Caffeine

Caffeine can help someone feel awake without removing the need for sleep. It blocks the signal of growing sleep pressure, and its effects can last for hours. That makes the amount people drink and the time of their last drink relevant when they report a short night (Magkos & Kavouras, Crit Rev Food Sci Nutr, 2005).

The WHI survey shows a consistent association: reported short sleep becomes more common across the five drink-count bands, from no caffeinated drinks on a workday to four or more. The answers cannot tell whether caffeine came before the short sleep or followed it. They do give a reason to consider the two together. The survey counted drinks—tea, coffee or energy drinks—not caffeine in milligrams, so the same number of drinks can mean different doses.

Short sleep rises with every extra drink#

Short sleep here means the two shortest answers to the survey's sleep-duration question: under five hours, or five to six. Across the 7,569 responses to that question, 30.7% fall in those two bands. Among responses that answered both the sleep question and the drinks question, the share climbs at every step.

Every band is higher than the one before it, with no reversal across the five. The ends are 16.8 percentage points apart: two in five responses at four or more drinks report short sleep, against one in four at none. The two-drink group, the largest of the five, is the first to sit above the survey-wide rate of 30.7%.

The survey cannot tell whether caffeine use came before short sleep or followed it. The answers describe the same response, but they do not establish cause. The Sleep page carries the full distribution of reported sleep duration and difficulty.

The timing results cover a selected group of 3,339 responses: almost all report at least two drinks a workday. They do not describe all caffeine users. The survey measures drinks, not caffeine dose. Figures count responses, including repeat submissions. Methodology.

Two drinks is the most common answer#

76.6% of the 7,211 responses report at least one drink on a workday. WHI 2025 put that share at 76.3%. For scale, a U.S. beverage survey of 37,602 consumers aged two and over, using seven-day diaries, found 85% having at least one caffeinated drink a day — a general-population, all-ages, any-day measure rather than the workday question asked here, but enough to place widespread caffeine use well outside one country (Mitchell et al., Food Chem Toxicol, 2014).

Two drinks is the most common answer, just ahead of one. Adding the first three rows, 6,129 of the 7,211 responses — 85.0% — report two drinks or fewer, including the 23.4% who report none at all. The group at three or more comes to 1,082 responses, 15.0% of the total, and 327 of those report four or more.

Three or more caffeinated drinks on a typical workday, as the person answering counted them, is the measure the rest of this page uses. It is a count of drinks. It is not a measure of caffeine in milligrams, of dependence, or of any clinical threshold.

That 85%-to-15% division is worth holding alongside the ladder above, because the largest single step in the ladder falls on the same line: 32.0% short sleep at two drinks, 38.9% at three, 6.9 points. The two highest rates in that table rest on 755 and 327 responses, respectively. The association runs across all five bands; its top end describes a minority.

Two in three finish their last drink at 3 PM or later#

The mid-afternoon band is the single most common answer. Add it to the evening one and 2,185 of the 3,339 responses — 65.4% — put the last drink at 3 PM or later, with 743 of them after 6 PM. Almost every response in this table also reports two or more drinks on a workday, so these four rows mainly describe people reporting at least two drinks, rather than caffeine drinkers in general.

That matters because of how late caffeine is still working. A controlled trial gave people 400 mg of caffeine at bedtime, three hours before bed, and six hours before bed; all three disturbed sleep against a placebo, and in the six-hour condition objectively measured total sleep time fell by more than an hour, while their own reports suggested they were unaware of the disturbance (Drake et al., J Clin Sleep Med, 2013). The authors concluded that the result supports avoiding substantial caffeine use for at least six hours before bed. The study did not establish a safe cutoff. A review of the wider evidence found that caffeine typically delayed falling asleep, reduced sleep time and worsened perceived sleep quality (Clark & Landolt, Sleep Med Rev, 2017).

Two things stop the trial transferring directly to this page. It used one fixed dose, while this survey counted drinks: the U.S. Food and Drug Administration describes 400 mg a day as about two to three 12-ounce cups of brewed coffee, and puts a single such cup anywhere between 113 and 247 mg, so a drink count is not a dose in either direction. And this survey recorded the band of the last drink rather than the gap before bed. A 6 PM drink is a different exposure for someone who sleeps at 10 PM and someone who sleeps at 1 AM.

In the paired answers, the two ends of the day separate.

One response in the after-6 PM group answered the timing question but not the sleep question, which is why this table shows 742 where the one above shows 743.

The gap between the earliest and latest drinkers is 10.1 points, 40.6% against 30.5%. But this is not a ladder like the first one. Read down the clock and the rates run 30.5, 39.0, 32.4, 40.6: the midday group, on the smallest base here at 344 responses, sits above the afternoon group that follows it. These four rows support a difference between the earliest and the latest drinkers — not a rate that climbs with the clock.

WHI 2025 reported the same direction: 41.2% poor sleep among those whose last caffeine came after 6 PM, against 35.2% among morning drinkers. Its sleep definition differs from this edition's.

Caffeine use climbs with age; three-plus use climbs faster#

Both tables cover ages 20 to 59 and the same 7,160 of the 7,211 responses. The other 51 gave an age outside that range, or one the form did not record as a usable whole number, so the youngest and the oldest respondents are in neither table.

Across the four age bands, the two measures move differently. Any caffeine rises 13.9 points, from 73.0% to 86.9%. Three or more drinks rises 16.1 points, from 12.3% to 28.4%, and keeps climbing in the 50s.

On the question of whether people drink caffeine at all, age separates these responses more than sex, work location or city does. Men and women report caffeine at 76.8% and 76.2%, on 4,917 and 2,269 responses: a 0.6-point difference. Work location separates 3.0 points — 77.4% working from the office, 76.3% hybrid, 74.4% from home. The five metros range 11.4 points, from 70.8% in Hyderabad (763 responses) to 82.2% in Delhi NCR (998), with Delhi NCR highest among these five survey samples. Cities are pincode-based reporting areas with unequal sample sizes; these rates describe the survey responses, not all residents.

Three or more drinks a workday is twice as common at high stress#

The survey asked people to rate how stressed they felt last week, from 1 to 10. Grouped into three bands, the drinks answers separate.

Those three bands hold 7,210 of the 7,211 responses; one gave a drinks answer and no stress rating. Each band is higher than the one before it, and the two ends are 10.6 points apart — 20.3% against 9.7%, twice the share.

Higher stress and three-or-more drinking occur together in these responses. The survey does not show whether one causes the other. The Stress page covers what people name as the source of that strain and what they do about it.

Heavy use is flat across most of the income range, then jumps at the top#

Every band below ₹25 lakh sits between 14.4% and 15.0%, none of them more than 0.6 points from the 15.0% across all responses. The break comes at the top: 18.2% at ₹25–40 lakh, and 25.5% at ₹40 lakh and above — about 70% above the all-responses rate, on 141 responses, the thinnest base in the table. 548 responses declined to give a band and report 13.5%; one left the question blank. Those eight rows and that one blank are all 7,211 responses.

Among the job functions with at least 100 responses, the highest and lowest rates are 17.2 points apart: 28.2% of the 124 responses in General Management / Strategy report three or more drinks, against 11.0% of the 2,133 in Engineering / Software Development.

General Management / Strategy is the highest of these twelve functions, which are the ones with at least 100 responses and which together hold 6,783 of the 7,211. "Others" is the survey's own catch-all option, chosen by 1,102 responses. The remaining 428 responses sit in smaller functions outside this ranking.

Income and job function do not establish seniority or working hours. The survey did not ask who runs a team, who is on call, or how long anyone's day is.