For consultants and all those who solve problems for a living, we know that “average” is a bit of a dirty word. A misnomer.
“Average” is a four-letter word
- Averages are lazy; shows a lack of effort.
- Averages are inherently reductive and often misleading.
- Averages are dehumanizing; no meaningful problem was solved by looking at averages.
- Averages don’t provide answers, only stereotypes.
- Sadly, averages are typically a little true – enough to keep this misnomer going.
1. Life is non-linear; Pareto everywhere
The Pareto principle is alive and well; yes, often a small # of inputs, drive a large # of output. For example, in the United States 5% of the population drive 50% of healthcare cost. In a similar way, the top 1% of the Americans own 35%+ of the total wealth. So when we talk about the “average patient” or the “average American”, what are we really saying?
2. Averages are approximate, like a US zip code
An average is a useful short-cut. A heuristic. Makes sure you are in right zip code, that’s about all. Once you have a sense of the magnitude (100 vs. 1,000, vs. 10,000), then it’s time to dig into the details and understand why.
3. So many different “averages”
Median, mean, mode are all estimates of central tendency. So which one are you talking about?
4. Benchmarks, who are you comparing against?
Benchmarks are tricky. It’s only as useful as the sample size and composition. I remember comparing my high school grades to any/all friends who got Cs. “Yeah, mom, but I did better than Jimmy. Jimmy got a C.” Trust me, this logic did not work on my mom, nor will it work on your investors, or your board.
5. Time frame matters, right?
No strategy is eternal. What might make NO sense over the next 12 months, could be exactly the tough decision, investments needed for the longer-term. This is why pie charts are not favored by visual designers because they only show a snapshot. 46% of people think XYZ. . . that’s kinda useless without knowing what was % 1 year ago, 3 years ago, and 10 years ago. Given a choice, trending charts are better because they show history, momentum, and context.
6. Descend into the particulars
Malcolm Gladwell likes to say that you have to “descend into the particulars” to really understand a story. The next time you say “average” during a business presentation, be prepared to field some questions from the functional experts:
- How does that vary by country, by channel, by product generation?
- How does that vary by customer segment?
- What does that look like year-over-year, and by same-store-sales?
- What’s that look like when you control for foreign exchange impact, and product mix change?
- What’s the total cost of ownership (TCO), not just the price of the equipment?
- Are there any outliers that are driving the “averages” up or down?
7. Definitions change
I remember looking at a S&P500 stock return chart recently, where it showed a 9% average return from (a long time ago until 2019). What’s not perfectly clear, and therefore potentially misleading, is that companies that go bankrupt (etc) are removed from the calculation. If we are going to calculate “average” return, we should factor in all the losers too, right? No, I did not have the patience to read the 41 page S&P500 indices inclusion methodology here.
So what? Why should we care?
It’s up to us. When we read, listen, watch. . . we are forming opinions, judging, and making decisions. Watch out for the “average” when we think about our health, elections, economic security. It’s a great opportunity to ask the 2nd and 3rd question. Stay curious. Don’t satisfice for the easy answer. Be THAT person.
So what? How to combat these deadly 7 sins?
There is no magical armor (read: Fornite skin) that will protect you from the lurking statistician. That said, a few smart moves:
- Find a sample size that is large (enough) and representative; This prevents the problem of calculating the average salary of graduating UNC geography majors in the year that Michael Jordan graduated. This let’s you answer the first question from the audience: “What’s the time frame of that data?”
- Document all your assumptions, sources, and calculations. This outlines the shape of the puzzle you are solving. It allows the audience to get their bearings. It makes your work easy to audit, confirm, and peer review. It’s not a black box.
- Read the audience (functional diversity, level of fluency with data). Let’s make sure we are answering the questions being asked. If we are trying to communicate in a way that is “heard by the audience”, let’s figure out who will be there, what they know, and what they care about.
- Think through the top 10 questions that you might encounter. What’s the standard deviation? What were the outliers?
- Represent the data in multiple formats. Show histograms, trend data. Even better, let them play with the cleansed data directly.
- Develop trust. If your audience knows your intention & objective with data they are more likely to hear what you are saying. No data is neutral.