I originally bought Whoop because I’m a sports enthusiast and data nerd, and I’ve been loving all the data it has collected so far.
However, I found myself running into a few issues:
-While I do speak with my Whoop Coach quite a bit, I noticed it has some inconsistencies and contradictions. For example, one month, sauna sessions seemed to hurt my recovery, and another, they didn’t seem to make a difference at all.
-Since I also use MacroFactor to track all of my calories, it’s frustrating that there’s no Whoop integration, meaning that Whoop Coach can’t reliably access my nutrition data. Even though it does get it through Apple Health, the conversations surrounding the impact of my nutrition on my Whoop metrics have been very hit or miss.
-I love the reports Whoop includes, but they lack the customisation for me (for example, I hate that I can’t see “all-time graphs” and I’m only limited to 6-month views, and that I can’t create custom graphs or cross-reference different metrics in a visual way).
So, over the weekend, I played around with Claude ("coding" version), a relatively new AI tool that I heard was great at processing large amounts of data. I was reluctant to use it at first as I don’t really love most AI models like ChatGPT, and I'm not a developer, but this one felt different.
In short, I was blown away by it.
Over just a few hours, I was able to connect data from different sources (Whoop data, as well as nutrition and workout data from MacroFactor), build customised dashboards for all of my data, and cross-reference it to find correlations.
This was so much easier than I thought it would be, and I found quite a few things that surprised me - which is why I want to share this all with you.
Here are a few key insights that I found:
Going to bed at 10 pm or earlier, instead of 11 pm or later, turned out to be the single biggest recovery driver in all of my data — bigger than any nutritional factor, workout type, or supplement. The earlier sleep window consistently produced meaningfully higher HRV and recovery scores. I knew sleep timing mattered, and that earlier was better than later, but I didn't expect it to be this important.
Fat intake is the strongest nutritional correlate with recovery (r = −0.314) — on low-fat days (<80g), my recovery is 12 points higher, and my HRV is 14ms higher than on high-fat days. What surprised me most: the effect persists for two full days, not just the night after a high-fat meal. WHOOP Coach never flagged this — because it doesn't have access to my MacroFactor nutrition data.
Added sugar is the second strongest predictor of poor recovery (r = −0.222) — on high added sugar days (>40g), my recovery averages 55% versus 68.1% on low days. Interestingly, total sugar showed almost no correlation. In my case, it seems like the source matters more than the amount.
My personal caloric sweet spot is 2,400–3,200 kcal/day (my maintenance calories are around 2,800-3,000). Above 3,200 calories, my recovery drops by ~9 points. I never would have considered that.
Two behaviours showed up consistently as recovery boosters: a 20-minute evening mobility session and daily outdoor walks. Both improved next-day HRV and recovery scores in a statistically meaningful way. In the past, this was something I did randomly - now I’m sure I prioritise it as part of my daily routine.
Heavier gym sessions actually correlate with better next-day recovery than lighter ones — which seems counterintuitive to me (I need to dig into that further). Climbing shows the opposite: a 1-hour session raises recovery, a 3-hour session tanks it. Same sport, completely different effect depending on intensity.
I also built custom graphs/dashboards for all the data I was curious about (you can see some of them in the photos in this post), such as:
-My total weekly activity count and duration (so I could see how active I was on a week-to-week basis - I counted gym sessions, climbing, vo2max sessions, mobility and walking)
-Strain vs total activity duration
-Gym workouts (sessions & duration)
-Gym Volume (weight × reps, kg)
-Separate graphs for climbing volume, mobility volume, and sauna sessions/duration
-Weekly reports and comparison graphs (so I can compare my metrics throughout specific months)
-Full dashboard tabs for nutrition, sleep, recovery, and training (so all of my data is in one place, one click away
Claude Setup
This is the part I dreaded the most - I didn’t want to spend hours figuring out some super technical AI, and the whole “terminal” thing was intimidating.
It turns out that installing it takes 2 minutes if you just follow their quickstart guide, which you can Google.Just do it, it’s SO much better than using the desktop version of Claude or other AIs for data analysis like this.
Now, in order to use Claude in this way, you will unfortunately have to use the Claude paid plan for now. I suggest starting with Claude Pro ($20/month), I’ve found that it has more than enough tokens (credits) to analyze all of your data, and build your first few dashboards.
I did end up switching to Claude Max after a few days (because I’m obsessed with data and wanted to analyze my body fat on progress photos and create tens of different charts), but for most people, this won’t be necessary. The tokens reset every 5 hours, and unless you want to work on this all day every day, the Pro plan should more than cover you.
