(Re)FÜL – The Algorithm for Human Endurance
(Re)FÜL – The Algorithm for Human Endurance
In endurance sports, you can’t run on talent alone.
By the middle of a marathon, the body has burned through hundreds of calories. The athlete is digging deep, into months of training. And together—that training, those nutritional decisions (from dinner last thursday to breakfast this morning)—form a veritable prediction of the outcome: Will she or won’t she reach her goal?
(Re)FÜL founder Cliff Duhon has spent a lot of time with athletes. Most still have full-time jobs, he explains. Their training times can vary from one hour on Monday to eight hours on Saturday—which means diet should too. A smart athlete will obey the law of thermodynamics: Want more energy output? Get more energy input. But how do regular people find the time, patience, money and/or expertise to script a highly-tailored diet?
Cliff lays out the options: You bite the bullet and do the research yourself. (For every article that says “Eggs are good,” you can find one that says “Eggs are bad.”)
Or, you use the current popular technology: the food log app. You enter your weight, your height, the number of calories you want to eat (which doesn’t take into account other nutritional values). Then you enter your meals and snacks.
After Cliff made a career jump into the health world in 2012 (selling fitness snacks like GU and PowerBar products online), he decided to pursue a third option.
The concept of a log seemed antiquated—reactive rather than proactive. Plus, how many people are going to take the time to analyze their own data, much less analyze it well?
This new way would seamlessly incorporate fitness and nutrition:
One, it would take into account more than ‘weight loss’ (which doesn’t distinguish body fat from muscle gain). Two, it would consider the type of athlete (distinguishing the needs of a marathoner from that of a powerlifter).
And three, it would ‘close the loop between feedback and performance.’ In other words: Cliff wanted to address the knowledge gap that exists between post-workout and pre-workout—a crucial time during which many athletes (who have no way to interpret their data quickly enough) fail to eat or rest or adjust their next workout in a way that will enable them to successfully complete it.
(Re)FÜL is an app that uses real-time, individual performance data to provide automatic and immediate nutritional recommendations—on your phone and/or your fitness watch. It’s a meal plan (including sleep and training) that’s live.
In essence, Cliff’s pursuit of a third option has him facing his own endurance challenge: write an algorithm for the human body.
You Can’t Bastardize Physics.
Cliff approaches nutrition like a scientist.
His interest in nutrition started when he was a pre-med major in college. He read a lot of sports nutrition textbooks. He asked athletes to let him run tests on their performance. He did it all for fun.
Sports nutrition poses a difficult paradox. On the one hand, Cliff believes that—for goal-driven athletes—nutrition is black-and-white. The way your eating and training either works or doesn’t.
On the other hand, nutrition at large is “black and white with a ton of gray in between.”
The black and white parts are the basic principles, like the law of energy input and output. The gray is the dozens of other variables in a person’s life: sport of choice, body type, food intolerances. And that doesn’t even account for the biggest variable of all—life itself. Like the morning you forgot your breakfast, or that one night you just couldn’t fall asleep.
“You can’t bastardize physics,” Cliff says. Nutrition is a science. You also can’t create a static plan for a dynamic journey.
With the help of Garmin’s advanced hardware technology (most endurance athletes wear Garmin watches), the (Re)FÜL algorithm is a guide for the journey.
It tracks sleep, energy intake, fats burned, carbs burned, heart rate, distance, speed, energy expenditure through power, and other metabolic data. It uses that information to tell you what to eat and how to exercise next.
For example, you might not even notice that today’s meal has extra protein. But (Re)FÜL, taking into account your long run this afternoon, put your data to use and did it on purpose. With science.
You Get Out What You Put In
It would be so much easier to make another food log app, Cliff admits. It’s cheaper. It’s faster. You can make a ton of money. “It’s the path of least resistance.”
But Cliff and his team aren’t so different from their clientele. The path of least resistance isn’t the path they’re looking for.
They’re process people. Like any startup, (Re)FÜL is one big experiment. Cliff navigates ‘the gray area’ complexity of nutritional AI the way he always has: Hypothesis. Testing. Doggedly asking the questions, “What does this do? Is it actually going to help?”
Last week, the (Re)FÜL team performed a small test where they looked at an athlete’s daily training and expenditure. They took (Re)FÜL’s calculated predictions about that performance data (based on her nutritional nutritional input) and compared. They were off by less than 5% over the course of a week.
“Now we need to test ten. Then one hundred. Then one thousand,” Cliff says. That’s how the algorithm fine-tunes itself.
When the algorithm doesn’t match up with actual performance, the next question is “How does our algorithm know you.” When what you’re doing doesn’t work, (Re)FÜL is meant to see that, make note, and find what does.
“To be able to provide that real-world, working knowledge of how nutrition is actually impacting your body—that’s huge,” Cliff says. As deep as he is into the studies, the app, the systematic approach, he gets it.
“People don’t live in the lab. They live in the real world.”
Learn more about (Re)FÜL at www.re-ful.com.