Insight from 10,000 Weight Loss Diets
by Dr. Mike Israetel, Chief Sport Scientist / Anshuman Radhakrishnan / Dr. Spencer Nadolsky / Carolyn MacDonald |
Jul 07, 2020
The RP Diet App has been helping people lose fat and build muscle with custom, AI-generated and adjusted nutrition plans since 2018. Tens of thousands of people use the app every month to help them build, modify, and stick to diet plans, which means that we are accumulating a lot of data on dieting behaviors. To try to learn about what dieting choices and behaviors work best, we retrieved some select metrics from a recent cohort of over 10,000 fat loss diets each between 6 and 12 weeks in duration. These are real diets that real people performed, with, of course, varying degrees of success.
When we analyzed these diets, some very interesting trends surfaced, and we’d like to share them with you, as they might be insightful for your own dieting behaviors going forward, whether or not you are using the RP Diet App. Observed trends follow!
1. More people have more success with shorter diets than longer diets
In order to calculate diet success, we took the amount of weight lost on an average completed diet and multiplied it by the probability that the diet would actually be completed. It’s no surprise that people lose more weight on 12 week diets than they do on 6 week diets, but that information alone hides at least one lurking downside; not nearly as many people make it through a 12 week diet as do a 6 week diet. The longer the diet, generally, the less likely the average person is to follow through with it to the end. And though longer diets promise greater total weight loss, they don’t deliver as often, which has also been shown in other diet research. In fact, the drop off in diet completion between 6 and 12 weeks is so large, that, even accounting for the smaller total weight loss, 6 weeks diets lead to more diet success. That is, if you want to do a diet and you’d like your best chances of actually losing the most weight, the prior experience of the over 10,000 people in our sample seems to suggest that diets closer to 6 weeks in length are better bets than those closer to 12..
2. Moderate weekly weight loss rates seem best
When many people start a diet, they choose the most rapid weekly weight loss possible, which in the RP Diet App is 1.0% of weight loss per week. This means, for example, that a 200lb person might attempt to lose around 2lbs per week for every week of their diet. On the other hand, other people might see that as a rather Napoleonic pace, and might instead opt for the slow and steady approach, which in our RP Diet App is also an option, with loss rates of as low as 0.25% per week possible (around half a pound a week for our 200lb example). Which one is best? It turns out both have their place. Our analysis revealed that diets closer to around 6 weeks in duration actually seem to generate the best results (amount of loss multiplied by the chances of completion) when paced at the higher end of loss, around 0.6-0.8% body weight per week. This would be around 1.2-1.6lbs lost per week for a 200lb person. However, the longer a diet goal is between 6 and 12 weeks, the better the slower paced loss rates start to perform. By the 12 week mark, the best diet results actually come from diets that are in the 0.4-0.6% range, which would be around 0.8-1.2lbs for a 200lb person.
It seems that with the weight loss results seen in our sample of dieters, people were maximally successful when losing about 5.5% of their body weight over the course of a 12 week period, whether they did that more quickly in 6 weeks or more slowly over the whole 12. In fact, without adjustment for the probability of finishing the diet, the 12 week dieter would have lost more weight. But, they failed to reach their end goal much more often than those who dieted for just 6 weeks, giving the shorter diet group the slight edge in overall results. We’re not yet sure if this relationship occurs similarly in all subroups (males, females, very overweight, not very overweight, etc.), but we will be looking into that in later research.
3. Following the plan is the better choice
As you follow the RP Diet’s AI-generated diet plan, you record your body weight over the week, and at the end of the week, the AI gives you a few options. The first option is to accept its recommendations. It generates these by calculating to see if you’re on track, and recommending increasing, decreasing, or maintaining food intake over the following week to keep you on track to your chosen goal. The second option is to go your own way. You can choose to increase, decrease, or maintain food intake against the recommendations of the AI. For example, if the AI tells you that maintaining your current food intake over the upcoming week will put you on the most likely path to your goal, and you tell the app to reduce food intake in the upcoming week instead, you’ve chosen the second (manual) option.
As app designers, we were very curious to see how the advice our AI algorithms were generating fared, and of course we were hoping they weren’t taking app users down a path to poor results! But on the other hand, we did program the manual overrides for a reason; we knew that many of our users would need a diet break here and there, or know their body responses to dieting a bit better than the AI, so we were expecting the manual option to perform pretty well in at least some instances.
Once we ran the numbers, we were pretty surprised. The correlation between AI-recommended adjustment use and diet success was quite robust. In plain English, the most successful diets were more often the ones in which users accepted every single one of the app’s AI recommendations, and ventured out on their own adjustments zero times. There was a very small cohort of super-dieters (about 2% of all diet users) that dieted much longer (12 weeks) and faster (1.0% per week), and seemed to get their best results with occasional use of just one “go your own” adjustment per diet. But everyone else seemed to do best when they followed the app’s recommendations each week.
We considered that perhaps the results of this analysis were survivor-biased. Of course the people that actually followed the AI’s recommendations would lose more weight; but all those that tried and failed because the adjustments were too aggressive wouldn’t be counted! It’s like declaring that the Hunger Games have a 100% survival rate by just interviewing the winners! To examine this possibility, we looked at the percent chance of completing a diet based on how many AI adjustment recommendations were taken. It turns out that people who accepted more adjustments actually had higher rates of diet completion than the average! It would seem that there wasn’t much survivor bias here at all. People that accepted more AI recommendations just seemed to get closer to their diet goals.
What can we conclude from this analysis? Well, we have to be very careful with how we phrase any conclusions. Because the diet is all correlational, we cannot imply any causation from it. In addition, the entire sample of dieters is self-selected, so this kind of analysis doesn’t even come close to the conclusive ability of any kind of controlled diet experiment. However, the sample size is very large, and the behavior patterns very consistent, so some tentative insights might be appropriate.
If you’d like to maximize your chances of success based solely on what has seemed to work best for the diet users in this sample, your best bets might be to:
- Try a shorter diet first. Perhaps a 6-8 week diet, especially if it’s your first time trying to change your eating to lose some weight.
- Avoid extremes of weekly weight loss. Aiming for about 0.5-0.7% weight lost per week (about 1-1-1.5lbs per week for a 200lb person).
- Buckle down and follow the plan. If the app, or whatever diet you’re using (including what your coach says if you have one) tells you to increase, maintain, or drop calories, it’s probably a good idea to take that input seriously. It’s very tempting to go your own way on a diet, especially when impatience or cravings get the best of you, but staying the course seems best.
Thanks a lot for reading, and we’ll continue to update the app and analyze diet results so that we can learn a bit more about how help more folks become more successful with their diets in the future! Stay tuned!