Why I built Distil.

A supplement report business that doesn't sell supplements.

I tried to sort out my own supplement stack a couple of years ago. I had read enough research to know what mattered, what the forms were, what the doses should be. What was actually for sale didn't match. The retailers didn't tell you what to take or what to avoid. The brands made claims their citations didn't support. The cheap supermarket products were the wrong forms at sub-therapeutic doses. The premium products were the right forms at twice the price they needed to be.

So I went looking for a personalised service that could close the gap. Someone to look at my profile, name what I needed, name what I should skip, and trust me to buy it. Every personalised service I tried was a sales funnel. Fill in a questionnaire, get a slick report, end up paying for a monthly box. The personalisation was packaging. The product was the subscription.

And once you see that, you can't unsee it. The longer the stack, the bigger the box. The bigger the box, the bigger the subscription. The report follows the money. Diet rarely gets a substantive section, because food doesn't ship in a sachet. So I built Distil.

The pattern is structural, not cosmetic.

The reason these services felt off wasn't poor taste. It was the business model bending the output. Three patterns repeat across the sector, and they are not accidents:

Personalisation coupled to subscription. A monthly box arrives because that is the revenue mechanic, not because monthly delivery is clinically useful. A capsule's optimal dosing schedule has nothing to do with billing cycles. But the questionnaire is the funnel into the subscription, so the personalisation is shaped by what the subscription needs to look like, not by what the evidence says.

Stack inflation as the path of least resistance. Every additional compound is more revenue. Honestly removing items from someone's stack hurts the P&L. So reports drift toward inclusion. A guide that recommends thirteen things and rejects two will outsell one that recommends six and rejects nine, even if the second is the better recommendation. Reports follow the money.

Food and blood work disappear. Substantive dietary analysis is rare because food doesn't ship in a sachet. Blood work that might mean fewer compounds (test ferritin before recommending iron, test 25-OH vitamin D before assuming a deficient dose) is also rare, because confirming you don't need a supplement reduces the order. The information that would help you most is exactly the information that hurts the funnel most.

None of this requires anyone in the industry to be acting in bad faith. It only requires a business model where stack length and compound count drive revenue. Once that is the model, every report tilts in the same direction.

The fix has to be structural too.

You cannot solve a structural conflict of interest with better intentions. So Distil has none of those structures. We do not stock supplements. We do not earn commission on anything we recommend. We have no preferred retailer. The business model is one questionnaire, one report, paid once. The report's incentive is to be the best report, not the gateway to anything else.

Where the report tells you to buy magnesium glycinate, we give you UK search terms so you can buy it from any retailer who sells it. We earn nothing from your purchase, regardless of where you buy. The report recommends the work; it does not channel the inventory. The single sharpest sentence I can write about what Distil is: a guide that recommends everything has not selected anything. The recommendation is the selection. The exclusions are the proof.

The considered list is sometimes longer than the recommendation list. That asymmetry is the product.

What "evidence-based" actually means here.

"Evidence-based" is a marketing word in this industry. To anchor it back to something checkable, here is the architecture behind every Distil report:

Ninety-eight compounds in a working database. Each graded A, B, or C against the published research, by published rules. Grade D is reserved for compounds we have considered and rejected; Grade D never appears in a report. The full grading rules are at distil.health/about/methodology.

Hard exclusions, not "use with caution". Around forty rules where a compound is off the table by mechanism for a specific client. Iron without a confirmed-low ferritin: drop. 5-HTP on an SSRI: drop, because of serotonin syndrome risk. NMN with any cancer history: drop, because NAD+ raising compounds may support cell proliferation. Curcumin on tacrolimus: drop, because of CYP3A4 induction interfering with calcineurin-inhibitor levels. These are not soft warnings. The compound does not appear in the stack.

Two-pass generation, with an independent safety check. The report is written in stages: a primary model selects the compounds and writes the recommendations against the locked profile; a separate model audits the output against the safety checklist. If the audit fails, on dose, on exclusion, on missing GP review notice, on banned-claim language, the report is held for manual review. It does not deliver. This has happened. Reports that were technically deliverable but visibly imperfect have been rebuilt before delivery.

Public quarterly calibration. The compound database is recalibrated every three months, with the log published. The April 2026 review changed thirty-three citations across seventeen parallel evidence audits and integrated seven new 2024–2026 RCTs and meta-analyses. Every change is itemised. The full log is open at /about/methodology.

"Evidence-based" should mean that the rulebook is published, the compounds are graded, the exclusions are absolute, and someone is checking the work. None of that costs a thing. Most of the industry doesn't do it because hiding the reasoning is how you sell more.

Inside the machine: what coordination actually looks like.

