Seven milligrams. That’s the British Heart Foundation’s figure for the caffeine left in an average 200ml cup of decaf. Small, by any reasonable standard. Not zero, by any standard at all.
UK law caps the residual caffeine at 0.1% of dry mass for roasted decaffeinated beans, and 0.3% for soluble (instant) decaf. In a brewed cup that lands at roughly 2 to 15mg of caffeine, against 95 to 200mg in a regular cup. The British Heart Foundation does the arithmetic at around 93% less caffeine, not 100%.
For most drinkers those few milligrams disappear without consequence. For pregnant readers counting intake, anyone carrying the slow-metaboliser CYP1A2 variant, or anyone who’s lost a night’s sleep to a single black tea, the gap between “almost zero” and “actually zero” starts to matter.
No, decaf isn’t caffeine free
Every credible source agrees that a brewed cup of decaf retains a small amount of caffeine. UK and EU rules permit a residue. They do not require zero.
The residue is small. The British Heart Foundation cites around 7mg per 200ml cup of decaf, against 100mg for the regular equivalent. Holland & Barrett puts the range at 2 to 15mg for decaf versus 95 to 200mg for regular. Swiss Water, citing a 2006 University of Florida study, lands at 3 to 15mg per cup. The numbers move within that band because brewing strength, bean origin, decaffeination method and roast all push them up or down a few mg.
So: caffeine at roughly one tenth to one twentieth of regular coffee. Low enough that most drinkers will never notice it, and never quite zero.
How much caffeine is in a cup of decaf, really
The number depends on what cup you are holding.
| Type of decaf | Caffeine per serving |
|---|---|
| Brewed filter or cafetiere | 2 to 15mg per 8oz |
| Espresso decaf shot | 4 to 12mg per single shot |
| Instant or soluble decaf | 2 to 7mg per 3g serving |
| Chain-shop brewed (Starbucks brewed decaf) | 15 to 30mg per serving |
| Regular brewed coffee, for comparison | 95 to 200mg |
Worth flagging the chain figure. Caffeine Informer’s tested measurements put Starbucks brewed decaf at 15 to 30mg per serving, well above the BHF average. A decaf brewed weak at home and a decaf pulled strong at a coffee shop are different drinks. If the scenario is “I had a decaf at a chain before bed and slept fine”, fine. If the scenario is “I am counting milligrams”, that scenario is worth knowing.
Instant carries more variance than ground because most instant decaf does not declare its Arabica/Robusta blend, and Robusta has roughly twice the caffeine of Arabica. The 3mg per serving figure widely repeated is an average.
The UK and EU rule on decaffeinated coffee
This is the bit most of the front page of Google misses.
UK statute (S.I. 1987 No. 1986, Regulation 5(1)(b)) sets the threshold. To be sold as “decaffeinated” in the UK, residual anhydrous caffeine must not exceed 0.10% of the coffee-based dry matter for roasted beans, or 0.30% for soluble extracts. The underlying EU framework sits behind the soluble figure (Directive 1999/4/EC, which replaced Council Directive 77/436/EEC in 2000; Commission Directive 79/1066/EEC sets the analytical methods for caffeine determination). The 0.10% roasted cap is UK statute rather than a standalone EU roasted-coffee directive. Post-Brexit, the UK retained both thresholds in domestic law.
The American equivalent is looser. US guidance cites 97% caffeine removal as the working industry standard. The UK figure works out to 99.9% removal for roasted beans.
That is not a small gap.
Most explainer articles on the front page of Google quote the US figure as if it is universal. A UK reader on a UK search result gets a US answer. If you are buying decaf sold into the UK, the bean has hit a tighter standard than most blog posts on the subject would have you believe. See the word “decaffeinated” on the pack and you are already inside the 99.9% threshold.
Does the decaffeination method change the residual caffeine
A little, but probably less than you would think.
Four methods are in everyday use on UK shelves. Swiss Water decaf uses water and activated carbon, with no solvents at any stage. Sugar cane decaf uses ethyl acetate, a solvent that occurs naturally in fruit and is often produced from sugar cane fermentation in Colombia. CO2 decaf uses pressurised liquid carbon dioxide as a selective solvent. Chemical solvent decaf, the so-called European method, uses methylene chloride, and it is still the cheapest option common in mass-market instant.
All four must hit the same 0.1% legal cap for roasted beans. The threshold does not care which mechanism removed the caffeine. In third-party testing, water-process decaf (Swiss Water and the Mexican Mountain Water variant) tends to test at the lower end of the cup-level range, around 2 to 5mg, against 5 to 10mg for the solvent methods. A real gap, but a small one. Most drinkers will not taste the few mg, and will not feel them either.
The Decaffeinate directory currently lists 55 specialty decaf coffees from 52 independent UK roasters. Of the method-tagged listings, 18 are sugar cane, 11 are Swiss Water, 7 are CO2 and 6 are Mountain Water. Filter by method to see which roasters use which.
Who actually needs to worry about the leftover
For most drinkers, decaf at 2 to 15mg is functionally a non-event. Three cups of decaf still totals less caffeine than one short black.
A few groups should pay attention.
Pregnant or trying to conceive. NHS guidance is to keep total caffeine intake below 200mg per day during pregnancy. Decaf leaves a large headroom even at multiple cups, and the NHS lists decaffeinated tea and coffee among the recommended swaps. It is not zero, though, so it still counts towards the daily total. Follow NHS or midwife guidance over anything written here.
Severe caffeine sensitivity. Roughly 10 to 15% of people carry the slow-metaboliser CYP1A2 variant, the high-sensitivity ADORA2A variant, or both. For these drinkers, even 5 to 15mg can produce mild anxiety, palpitations or sleep disruption. If half a cup of regular tea is enough to notice, treat decaf as low-but-not-zero.
Late-evening drinkers. The foundational sleep-lab work on this (Karacan 1976) found decaf had no measurable effect on sleep disturbance. For the sensitive group above, sleep researchers tend to suggest 2 to 3 hours before bed as a sensible cut-off.
People on specific medications. Don’t take medical advice from a coffee directory. Ask your GP.
If you want the lowest possible caffeine, what to look for
A few rules of thumb that fall out of the research.
Choose water-process where possible. Swiss Water and Mountain Water consistently test at the lower end of the cup-level range, around 2 to 5mg per cup versus 5 to 10mg for the solvent methods. Not a guarantee on any individual cup, but the most consistent across batches.
Avoid instant if consistency matters. Robusta/Arabica blend ratios are rarely declared on instant packs, and Robusta has roughly twice the caffeine of Arabica. The averages are reassuring. The variance is not.
Don’t trust the “darker roast equals more caffeine” myth. Darker roasts burn off slightly more caffeine, not less. The difference is small but real, and the myth goes the wrong way round.
Filter the directory. The full Decaffeinate listing has filters for decaffeination method. If lowest caffeine is the priority, filter to Swiss Water or Mountain Water and start there.
The verdict
Decaf coffee is not caffeine free. UK law caps the residue at 0.1% by mass for roasted beans (99.9% removal) and 0.3% for soluble. In a cup, that lands at 2 to 15mg, against 95 to 200mg for regular coffee. The British Heart Foundation’s neat figure, 93% less caffeine, is the one to remember.
Yes if you are being loose. No if you are being precise. For the genuinely caffeine sensitive, the late-evening drinker, or the pregnant reader counting intake, water-process decaf gives the lowest reliable residue. For everyone else, the leftover is functionally a non-event.
If you’d like to see which UK specialty roasters are working with which decaffeination methods, browse the full directory. 55 coffees, 52 independent roasters, filterable by method, with current prices and live stock.