The Levels Theory of Behavior Change

The Levels team describes how continuous glucose monitoring affects behavior change.

Millions of Americans want to eat better and exercise more, and yet find it exceedingly difficult to make these changes and stick to them. Despite our best efforts, only 5% of Americans get the recommended amount of physical activity, 72% of Americans are overweight or obese, and 59% say that the conflicting information about nutrition makes them doubt their food choices. As a society, we find it difficult to make healthy choices consistently, and this is the code that needs to be cracked to modernize healthcare and improve health at scale.

There are a lot of reasons why it is so hard to choose healthy behaviors, many of which are entrenched into our society. These range from cultural norms regarding nutrition and “comfort food,” relentless food marketing and advertising, the addictive nature of ultra-processed food, culturally-reinforced taste preferences, complex psychological and social dynamics around food, limited access to healthy foods for many individuals, public policy that financially bolsters production and purchasing of disease-promoting foods, and a culture that promotes sedentary behavior. What’s more, we have a healthcare and medical education system that still largely emphasizes reactionary sick-care over prevention and disease reversal.

Currently, the leading causes of illness and death in the United States are chronic diseases like heart disease, cerebrovascular disease, Alzheimer’s disease, diabetes, and infections.

The commonality between all of these diseases is that you are significantly more likely to get these diseases and die from them if you have a poor diet and lifestyle. Even something like getting extremely sick from the flu is directly related to how we eat and live, as evidenced by individuals with diabetes being six times more likely to need hospitalization for influenza than non-diabetic individuals. Meaningful behavior change in how we eat and live is at the center of reversing our chronic disease epidemic, and without effective tools to achieve these aims, we will never dig ourselves out of our current chronic disease epidemic hole.

Fighting chronic illnesses primarily with drugs and surgery is the wrong approach. If the goal of the battle is to improve poor diet and lifestyle at scale, the weapon needs to be tools to improve diet and lifestyle, and drugs and surgery do not do that.

The vast majority of chronic diseases are strongly linked to insulin resistance, a condition of poor metabolism where cells become less responsive to the hormone insulin, which helps cells take up glucose for energy. Spiking blood sugar over and over throughout your life will cause insulin resistance, and minimizing blood sugar spikes throughout life will greatly decrease the possibility of it ever developing, and can help reverse it.

It seems simple enough.  But if it were that simple, then how have we ended up with half the adult US population (and likely many more) having overt insulin resistance?

Epidemic levels of insulin resistance is a result of our changing diets, lifestyles, and toxic exposures over the past century. This condition of insulin resistance feeds into almost every symptom and condition you could imagine, and contributes to:

  • Obesity
  • Alzheimer’s dementia
  • Cardiovascular disease
  • High blood pressure
  • Cancer
  • Chronic kidney disease
  • Nonalcoholic fatty liver disease
  • Depression
  • Anxiety
  • Stroke
  • Peripheral vascular disease (and resultant amputations)
  • Preventable blindness
  • Dry eyes
  • Infertility (PCOS)
  • Erectile dysfunction
  • Impaired immunity
  • Gout
  • Arthritis
  • Chronic fatigue
  • Chronic pain/fibromyalgia
  • Gastric reflux
  • Acne
  • Skin tags
  • Hidradenitis suppurativa
  • Psoriasis
  • Increased flu severity
  • COVID-19 mortality

There is virtually no symptom or disease that is not touched by insulin resistance. Improving insulin sensitivity and metabolic fitness is an unequivocal and high value way to reduce risk of chronic disease and improve daily life.

Unfortunately, the American healthcare system has not yet figured out how to inspire people at large to adopt a diet and lifestyle that prevents and reverses metabolic dysfunction and insulin resistance.

This is because the healthcare system has not developed tools that tap into key features of sustainable behavior change, and has not had economic incentives to do so. A doctor or coach telling a patient general recommendations about what to eat or how to exercise is not a behavior change tool, it’s a conversation. And these conversations do not tend to have their desired effect, as evidence by research that shows that up to 40% of patients do not adhere adequately to physician instructions for lifestyle changes, with the prevalence rate rising to 70% or more when significant lifestyle modification or complex behavior changes are required. Even with something as simple as taking a medication, 50% of patients do not take their medications as prescribed by their doctor. Top-down advice given in a typical healthcare visit does not work at scale to fundamentally improve the health of individuals.

