Evidence methodology
StackTerminal is built to be explicit about what the evidence does and does not support. We do not treat all supplement claims as equal. Study design, population fit, dose realism, replication, and outcome quality all affect how strongly we speak.
How we think about strength of evidence
We grade evidence at the record level, then surface that context on ingredient and stack pages. A well-run meta-analysis of randomized trials generally carries more weight than a small pilot trial, which carries more weight than observational data, mechanistic reasoning, or animal work alone.
We also care about whether the studied population matches the claim. A result in older adults, sleep-deprived athletes, or vitamin-D-deficient participants may not transfer cleanly to everyone.
What moves a claim up or down
Study design: Meta-analyses and randomized controlled trials usually outrank observational or mechanistic evidence.
Replication: A finding repeated across independent studies is stronger than a one-off positive result.
Population fit: We ask whether the people studied actually resemble the intended user or claim context.
Dose realism: Evidence is less useful if the effect only appears at unusual doses, forms, or protocols.
Outcome quality: Hard outcomes and direct biomarkers usually matter more than vague self-report alone.
Signal consistency: Mixed, conflicting, or fragile findings lower confidence even when some studies are positive.
How to interpret our grades
Strong support from multiple credible studies with reasonably consistent findings in relevant populations.
Meaningful support exists, but there may be fewer trials, smaller samples, some inconsistency, or narrower population fit.
Early or limited support. A signal may be present, but confidence is not yet strong enough for broad conclusions.
Highly preliminary, indirect, or weakly supported evidence. Use caution when interpreting claims at this level.
What we do not do
- • We do not treat ingredient popularity as evidence.
- • We do not assume a positive mechanism automatically means real-world benefit.
- • We do not present AI summaries as a substitute for linked studies and explicit uncertainty.
- • We do not give medical advice, diagnosis, or treatment recommendations.
Want the broader safety and scope details too?
See Safety →