The number that got my attention was fourteen.
That's roughly how many percentage points of sleep efficiency a small randomized trial reported gaining in adults with chronic insomnia after a course of capsule-delivered donor microbiota — the fecal transplant insomnia result that circulated widely this year. Fourteen points is not a rounding error. Against an eight-hour window in bed, it's the difference between spending an hour and forty minutes awake and spending under an hour.
The verdict: the trial is real and worth reading, the effect is larger than anything I produced in my own kitchen, and it is still not a reason for you to pursue a stool transplant. I spent eight weeks running the closest grocery-store approximation and got about a fifth of the benefit.
The comparison, before the argument
| Approach | Evidence behind it | Effect on sleep | What it costs |
|---|---|---|---|
| Donor microbiota capsules | One small placebo-controlled trial, single site, under 100 participants | ~14-point gain in sleep efficiency, as reported | Unavailable outside research; unquantified long-term risk |
| Fiber + fermented foods (my n=1) | Controlled feeding work on diversity and inflammation; sleep a secondary outcome at best | 8 min faster onset, ~3 points of efficiency | ~$14/week, two weeks of bloating |
| CBT-I | Dozens of RCTs; first-line in clinical guidelines | ~19 min faster onset, ~26 min less wake after sleep onset in meta-analysis | Five to eight weeks of doing something you will hate |
| Moving the last coffee earlier | Dose-timing studies with polysomnography | 400 mg six hours before bed costs ~40 min of total sleep | One flat, unpleasant afternoon |
What most people do
The headline arrives and the response is almost always a purchase. A probiotic chosen because the strain count on the label is a large number. A mail-order gut test that returns a colorful chart and a list of foods to rebalance toward. Melatonin at three or five milligrams, which is many times what the pineal gland releases on an ordinary night.
What most people don't do is the thing with the evidence behind it. Cognitive behavioural therapy for insomnia is first-line in essentially every clinical guideline, and its central move — sleep restriction — requires spending two weeks more tired than you already are. Compliance is poor because the treatment is genuinely unpleasant, not because patients are lazy.
And most people, in my experience of asking, believe they aren't caffeine-sensitive. Drake and colleagues gave subjects 400 mg six hours before bed and measured with polysomnography: roughly forty minutes of total sleep lost, in people who reported no subjective effect. Self-report and sleep architecture disagreed, and sleep architecture is the one that was right.
So this is the ground the microbiome headline lands on: a reader who has tried five supplements, hasn't tried the therapy, and is drinking a 3 p.m. coffee they've decided doesn't count.
What the evidence suggests
Read the trial and the picture narrows. Randomized, placebo-controlled, capsules rather than the older colonoscopic delivery, participants with clinically diagnosed chronic insomnia, sample under a hundred, one centre. Within those bounds it held: sleep efficiency climbed substantially, time awake after falling asleep dropped by a meaningful fraction of an hour, and — the detail that makes it worth taking seriously — the gains tracked with measurable shifts in recipients' gut communities toward their donors'. An outcome improving alongside the thing the intervention was supposed to change is a different class of finding from an outcome improving on its own.
What it doesn't establish is mechanism, durability, or generalizability. The candidate explanations — short-chain fatty acids modulating inflammation, vagal afferent signaling, altered availability of tryptophan as a precursor — are plausible and none were demonstrated here. You will also see it argued that the gut makes most of the body's serotonin, therefore gut, therefore sleep. Peripheral serotonin doesn't cross the blood-brain barrier; that syllogism doesn't work. The tryptophan route might, but it's a slower and more careful argument than the one usually made.
The procedure also isn't risk-free. The FDA issued a safety alert in 2019 after a patient died from a drug-resistant organism transmitted in donor stool. Rare, from a screened supply — but exactly the kind of rare that ends any conversation about improvising at home.
The adjacent evidence, the part you can act on, is thinner but real. A ten-week controlled feeding study at Stanford found a high-fermented-food diet increased microbiota diversity and lowered inflammatory markers, while the high-fiber arm moved immune signaling without moving diversity. Sleep was not an endpoint in that work. Anyone telling you it was is extrapolating.
What I actually do
I ran it on myself for eight weeks, February into April, after two weeks of baseline.
The protocol: fiber from about 19 g/day to a 38 g/day target, food-logged, mostly legumes, barley, and leaving skins on. Two fermented servings daily — 200 ml kefir at breakfast, about 60 g kimchi or sauerkraut at dinner. No capsules; I wanted food I could see rather than a supplement I couldn't verify.
Caffeine I held constant rather than optimizing: two cups, roughly 190 mg, last one before 11 a.m., every day of all sixteen weeks. Moving two variables would have made the whole thing uninterpretable, and caffeine timing was a win I'd already taken years ago.
Measurement was a finger-worn tracker for efficiency and wake after sleep onset, plus a paper diary each morning for how long falling asleep took. The diary matters because trackers are weakest at precisely the number insomniacs care about most.
Baseline fortnight versus final fortnight:
- Diary sleep onset latency, median: 34 min → 26 min
- Tracker sleep efficiency: 84.1% → 87.2%
- Wake after sleep onset: 41 min → 38 min (noise)
- Weeks 1–2: worse, with three nights of real gas-driven wakefulness before it settled
Eight minutes off onset, three points of efficiency, and it held after the window closed. I'll take it. But I can't clean it. The experiment was unblinded, I knew what I was testing and wanted it to work, and February to April adds daylight at both ends of the day, which is itself a sleep intervention. Any of that could be the entire effect. One subject with a known expectancy bias produces a hypothesis, not a result.
The honest ranking: CBT-I beats caffeine timing beats fiber and ferments — and the transplant isn't on the menu.
Who this is for, and who it isn't
Worth doing if the fundamentals are already in place — fixed wake time, caffeine cut off by late morning, no alcohol nightcap — and you want a low-risk next lever. It costs about $14 a week, has upside beyond sleep, and the worst realistic outcome is two uncomfortable weeks.
Not for you if you have moderate-to-severe chronic insomnia and haven't done CBT-I. You'd be trading a treatment with a large replicated evidence base for one with an interesting first trial and a good story. That's a bad trade no matter how novel the science reads. Also not for you, at least not without a clinician, if you're immunocompromised, pregnant, or living with IBD — and unscreened donor material is off the table for everyone, permanently.
Fourteen points, then. It's a real number from a real trial, and it isn't mine, and it probably won't be yours — not this year, and not from anything on a shelf. I got three. The distance between three and fourteen is the honest measure of how early this field is: enough signal to make fourteen worth chasing in a properly powered replication, nowhere near enough to justify chasing it yourself.