What’s worse, many doctors aren’t even discussing lifestyle and diet with patients, despite unequivocal evidence that it improves chronic disease outcomes. Why might this occur? Many physicians get less than 10 hours of nutrition training in medical school, and insurance tends to reimburse poorly for extended counseling. The result is a medical culture where patients with clinical obesity receive dietary and lifestyle counseling only about 30% of the time by their primary care physician.

Simply put, our medical culture does not currently incentivize or achieve healthy behavior change at scale, and this is costing lives.

Levels solves this type of behavior change. We believe that effective behavior change stems from 4 factors:

  1. Closed loop systems
  2. Body-awareness and interoception
  3. Accountability
  4. Personal control

These features are built into the Levels product. By tracking glucose in real time, having personal access to biometric data, and having granular visibility into how choices are affecting health in real time, transformation occurs, is motivated from within, and is decoupled from emotion or deprivation. This is a patient-driven approach that leverages behavior change research and insights to meaningfully inspire behaviors that have a direct relationship to the physiology that underlies a majority of the diseases and symptoms faced by our population.

1. Closed Loop Systems

The thesis of Levels is that behavior change comes from closed loop systems. The concept of “closed loop” and “open loop” systems comes from control theory:

In open loop control, the control action from the controller is independent of the “process output”.

In closed loop control, the control action from the controller is dependent on feedback from the process in the form of the value of the process variable.

Diet and nutrition is a notorious open loop system, in which people have never had a direct feedback mechanism for how their diet is affecting their health and lifestyle. Many of our customers describe diet as a “black box”, where they make choices based on what they’ve read, but they really have no idea if it’s working.

The current feedback mechanisms for diet are all open loop, and include things like:

  1. DNA tests, in which you get a result along the lines of, “bananas are bad for you”, but it’s hard to connect that result to real outcomes after eating a banana.
  2. Food journaling, in which you write down what you eat and how you feel, and you can sometimes make tenuous connections between cause and effect.
  3. Blood tests, in which your results are a multi-month (or multi-year) trailing indicator of your health, and if your numbers are out of range, a doctor will frequently say something along the lines of, “You should make better choices”, which lacks the specificity necessary to lead to behavior change. Getting a yearly result of a fasting glucose of 101 mg/dL (early prediabetic range) does nothing to specify what behavior changes need to happen to reverse this.
  4. Monitoring weight, which can be hard to interpret due to dynamic influence of weight by hormonal fluctuations, water, and inflammation.

It’s worth further clarifying how a closed loop system differs from an open loop system. In a closed loop system, it’s possible to directly attribute action with consequence. This is not to say that other behavioral mechanisms that give real-time feedback (like gamification, streaks, or community support) are not useful, but they are not closed loop systems, as they don’t engender solid attribution.

Continuous Glucose Monitoring (CGM) is the first tool that we can use to measure real time biomarkers and close the loop on diet and nutrition—a notoriously opaque and misleading market that leaves most people feeling helpless and confused.

Glucose is not a panacea for metabolic health, but it’s the single most important thing to track for the 80%+ percent of the population who exhibit signs of metabolic dysfunction, and more than 50% with overt insulin resistance. It is also the only blood based biomarker that is able to be monitored at home continuously, in the “background,” without being initiated by the user/patient. This is what makes it a great place to start.

When we consider next steps for what we should focus on, we need to keep our thesis in mind. There’s no limit to the functionality we can add to a software product that leverages a glucose data stream, and many of these will be distractions from our thesis. For example, we could easily add genetic data into Levels, which would add value but would not close a loop and would likely not lead to behavior change.

Fitness, sleep, heart rate, and activity data, on the other hand, is real time data and can have an immediate and measurable impact on how somebody responds to a given choice, helping us move towards a more intelligent technology that is able to determine the “weight” of the impact of various choices (sleep, exercise, stress) on overall glycemic control.  This makes it a great candidate for a data stream we should consider adding in the near-term.

Similarly, automated event detection allows us to prompt our customers with quick feedback during (or immediately after) an event that can close the loop on a given behavior, which makes it a good candidate as well.