Now, to use Claude, you don’t actually need to use the terminal, you can use the free tool from Google called Antigravity (which has a much friendlier UI for non-developers like me) - setup takes 1 minute: https://antigravity.google/Next, I created a folder where I wanted Claude to run. Claude runs locally on your computer (so you don’t actually share your data with anyone, if you have privacy concerns), and I suggest you run it in a specific folder (so it doesn’t have to access ALL the files on your computer).
So, in my case, I created a folder desktop/claude/health, where I put all of the data that I wanted to analyze.
In terms of giving Claude data to work with, the easiest way to do it is to export a .csv file directly from your Whoop App. You go to Whoop, click “More - 3 vertical lines button on the bottom”, App Settings, Data Export. It says the export can take up to 24 hours; in my case (9 months of data), it took about 20 minutes. You can do this once per day.
You can export data in a similar way from other apps like MacroFactor, Apple Health, Strava, etc. - so you can get any data you like quite quickly.
If you want to make things a bit more automatic (so you don’t have to regularly manually export data), you can do this by letting Claude connect to your Whoop API. This has a few extra steps, but you can just ask Claude to walk you through them - it tells you exactly what to do, step-by-step.
Now, this did take some fidgeting for me as my data wasn’t syncing properly, but I solved it by just telling Claude to “go fix it because it still doesn’t work”, and in a few attempts, it did. But to be honest, if you just want to get started with analyzing your data, don’t get hung up on this step - just use the manual data export from Whoop.
Okay, now that you have your folder with your data, there’s one more thing you want to do - to create a “claude.md” file. In plain English, this is the file that tells Claude the context about your project - what you want to do in this folder. It makes it run more efficiently, spend fewer tokens, and get better results.
You can actually have Claude create this file for you. Here’s what you can say:
“In this project, I want you to analyze data from Whoop and find insights in my data. Create a claude.md file and let me know which information you need from me.”
It will ask you a few basic questions - answer them, and you’re ready to get started.
I used Claude to analyze my data in two ways: data analysis and dashboards.
Data Analysis
I used simple prompts, such as:
“Can you map my nutritional data (specific macros and calories) to my HRV, RHR, and next-day recovery scores and see if you can find some correlations?”
“How do different bedtime windows impact my next-day recovery scores?”
“Analyze all of my data (nutrition, workouts, whoop data), and see if you can find anything that has a strong correlation to HRV, RHR, or Recovery %.
“Does doing mobility in the evening make any difference in next-day recovery metrics?”
I used broader prompts (to see what data is actually available, and find specific insights I couldn’t find on my own), as well as specific prompts (with things I was actually curious about).
I also played around with prompts like “act as a nutrition coach/fitness coach/longevity coach/data scientist, what are some red flags you can find in my data?” to analyze data from different angles.
Now, of course, not all responses were great, and many prompts didn’t find anything useful. Now, I’m no data scientist (someone who is might be able to suggest better prompts), but from my knowledge of nutrition, fitness and longevity, I didn’t see any obvious mistake from Claude, which was surprising.
What I love about it is that it worked off my data, not articles from the internet - so all of the responses were based on what I fed him, instead of how things “should be”, which was extremely valuable to me.
Dashboards
Finally, I also built some comprehensive dashboards that show me all of my data in one place.
The first prompt was simple: “Can you build me a dashboard that includes my Whoop and MacroFactor data?”
Next, I played around with different ideas, such as:
-New graphs for metrics that weren’t there yet
-New graph features (such as daily/weekly/monthly time frames)
-Week-by-week comparison reports (to see differences in my weight loss, or recovery scores, for example)
-Creating new metrics (merging all of my gym sessions, walking, vo2max training, mobility, and climbing into one graph)
-Mapping one graph on top of another (i.e., seeing my total activity time alongside my daily strain data)
There’s so much you can do (if you can imagine it, you can probably do it), and creating it all is as easy as asking Claude to do it and giving him feedback when necessary.
Limitations
Now, sadly, there are a few limitations I ran into:
-First, step data isn’t available in the Whoop API, which is a shame (as I personally find step counts to be more accurate on Whoop than Apple Health)
-Also, the vo2max data doesn’t get exported
But other than that, I’ve very rarely found myself limited by the data or technology (and hopefully, Whoop AI can do some of these things in the future inside the app, and integrate with nutrition apps like MacroFactor, so it has access to it).
If you’re a data nerd like me who wants to do more with your Whoop data, I really encourage you to try something like this out. And if you have any questions, just reach out or leave a comment - I’m happy to help with what I learned.
Also, if you did something similar with your own Whoop data, and you’re keen to share, I’d love to hear about any cool insights you had or graphs/dashboards you built that I could experiment with.
