StatReason

Cronbach's Alpha Calculator

Estimate how consistently a set of survey or test items measures the same underlying thing, using every respondent's answers across every item.

One subject per line, comma-separated item scores. Every subject needs a score for every item.

What this answers

Cronbach's alpha answers "do these items on my survey or test tend to move together, the way you would expect if they are all measuring the same underlying trait?" It is the most widely reported reliability statistic for multi-item scales, from personality inventories to customer satisfaction surveys.

How it is calculated

Alpha compares the sum of each individual item's variance across subjects to the variance of the total score (every item summed together per subject). When items are highly correlated with one another, the total score's variance grows faster than the sum of individual item variances, pushing alpha toward 1. When items are unrelated or contradictory, alpha can be low or even negative.

Worked example

For a two-item scale where every subject's second item score is exactly one point higher than their first (a perfect linear relationship between the two items), alpha comes out to exactly 1, since the two items are perfectly internally consistent with each other. Real surveys almost never reach exactly 1; values in the .70 to .90 range are commonly treated as acceptable to good.

Assumption audit

Calculated from your data: the number of subjects and items, each item's own variance, and the total score's variance across subjects.
Evidence to review: whether any single item looks out of step with the rest (check the individual item variances above). A reverse-worded item entered without being reverse-scored first will silently drag alpha down.
You must verify: that every item is genuinely intended to measure the same underlying construct. Alpha does not prove unidimensionality or validity; it only measures internal consistency among whatever items you provided.

What this result does not mean

A high alpha does not mean the scale measures the right thing, only that its items are internally consistent with each other. Alpha can also be artificially inflated simply by adding more items to a scale, even redundant ones, so it should not be treated as proof of quality on its own.

Limitations

Alpha assumes all items are on the same scale and (in this default, unstandardized form) the same units. If your items use different response scales, standardize them before combining, or note that this calculator's raw-score alpha will not be meaningful across mismatched scales.