2. Interoception

One of the reasons closed loop systems are particularly powerful in this context is because they improve interoception (also known as “somatic awareness”), which is a person’s sense of the internal state of their body at a given moment in time.

Having awareness of signals originating within the body (ie, a physical awareness of one’s heartbeat or respiratory patterns) has been associated with better physical and mental health outcomes, and increased motivation. Conditioning to improve interoception capabilities has been hypothesized to be important in chronic pain, high blood pressure, gastrointestinal disorders, mood and anxiety disorders, addiction, and asthma.

Interestingly, training to improve interoception capabilities — such as structured heartbeat biofeedback that improves heartbeat discrimination capabilities — has been shown to significantly improve health outcomes. In short, it appears that becoming more sensitive to bodily workings through biofeedback — which can “close the loop” between internal processes and psychological state —  is important for wellbeing.

It’s common for people to feel tired, energized, angry, unmotivated, depressed, distracted, and a host of other states without understanding the reasons why they feel that way. When it comes to subjective mood states, it’s easy to misattribute causality because there are so many variables that can affect how one feels—sleep, hydration, caffeine, exercise, stress, food intake, food timing… the list goes on.

In fact, it was exactly this type of misattribution error that led to the first lightbulb moment that Sam, Levels Co-founder, had while using Levels. On his second day using a continuous glucose monitor, after his normal healthy breakfast of steel cut oats, he started to notice that his hands were shaky, and he felt tired and hungry.

In his head he was thinking, “Oy… Well I did have 3 cups of coffee this morning, and I didn’t really sleep that well… This caffeine crash is getting out of control. I’m switching to tea starting tomorrow.” Then he checked his levels and saw that his glucose levels had spiked to 215 mg/dL and he was crashing down to 50 mg/dL (for reference, individuals without diabetes rarely exceed a glucose level of 140 mg/dL; for additional context on what’s normal, check out our blog post). He was experiencing severe hyperglycemia followed by reactive hypoglycemia and the symptoms that come with it, and he misattributed it to coffee or sleep.

In fact, this is something he’d been doing for his entire life, and this is the first time he had data to guide his decisions. He switched to eggs for breakfast and no longer experienced these symptoms. It was eye opening for Sam, who has read dozens of books on diet and nutrition and was actively trying to make healthy choices, was still doing things that made him feel bad throughout the day—not to mention that they were negatively affecting his long-term health.

In the context of glucose and metabolism, we call this “metabolic awareness.” It’s a newly discovered intuition for most people in which you can start to associate subjective feelings with specific choices, confirmed by data. When your heart is racing and you feel warm, you might start to recognize that as a symptom of hyperglycemia after eating something with a lot of sugar. Likewise, you can start to connect your crashing energy levels and lack of focus with the carb-heavy lunch you had earlier and the subsequent hypoglycemic crash.

If you can close the loop by allowing people to specifically attribute how they feel with a decision they’ve made, you not only allow them to consciously improve their decision making, but you also change their subconscious decision making processes.

What we’ve found amongst our team and many of the individuals who have gone through the Levels program is that glucose biofeedback removes much of the emotional or conscious elements of behavior change. With a closed loop between glucose and a negative sensation like fatigue or anxiety, and the attribution of the negative sensation linked to a specific choice (food, lack of sleep, etc), people have found diminished enjoyment and reward in the trigger, making it inherently less desirable and easy to eliminate.

It’s hard to explain unless you’ve experienced it yourself, but when the loop is closed after making a particular dietary choice—especially if it’s associated with a negative lifestyle event like feeling tired—you don’t need to make a conscious decision to change your behavior because the decision is made by your body. It’s more akin to getting food poisoning, where you don’t consciously decide, “I’m not eating clams ever again,” because the decision is “made” for you by your body; after food poisoning, the idea of clams is inherently no longer associated with pleasure or reward.

Glucose biofeedback has a similar impact, subtly nudging us towards more metabolically-friendly food choices by conditioning us with feedback that changes the reward equation. It also helps us understand with granularity how choices relate to subjective experience, thereby potentially strengthening a novel form of interoception. It’s these improvements to interoception that we expect will make for incredibly powerful and lasting behavior changes.

