Our methodology
How crowd-reported outcomes become the scores you see on this site.
1. Where the data comes from
Every number on Works for Us starts as a personal report from a signed-in user. For each report they tell us:
- the compound they took (and the dose, where known)
- the condition they took it for
- how well it worked for them, on a 1-5 scale
- any side effects they noticed
- whether they would recommend it to someone else
Reports go through moderation before they count towards public figures. We remove spam, and we never verify anyone's diagnosis or medical records. This is shared experience, not a clinical study.
2. When a score is shown
A compound-condition pair only shows an aggregate score once it has at least 10 approved reports. Below that, we show a "not enough reports yet" state instead of a score. Small samples mislead - a single glowing report is not evidence, and we refuse to present it as if it were.
3. What the score means
The headline score for a pair is the average (mean) effectiveness rating from all approved, non-deleted reports for that compound-condition pair. It is a measure of crowd-reported experience. It is not a measure of clinical effectiveness, and it never claims that a compound treats or cures anything.
An "adjusted score" that accounts for report quality and volume is planned. It is not in use yet - when it ships, this page will describe exactly how it works. We would rather show you a plain average we can defend than a clever number we cannot.
4. Ranges and uncertainty
Alongside the average we show a range (a 95% confidence interval). Think of it as the honest spread around the average: the true community experience most likely sits somewhere inside that range.
The range depends entirely on how many reports exist - ranges narrow as reports accumulate. A pair with 12 reports will have a wide, cautious range. A pair with 500 reports will have a tight one. If you see a wide range, treat the average with proportionate caution.
5. Deletions and updates
Every time a report is added, edited, withdrawn or deleted, the aggregate for that pair is recomputed from scratch from the current set of approved reports. Nothing is cached beyond the moment of calculation, so a deleted report stops counting immediately - this is also how we honour GDPR erasure requests for aggregate data.
6. What this data cannot tell you
- Self-report bias. People who had a dramatic experience - very good or very bad - are more likely to report than people who noticed nothing.
- No clinical verification. We do not verify diagnoses, doses, or outcomes. Reports are personal accounts.
- No placebo comparison. Without a control group, some of any reported benefit may be placebo effect, regression, or coincidence.
- Small subgroups. Even at 10+ reports, a pair can be dominated by a handful of very similar people, which may not generalise to you.
Use these figures to form questions, not decisions. Bring them to a qualified clinician - see our safety guidance.