Diagnostics as a Service and what a supplement retest can prove

Does this product do anything a customer can measure? A supplement brand that buys Diagnostics as a Service can answer with a before and after pair of measurements. That pair is the most persuasive thing the brand will ever put in front of a customer, though it is also one of the easiest to read too much into. A second value that is better than the first is evidence about one person on two days, so the interesting work is deciding how much of the difference is the product and how much of it is everything else.

What can a supplement retest prove for one customer?

None of this is legal advice, so a lawyer working in food and competition law should review any claim you build on a measured result.

A retest can show that a named customer's measured value on the second day was different from the value on the first day, by an amount that is larger than the noise of the measurement. That is a real and useful fact for that person, but it is not a demonstration that your product caused the difference.

The gap between those two statements is where supplement brands get into trouble, because the second one is the sentence the marketing team wants. A single customer with no control condition cannot separate your capsule from the four other things that changed in the same twelve weeks, such as more sleep, more daylight, a different diet and a resolved infection.

Doing this is still worth the money. A customer who can compare their own two numbers has something no competitor's label can give them, and honesty about what the pair means costs nothing while being wrong about it costs the brand.

Why does a second supplement measurement often look better?

Because of regression to the mean, which is the tendency of an extreme measurement to be followed by a less extreme one for statistical reasons alone¹. Any single result contains a real value plus random variation, so a person selected because their first number was unusually poor will often measure closer to their own average the second time with no treatment at all.

This matters more to supplement brands than to almost anyone else, because of how customers arrive. People buy an iron product after a low ferritin result, and they buy vitamin D in February after a low winter reading. Your customers are therefore selected on exactly the extreme values that regression to the mean will partly correct on their own.

• Selecting customers by a low first value makes some apparent improvement likely with no product involved.
• The effect is largest for markers with wide day-to-day variation, such as many nutrient measurements.
• Averaging two baseline measurements before starting reduces the problem, though it costs a second sample.
• A control group is the only clean answer, which is why studies have one and why marketing rarely does.

How much of a change is the laboratory rather than the customer?

Every analytical method has its own imprecision, so the same tube of blood measured twice does not give exactly the same number. Method differences are larger still, because two platforms measuring the same sample for ferritin, vitamin D or a thyroid marker can report values that differ by more than the change a customer is hoping for.

Some markers have international standardisation behind them and some do not, which means the size of this problem is different for every line on the report. A brand that changes laboratory, method or collection type between the first and second test has introduced a difference it cannot separate from the customer's own biology.

The accepted way of handling this is to work out how large a difference has to be before it counts as real, using the published imprecision of the method together with the person's own day-to-day variation². Below that threshold, a change is not distinguishable from noise, however satisfying the two numbers look beside each other.

How much of a change is ordinary biology in a supplement retest?

Every marker moves within one healthy person from week to week, and for some markers that movement is large. Reporting a difference without knowing the ordinary range of movement for that marker is how a brand ends up claiming credit for a Tuesday.

• Vitamin D varies with the season, so a spring retest after a winter baseline is measuring sunlight as well as capsules.
• Ferritin rises during inflammation, which means a recent cold can raise it while iron status is unchanged.
• Creatine kinase rises after hard training, so a sample taken two days after a heavy session is not a resting value.
• Cortisol and testosterone follow a daily rhythm, so the appointment time changes the number.
• Fasting status, hydration and recent alcohol all change several routine markers at once.

The published figures for how much each marker varies within one person are maintained by the EFLM in its biological variation database, and they are specific to the marker rather than general². Any retest protocol worth running fixes the time of day, the fasting state and the interval, because those three choices remove more noise than any amount of interpretation afterwards.

Why is one customer's result treated differently from a supplement claim?

Because a claim about a product speaks to everybody who reads it, while a measurement speaks about one person on one day. European law treats commercial communication about a food as a health claim that has to be authorised and listed, or used as one of the on-hold claims³⁴. A before and after pair from your own customers does not create an authorisation.

This is where the trouble starts for otherwise careful brands. An internal chart showing that many customers improved is an interesting reason to run a study, though publishing it as a product effect is a claim. A claim with no authorised entry and no on-hold entry behind it is one letter from a competitor's lawyer away from being expensive.

The safe and honest line is narrow enough to write down. You may tell one customer what their own values were and what changed, you may explain what the marker means, and you may not turn that pair into a general statement about what your product does for people.

What can a supplement brand do with aggregate anonymised data?

Aggregate data is legitimate for improving the product, choosing panels and generating a hypothesis worth testing properly. It becomes a problem when an uncontrolled internal series is presented as evidence of effect, because the design cannot support that conclusion whatever the sample size is.

• Ask for separate, explicit consent for research use, since health data has its own protection under European law.
• Anonymise properly rather than removing a name, because a small cohort with dates is often identifiable.
• Report the customers whose values got worse, because a series that only counts improvements is not a result.
• Describe the cohort honestly, including who was excluded and how many people never returned for the second test.
• Treat the output as a reason to fund a controlled study rather than as a substitute for one.

A brand that wants a defensible population claim eventually has to run a controlled trial with a protocol written before the data is collected. That is a real cost, so most brands decide against it, and the decision that follows honestly is to stop making population claims rather than to make them from customer averages.

How does Diagnostics as a Service keep a retest comparable?

The laboratory layer has to be the same on both days or the comparison means nothing, so the value of a partner here is consistency rather than novelty. One contract covers the analysis, the kits, the courier and the software, which means the second sample is handled the way the first one was.

• The same panel is available for repeat testing, and a trend view across repeated panels is in the customer dashboard.
• Six collection methods are supported, and a retest should use the same one as the baseline.
• Over 2,500 orderable parameters across eight modalities are available, so a panel rarely has to be rebuilt mid-programme.
• Analysis is performed by an accredited laboratory partner, which holds ISO 15189 accreditation and operates under RiliBAEK.
• GDPR terms and an AVV are part of the same partner contract.
• Panels are built with practising clinicians and researchers, and the protocols are reviewed by an advisory board.

A change of method is sometimes unavoidable, so the honest handling is to say so on the report rather than to let the customer read a method change as a change in themselves. Book a 30-minute demo if you want a measurement layer that is the same in month six as it was in month one.

Notes and sources

Last updated: 31 August 2026

1. Barnett AG and others, Regression to the mean: what it is and how to deal with it, International Journal of Epidemiology, volume 34, 2005

2. European Federation of Clinical Chemistry and Laboratory Medicine, Biological Variation Database, searchable per marker. EFLM Biological Variation Database homepage

3. Regulation (EC) No 1924/2006 on nutrition and health claims made on foods, EUR-Lex

4. EU Register of nutrition and health claims made on foods, European Commission