3. Accountability

One commonly understood mechanism of behavior change is adding an obligation that brings with it social pressure to stay adherent. Many organizations have used this tactic successfully, including Alcoholics Anonymous (where you’re accountable to your peers in the group), Omada, and simple arrangements like hiring a personal trainer to force you to work out at a regular time.

When we started Levels, accountability was not a value proposition we considered, but the more feedback we gather the more we realize that for certain markets, adding an accountability layer is extremely powerful.

Accountability also covers much of what is commonly referred to as “gamification”. In our case, that would include features like “streaks” and other mechanisms that show longitudinal progress.

Another aspect of accountability might be better defined as competition. One of Fitbit’s breakout moments was when they introduced the ability to compete with your friends and family members to see who hit their goals and took the most steps. This not only encouraged adherence to your goals through social pressure, but it also built in a viral sales loop, where people would encourage their friends to purchase the product so they could compete with each other. The nature of accountability within a competitive framework falls into the “autonomous accountability” category of the Self-Determination Theory, whereby individuals pursue accountability from a foundation of internal motivation to execute a desired behavior.

4. Control

Finally, an additional feature of behavior change captured by the Levels product is increased control.  Recent research shows that in 2018, 80% of consumers found conflicting information about food and nutrition, and 59% say that the conflicting information makes them doubt their choices. Individuals are currently faced with seemingly endless food marketing claims, competing popular nutritional ideologies, and nutritional recommendations from different medical governing bodies, and this can create a counterproductive environment rooted in the “paradox” or “tyranny” of choice. In this model, individuals with too many choices have decreased motivation, less satisfaction with choices, and creates decision-related anxiety and regret.

We believe that the objective datastream generated by glucose monitoring makes the myriad daily choices drastically simpler and allows individuals to execute on better choices with increased control. For example, knowing that oatmeal spikes glucose to 180 mg/dL effectively eliminates this as a positive nutritional option. This is now no longer a food choice. Knowing that taking a 20 minute walk before eating a banana reduces a subsequent glucose spike makes walking a clear positive choice.  Objective data that CGM provides can engender a unique and renewed sense of objective control, which has an impact on behavior change motivation; research suggests that “perceived control is associated with emotional well-being, reduced physiological impact of stressors, enhanced ability to cope with stress, improved performance, and a greater likelihood of making difficult behavior changes.”

What’s more, this process is extremely efficient. Using Levels for just a month can clarify eating and lifestyle decisions that an individual may have been waffling about for decades.


Successful behavior change at scale is required to solve the metabolic health crisis. If hundreds of millions of Americans do not make the consistent daily choices to eat and live in a way that minimizes glucose variability, we will continue to see epidemic rates of insulin resistance and related chronic disease. Drugs and surgery cannot be the lifeline thrown to everyone; they are more often a band-aid that does not actually reverse the dysfunctional physiology that causes disease.

Drugs and surgery are what we refer to in medicine as “tertiary prevention,” meaning prevention of complications if someone has already developed a disease.  If there is anything we have learned about chronic lifestyle related diseases, it is that we need better solutions than waiting till someone is in florid diabetic ketoacidosis (a severe manifestation of uncontrolled diabetes) to begin proper treatment. At Levels, our long-term goal is to enable primary prevention, meaning prevention before someone has overt disease. If people are to avoid diabetes, they have to be equipped with tools that help them monitor the drivers of these diseases, well before they manifest in full blown pathology.

The conventional top-down approach to healthcare advice, coupled with a bias towards drug and surgical interventions, will not work to meaningfully improve human health, resilience, longevity, or happiness. Advice does not inspire sustained behavior changes at scale, and drugs and surgery manage downstream symptoms and manifestations of disease, but do not typically reverse or prevent disease.  What’s more, a physician or coach cannot be with a patient 24 hours a day intelligently inspiring and informing choices, and yet this is fundamentally what is needed, because disease is a result of the hundreds of choices we make daily.

Digital health technology is poised to take on the chronic disease epidemic in a radically different way, and we are fortunate that glucose monitoring tools exist to make this sustainable positive behavior change a reality for anyone who wants to optimize their health and prevent future disease. By focusing on aspects of behavior change and human nature that are robust and leverageable, digital health technology can catalyze progress in ways that the traditional practice of medicine cannot.