Most sleep advice tells you to stop drinking coffee after 2pm. Some people follow this rule and still sleep terribly. Others drink espresso at 8pm and sleep fine. Neither person is broken — they just have different caffeine metabolisms.
Caffeine's half-life — the time it takes your body to eliminate half the dose from your bloodstream — averages 5 to 6 hours in healthy adults. But "average" hides a range of 3 to 9 hours depending on factors you may not have considered. If you're on the slow end of that range, a 2pm cup of coffee still has meaningful caffeine in your system at midnight.
This post walks through the math, the variables that shift your personal half-life, and how to use wearable data to find your actual cutoff time.
The Half-Life Math
Here's the core calculation. Caffeine follows first-order kinetics — each half-life period eliminates half of what remains.
Starting dose: 200mg (roughly a 16oz drip coffee) Half-life: 5.5 hours (population average)
| Time After Consumption | Caffeine Remaining |
|---|---|
| 0 hours | 200mg |
| 5.5 hours | 100mg |
| 11 hours | 50mg |
| 16.5 hours | 25mg |
| 22 hours | 12.5mg |
For context: caffeine starts noticeably affecting sleep architecture at doses around 100mg in the system at bedtime. At 25mg, most people experience minimal impact. At 50mg, it varies.
The practical target: get below 50mg (ideally below 25mg) in your system by your intended sleep time.
Working Backwards from Bedtime
If you want to be in bed at 10:30pm with less than 25mg of caffeine in your system, and you drank 200mg at some point:
- 200mg → 25mg requires 3 half-lives
- 3 × 5.5 hours = 16.5 hours before bedtime
- 10:30pm minus 16.5 hours = 6:00am
By average metabolism, a 200mg coffee after 6am still has 25mg+ in your system at 10:30pm. This is why the standard "no coffee after 2pm" advice is too permissive for many people — it only works if your target bedtime is 11:30pm or later, and you have an average metabolism.
A standard 8oz drip coffee contains approximately 95mg of caffeine. A 16oz drip (medium size at most coffee shops) contains approximately 190-210mg. An espresso shot is about 63mg, but most espresso drinks contain 2 shots (126mg). Energy drinks vary widely — check the label.
Related: Our Caffeine Calculator can help you apply these ideas. For the complete picture, see our The Complete Guide to Supplement Tracking.
Why Your Half-Life Isn't the Average
The CYP1A2 enzyme in your liver is responsible for breaking down roughly 95% of the caffeine you consume. Variations in the gene encoding this enzyme create meaningfully different metabolic rates.
Fast Metabolizers (Half-life: 3-4 hours)
People with the CYP1A2 fast variant process caffeine quickly. For a fast metabolizer, a 2pm coffee is largely cleared by bedtime. These are the people who can drink a cup after dinner and fall asleep without issue.
Fast metabolizers often need to consume more caffeine for the same effect, and they may experience a more pronounced "crash" a few hours after consumption.
Slow Metabolizers (Half-life: 7-9 hours)
People with the CYP1A2 slow variant are at the other end. A coffee consumed at noon may still contribute 50-75mg to their bloodstream at midnight. Research has associated slow caffeine metabolism with higher cardiovascular risk from coffee consumption — the coffee itself isn't the problem, it's the prolonged exposure to elevated caffeine.
You can get a rough sense of your CYP1A2 status through consumer genetic testing (23andMe, AncestryDNA — look for the rs762551 variant). But the most practical way to determine your personal half-life is tracking, not testing. More on that below.
Other Factors That Shift Your Half-Life
Oral contraceptives: Estrogen competes with caffeine for CYP1A2. Women taking oral contraceptives typically have a caffeine half-life 50-100% longer than baseline — meaning a 5-hour half-life becomes 7.5-10 hours.
Pregnancy: Similar mechanism to oral contraceptives. Half-life can extend dramatically, especially in the third trimester.
Liver function: Any condition that impairs liver function (including heavy alcohol use) will slow caffeine metabolism.
Age: Caffeine metabolism tends to slow slightly with age, though this effect is modest compared to genetic variation.
Smoking: Counterintuitively, smokers metabolize caffeine approximately 50% faster than non-smokers. This is why ex-smokers sometimes find that their caffeine tolerance drops when they quit.
