you to where you meant to be.