The clearest way to show what Distil does that most stacks do not is to look at three small pieces of arithmetic that appear in almost every report.

One: the magnesium glycinate / glycine arithmetic.

Magnesium glycinate is magnesium chelated to two molecules of the amino acid glycine. A 400mg elemental magnesium dose, the typical evening dose for sleep and HPA axis support, delivers approximately 2.4g of glycine as the chelated co-amino acid. This is not theoretical: the glycine is hydrolysed off the chelate during absorption and enters circulation as free glycine.

Standalone glycine for sleep onset is conventionally dosed at 3–5g before bed. The mechanism is partly through NMDA receptor modulation, partly through a measurable reduction in core body temperature ahead of sleep onset. If a report recommends both magnesium glycinate at 400mg elemental and standalone glycine at 5g, the actual cross-compound glycine load is closer to 7.4g. That is well above the dose the trials use, with diminishing returns.

Distil sets the standalone glycine dose at 3g when magnesium glycinate is in the same stack, because the combined cross-compound load of approximately 5.4g is the effective dose, not 3g and 5g independently. This is a small piece of arithmetic, but it never happens in a report that has not been consciously coordinated. It is the difference between a list of compounds and a stack.

Two: zinc, magnesium, and the divalent metal transporter.

Zinc and magnesium are both divalent cations. Both are absorbed in the small intestine partly via the divalent metal transporter 1, DMT1. They compete for the same transporter. Take both at the same meal, in supplemental doses, and you measurably reduce the absorption of both. The mechanism has been characterised in human studies for decades; it is not contested.

The implication is that zinc and magnesium taken together do not give you the dose that the label claims either of them is. So Distil places zinc at lunch and magnesium glycinate at bedtime, separated by enough hours that the transporter is not the bottleneck. This is not a timing preference based on convenience. It is a coordination decision driven by a specific transporter mechanism.

The same logic propagates through the rest of the stack. Phosphatidylserine is placed in the morning if there is a sleep goal, because evening dosing impairs sleep onset in a meaningful proportion of people. Rhodiola Rosea is morning-only because afternoon or evening dosing consistently disrupts sleep onset across the clinical literature. None of these are timing defaults; each is a placement decision driven by a specific physiological reason.

Three: sleep is not one problem.

The intake form distinguishes between two distinct sleep problems: difficulty falling asleep (onset), and waking during the night (maintenance). These are different physiological problems, addressed by different mechanisms.

Onset problems are best addressed by alpha-wave promotion (L-Theanine has the strongest evidence here, via the L-theanine-induced alpha rhythm seen on EEG within roughly 40 minutes of dosing), measurable core body temperature reduction (Glycine, again with the temperature mechanism), and GABA-A binding (Apigenin, with smaller but consistent evidence). Maintenance problems, by contrast, are usually a downstream consequence of HPA axis dysregulation: elevated evening cortisol that does not decline normally through the night. The mechanism-matched response is Ashwagandha, dosed in the evening rather than the morning, with a clinical literature showing measurable cortisol reduction across multiple RCTs of standardised extracts.

The same compound at the wrong time, or the wrong compound for the wrong subtype, is not personalisation. It is a guess with good packaging.

Why this matters at all.

The supplement industry runs on information asymmetry. The buyer cannot easily verify what the formula does, what form the active is, what dose the trials used, or whether they need it at all. Each layer of asymmetry is a margin opportunity for someone. The gap between knowing supplements matter and knowing what to do about them is exactly where most of the industry's revenue lives.

The Distil report is one way to close that gap, for one product category, for the people who are willing to pay once for a definitive read. The next thing we are building is a free tool: a rule-based interactions checker covering the top 250 UK medications against all 98 Distil compounds, with citations, plain-English translations, and a public methodology. No login, no LLM behind it, no upsell. Free for everyone forever, because giving the work away is what the work is.

If a piece of the industry can be made transparent, that piece should be made transparent. The customers find the people doing the transparent thing eventually. That is the bet.

"On your side" means we will always tell you what the evidence does not support as clearly as what it does.

What Distil is, and is not.

Distil exists to give you the report you would write for yourself, if you had the time to read the literature, the constraints to know what to safely combine, and the patience to track every cross-compound dose interaction. It is not a substitute for a clinician. It does not replace blood work where blood work is what you need; in fact, the report flags the blood work and tells you to do that first. It does not produce stacks for active cancer treatment, pregnancy, breastfeeding, or active eating disorders, because those situations need clinical care and not a supplement report.

What it does is replace the gap between knowing supplements matter and not knowing where to start, where to stop, what to skip, what to test before spending, and which forms are worth paying for. If that is the gap you are in, the report is for you.

A guide that recommends everything has not selected anything. The exclusions are the proof.