Improve your diet compliance by 3,000 bps in 5 days
I have a system for losing weight. The system works if I follow it. And no, I don’t always follow it. I tend to do well for two months and then I stall out. I’ve lost fourteen pounds since June 2020, so I know it works. But I’ve only been compliant maybe two out of every five days (40%) and I’ve stalled out since Halloween. This past week, I found a way to increase daily compliance to 80% and I’m dropping weight again. The key was changing my key performance metric from a daily score (1/100) to a rolling ten day average that measured binary compliance.
For years I’ve been testing and developing an approach to operational excellence as part of my day job. Like many things I’m trying to learn at work, I apply it to problems at home as well. The philosophy is that all problems are people problems. My weight is a people problem. I’m the people. And the same techniques around data, feedback loops and constant improvement I use at work should work for me too. Since I can’t readily share specific stories about data and management feedback loops from work, I use my weight regime as an illustration of the same principles.
When I first started trying to lose weight, I was committed to being “data-driven.” I had a plan to record what I ate, exercise three times a week, and eat every four hours. I hadn’t measured any of these things, so I decided to measure them all. I tracked each day with a checklist and gave myself a one or a zero for the day. I either did 100% of the things I was supposed to do, or the day was a zero. Of course, with that criteria, almost every day was a zero and I was discouraged. I had unwittingly created a disincentive to compliance.
Being a gamer, I got rid of the binary daily metric and created a gamified system of “points for actions.” Instead of every day being a one or a zero, I would give myself points for not eating after 7pm, points for recording my steps for the day, etc. I was giving myself credit for the steps that I did take. This worked for a little while and helped me stay engaged to overcome the discouraging series of zeroes. But soon I felt like I was wasting my time, putting in effort that wasn’t producing results. It’s like when I took pictures of everything I ate because I was told you needed to track what you eat to lose weight. I couldn’t understand why I was gaining weight despite taking pictures. “See? I take pictures of the pizza, the cookies, the Doritos, everything! Why am I not losing weight?”
Upon reflection, the original binary approach was solely output-focused. I only measured the final result of the day, which was discouraging. I wasn’t giving myself feedback on the things that needed to happen to produce those results. That made me feel like I was being judged for something I didn’t have visibility into or control over, which caused frustration and shame. (Any system that creates even a whiff of shame is DOA. Humans cannot tolerate shame and they will avoid it at all costs.) The gamified approach was also ineffective because it rewarded only effort and not results. It was also discouraging over the long run because I could work hard, get credit for trying despite not producing the results I wanted. It made me feel better short term but the motivation dissipated over several weeks.
It turns out I need three things: 1) visibility into the things I can control, 2) positive feedback for the effort, and 3) measurement of the final outputs. Since the gamified approach already has visibility into things I can control and positive feedback for effort, I added a new results measurement: a ten-day rolling average of daily compliance as measured by a binary goal tracking of the two penultimate output metrics.
Huh?
Ok, that is a very dense sentence. Let’s unpack it, underlined concept at a time. First, why a ten-day rolling average? Similar to your actual weight, things vary a lot day-to-day in ways that can be distracting short term. Like watching a stock price go up and down every hour, each additional data point takes attention and adds little actionable information. It’s noise. But a rolling average provides a trend that can still be acted on. Psychologically, a rolling average forgives short term fails and provides a ready opportunity to improve. It helps avoid “all-or-nothing” thinking like “Ah well, I screwed up today, might as well eat a pizza.” A ten-day rolling average gives me the minimum amount of information in an actionable and yet forgiving way.
It’s amazing how much of reporting and dashboard design is actually about the psychology of the people who have to react to it. You have to provide information that they know what to do with (actionable) and that they want to do something with (engaging).
Second, why daily compliance? Why not weekly or monthly? That one is fairly easy, I have to eat every day, so that gives me daily decisions to make, daily levers to pull. In order for data to be actionable, it has to inform to the next decision you need to make. If that’s every day, then report daily. If its every three years, then do a three-year plan. Daily reporting on something you can only impact every month is a waste of time. And monthly reporting on something you should consider changing weekly is ineffective.
What is “binary goal tracking”? Binary means a one or a zero. It either happened or it didn’t. Goal tracking is variance to a goal, i.e. did you hit the goal metric or not?
What is a “penultimate output metric”? Penultimate means “second to last,” the one right before the final one. It’s a pedantic word, but I like its accuracy. And in this case, it’s important to understand the position of a measurement in the sequence of causality. For my weight, there is one final metric: how many pounds show up on the scale. But what immediately determines that? What is one-step upstream from that weight measurement? My weight is directly determined by how much I eat, measured by caloric intake, and how much I exert, measured by steps on my Fitbit. Daily calorie intake and Fitbit steps are the penultimate output metrics. They are the second-to-last metrics that determine the final output, weight. They are the penultimate metrics.
But calorie tracking is not binary, its gradual, I eat 2,100, 2,500, 1,800 calories, not 1 or 0. Right, but achieving the target calorie is binary. My target calorie intake is 2,100. The calories for the day are less than or equal to 2,100 or they are greater than 2,100. Same with the steps, I either hit the goal of 6,000 steps or I didn’t. A day gets a 1 if I do both. The binary daily score of these two penultimate output metrics is a one if I eat less than 2,100 calories and get at least 6,000 steps. If I miss on either of those, it’s a zero. Then I add up the number of days that were a 1 in the last ten days, divide them and I get my ten-day rolling average.
And today it was 70%, meaning that 7 out of the last ten days I hit my goals of 2,100 calories and 6,000 steps. And guess what? I dropped 2.2 pounds during those ten days and I feel great about it.
— — — — — — — — — — — — — — — — —
I write about data design, leadership, business management, financial management, Cascadia, amateur film making and operational excellence. Please follow me on Medium, Twitter or LinkedIn.
