I drink two coffees a day. One at 6am before I train. One around 10am when the first one wears off. Nothing unusual. Nothing excessive by any objective measure.
Three weeks ago I stopped both. Not as a life decision. As a test. I wanted to see what caffeine was actually doing to the data I collect obsessively every morning.
What follows is what 21 days of tracked data showed me, what changed, what didn’t, and why the relationship between caffeine and HRV is more complicated than the recovery influencers on Instagram make it sound.
The Setup
Every morning I record HRV from a chest strap measurement on a 5 minute seated protocol. Not a wrist based optical reading. Not Apple Watch variability from overnight sleep. A dedicated morning measurement with a proper sensor.
My baseline HRV over the preceding three months sat at 52 milliseconds with a standard deviation of about 9ms. Typical range 42 to 62. Occasional outliers on either side related to training load, poor sleep, alcohol.
I also track resting heart rate on waking, sleep duration and stages, subjective recovery on a 1 to 10 scale, and training load from my Garmin.
The protocol was simple. Stop both coffees for 21 days. Don’t change anything else about training, sleep timing, nutrition, or alcohol. Just remove the two 200mg caffeine doses from the day.
Week One: The Mess
Days one through four were terrible.
Day one HRV was 48. A little low but within range. Day two dropped to 41. Day three hit 38. Day four was 36, the lowest morning reading I’d logged in four months. Resting heart rate was up 6 beats over baseline.
The subjective experience matched the data. Heavy head. Mild headache by 11am each day. Training sessions felt like running through wet sand. I went to bed earlier every night because I was exhausted.
This is caffeine withdrawal doing exactly what the literature says it does. A disrupted autonomic nervous system, elevated sympathetic tone, and a general sense that you are not okay. It showed up unambiguously in the data.
By day 5 the headache was gone but HRV stayed suppressed. Day 6 was 42. Day 7, 45. The body was starting to stabilise but not back to baseline.
Week Two: The Plateau
Something unexpected happened during week two. My HRV didn’t bounce back. It settled into a new, slightly lower baseline.
Daily readings during week two: 46, 44, 47, 45, 48, 46, 45. Mean of 45.9. My three month baseline had been 52. I was running about 6ms below my pre-experiment norm.
This was not what I expected. The dominant story in wellness media about caffeine is that it suppresses HRV while in your system. Remove the caffeine, get your real HRV back, it’s higher, you sleep better, recovery improves. That’s the narrative.
My data said something different. HRV was lower without caffeine, not higher. Consistently. For a full week after acute withdrawal had cleared.
Week Three: The Reckoning
Week three continued the pattern. Readings: 44, 46, 45, 47, 46, 45, 47. Mean of 45.7.
Three weeks in, no sign of a rebound. My resting heart rate had returned to baseline by day 10. My sleep duration was the same, maybe marginally longer but inside normal variation. My training sessions felt fine. Subjective energy was decent, not great.
But the HRV number, the thing I wake up and check first every morning, was persistently suppressed.
What Actually Happened
I put the data together and talked it through with a sports physiologist friend who has seen this pattern before. Here’s what’s going on.
Caffeine has two effects on HRV that pull in opposite directions.
Acute effect. In the minutes to hours after consumption, caffeine raises sympathetic nervous system activity. This suppresses HRV. If you measure your HRV an hour after drinking coffee, you will see a lower number. This is well documented.
Adaptive effect. With chronic habitual use at moderate doses, the nervous system adapts. Vagal tone, which drives high HRV, often strengthens. Morning HRV, measured before any caffeine, tends to be higher in habitual coffee drinkers than the narrative would predict. This is less well known and less discussed.
Remove the caffeine entirely and you lose the adaptive effect. The nervous system downregulates to a new baseline that reflects the absence of the stimulus it had been working with. HRV settles somewhere below where you started, not above.
This isn’t a universal finding. Some people see HRV rise when they quit caffeine, particularly people who were consuming large doses or drinking late in the day and disrupting sleep. But for someone consuming moderate doses before 10am with no sleep impact, the data can go the other way.
The Timing Question That Actually Matters
Here’s the insight from three weeks of tracking that changed my behaviour.
The caffeine is not the problem. The timing is.
During the 21 day washout, I was already experimenting with measurement windows. I tested HRV at 7am before coffee, at 9am one hour after coffee, and at 4pm in the afternoon.
When I resumed caffeine on day 22, the morning pre-caffeine readings went back to around 52ms within four days. The one hour post-caffeine readings dropped to around 44ms. The afternoon readings depended entirely on whether I’d had a second coffee and when.
This is the practical point. Your HRV is a snapshot. What it shows you depends entirely on when you take it relative to your behaviour. Measure before caffeine, you see one picture. Measure an hour after, you see another.
The wearable on your wrist is taking continuous or periodic readings all night and averaging. It’s not controlling for caffeine timing. It’s not telling you which readings are pre and post stimulant. It’s mixing them together and giving you a single score.
Your Oura or Whoop recovery number for a given morning can shift by 10 or 15 points based on whether the device happened to sample more readings before or after your last coffee. This is noise masquerading as signal.
What I Actually Do Now
I went back to my two coffees a day. The data during the experiment made clear that for me, habitual moderate caffeine was not suppressing my baseline HRV. Removing it suppressed it.
But I changed two things.
I moved both coffees earlier. First one at 6am, second before 9am. Nothing after 9am. This removes any possibility of caffeine affecting my night sleep architecture, which is the mechanism by which caffeine most robustly damages recovery.
And I stopped trusting any single HRV reading. Morning chest strap reading with protocol is my reference. Everything else is context, not truth.
The Broader Pattern
This is the general problem with interpreting any single number from wearable data. The number is always a snapshot of a complex physiological state, taken under conditions that the device is not controlling for.
Caffeine, alcohol, meal timing, hydration, time of measurement, sensor placement, skin contact, ambient temperature. All of these shift the reading in ways that the device cannot disambiguate for you.
The answer is not to track more. The answer is to track in a way that respects what each measurement can and cannot tell you, and to make decisions based on patterns rather than individual data points.
Three weeks of careful self tracking taught me something about my physiology. It also reminded me how easy it is to misinterpret the data a wearable is handing you every morning.
Green score. Destroyed legs. There are 6 blind spots in your wearable data. We wrote a free guide covering every one of them.
Green score. Destroyed legs. There are 6 blind spots in your wearable data. We wrote a free guide covering every one of them.
Download the Free Guide