CompoundBrief

Research literacy

Why “It Worked for Me” Isn’t Evidence: Personal Accounts vs. Trial Data

Published October 15, 2026 · 7 min read

Scope: This article is about how to evaluate evidence, not about any compound or product. It discusses why personal accounts and trial data answer different questions. It does not recommend, endorse or describe any use.

Online, a personal account is often the most persuasive thing a reader sees: a before-and-after, a first-person story, a thread of people reporting the same thing. It feels like evidence because it is vivid and specific. But a story and a trial are built to answer different questions, and the difference explains why researchers treat them so differently.

What a personal account can and cannot show

A first-person account shows that something happened to one person, who then described it. It cannot show that one thing caused the other. Many changes happen together: people alter several habits at once, conditions fluctuate over time, and expectations shape what gets noticed and remembered. A single account has no comparison group, so there is no way to tell what would have happened anyway.

The built-in distortions

Selection. People who feel they benefited are more likely to post, and people who saw nothing or had problems are less likely to. What is visible is a filtered sample, and the filter points in a predictable direction. Nobody posts a thread titled “nothing happened.”

Placebo and expectation. Believing a treatment will work can change how people feel and what they report, especially for outcomes that depend on self-assessment such as energy, mood or pain. This is a real, well-documented effect, and it is the reason trials compare against a placebo.

Regression to the mean. Measurements that are unusually high or low tend to move back toward average on their own. People often start something new at a low point, such as a bad week, a flare-up or a peak of concern, so a later improvement can be nothing more than ordinary fluctuation. Without a comparison group, it looks like an effect.

Natural history. Many conditions improve, worsen or cycle without any intervention. A story that begins at the worst moment and ends at a better one fits that pattern whether or not anything in between mattered.

Confounding and co-interventions. People who try one new thing often change their sleep, diet, exercise or stress at the same time. Any of those could account for what they notice.

Recall and survivorship. Accounts are written after the fact, shaped by memory and by the outcome. People who stopped early, dropped out or got worse rarely write the follow-up.

What randomized, blinded trials do about this

A randomized trial assigns participants by chance, so the groups start out alike on average, including on things nobody measured. A control group experiences the same expectations, fluctuations and natural history, so only the difference between groups can be credited to the intervention. Blinding keeps expectations from steering what participants report or what assessors record. Pre-registration of the primary outcome limits the temptation to look for a favorable result afterward. Reporting standards such as CONSORT ask for a full accounting of who was enrolled and who dropped out, which counters the survivorship problem. For how to read the numbers that come out, see our guide to reading a clinical trial result.

Trials are not perfect either

None of this makes every trial reliable. Small trials give imprecise results, sponsors may select favorable endpoints, and a result from one population may not carry over to another. Animal and cell studies, which make up much of the early literature on newer research compounds, cannot be assumed to predict outcomes in people. The point is not that stories are worthless and trials are flawless. It is that they differ in how well they can separate cause from coincidence, and the evidence for any claim should be weighed accordingly.

Questions to ask of any claim

Sources

← Back to all articles