Altitude: Some evidence suggests slower caffeine metabolism at high altitude.
Chronotype and Cutoff Time
Your chronotype — whether you're naturally an early riser or a night owl — affects when your caffeine cutoff time should fall, not just your metabolic rate.
Morning types (larks): Natural bedtime around 9:30-10:30pm. With an average metabolism and a 200mg dose, the latest safe consumption time is roughly noon-1pm.
Intermediate types: Natural bedtime around 11pm-midnight. The 2pm rule is approximately correct for this group.
Evening types (owls): Natural bedtime past midnight. A 3-4pm cutoff may be fine.
The key variable is always time from consumption to intended sleep onset, not the clock time itself.
How to Find Your Personal Cutoff With Wearable Data
This is where the approach becomes genuinely useful rather than theoretical. If you wear an Oura ring or WHOOP, you have the data to run this experiment yourself.
The protocol:
- Pick a 3-week tracking window where your schedule is relatively consistent.
- Log the time and approximate dose of every caffeinated drink each day.
- Track your sleep score (Oura) or recovery score (WHOOP) daily.
- After 3 weeks, look for the correlation: on days when your last caffeine was X hours before sleep, how did your sleep score compare to days when it was X+2 or X-2 hours?
You're looking for the inflection point — the cutoff time where your sleep scores systematically improve. Most people find a meaningful pattern within 3 weeks.
Be the first to try Prova
We're building an app to track whether caffeine and sleep tracking actually works. Join the waitlist.
Practical Cutoff Times by Profile
Use these as starting estimates, then adjust based on your own data:
| Profile | Half-Life Estimate | Cutoff for 10:30pm Bedtime |
|---|---|---|
| Fast metabolizer, no medications | 3-4 hours | 1:30-3:00pm |
| Average metabolizer | 5-6 hours | 10:30am-12:30pm |
| Slow metabolizer | 7-9 hours | 7:00-10:30am |
| Average + oral contraceptives | 8-10 hours | 6:30-8:30am |
| Average + smoking | 3-4 hours | Same as fast metabolizer |
If you're a slow metabolizer, even "morning only" coffee may affect your sleep. Consider reducing total dose (switching from a 16oz drip to a 12oz, or cutting from two cups to one) rather than just moving the cutoff earlier.
Beyond the Cutoff: Total Daily Dose Matters Too
Optimizing cutoff time assumes your total daily dose is reasonable. High daily caffeine intake (500mg+) keeps your adenosine receptor sensitivity suppressed even when caffeine clears — leading to blunted energy during the day and disrupted sleep at night regardless of cutoff timing.
If your wearable data suggests poor sleep quality even after optimizing timing, total dose reduction may be the next variable to test. A 30-day reduction from 400mg/day to 200mg/day, tracked against sleep scores, is a worthwhile experiment.
What the 43-Year JAMA Cohort Suggests About Dose
A study published February 9, 2026 in JAMA from investigators at Mass General Brigham, Harvard T.H. Chan School of Public Health, and the Broad Institute pooled 131,821 participants in the Nurses' Health Study and Health Professionals Follow-Up Study and followed them for up to 43 years. Participants who consumed 2-3 cups of caffeinated coffee per day — roughly 250-300 mg of caffeine — had an 18% lower long-term dementia risk than near-zero drinkers, and a 35% lower risk in the age 75-and-younger subgroup. The protective association leveled off above 300 mg per day, and decaffeinated coffee showed no benefit.
This is observational data, not a randomized trial — it cannot establish causation. But the dose-response is worth noting for anyone running a long-term caffeine self-experiment: more is not better past roughly 300 mg/day for the brain-aging endpoint, and the upper FDA limit (400 mg) appears to offer no additional cognitive upside while raising sleep and cardiovascular costs. The decaf null result also argues that caffeine itself — not the coffee matrix — is doing the work, which matches the mechanism literature on adenosine A2A antagonism and neuroinflammation.
If you track caffeine timing in Prova, also log total daily dose. A 30-day window comparing your sleep-score and HRV trends across "above 400 mg/day" vs "250-300 mg/day" weeks gives you a personal read on whether you tolerate the higher end — without needing to extrapolate from population averages.