What is Cronbach's Alpha?
Cronbach's alpha is a statistic that measures the internal consistency of a group of survey items — in other words, how well they hang together as a single scale. It's used to decide whether it's reasonable to combine several questions into one summary score.
Cronbach's alpha compares the variance of each item with the variance of the total score to estimate how closely related the items are. Values range from 0 to 1: higher values indicate that items tend to move together. Common interpretation guidelines treat values around 0.7 as acceptable for group-level measurement, 0.8+ as good, and very high values (e.g. >0.95) may indicate redundant items. Alpha assumes the items measure a single underlying construct and that item errors are uncorrelated — it does not prove unidimensionality or validity. It is sensitive to the number of items and their variances, and can be affected by translation or wording changes.
Usage example
You create a 5-item parent satisfaction scale on Hearo. After collecting responses, you compute Cronbach's alpha and get 0.82 — this suggests the items form a reliable scale and you can report an average satisfaction score. If alpha had been 0.56, you'd inspect item-total correlations and translations to find which questions are inconsistent and revise them.
Practical application
Cronbach's alpha helps you decide whether to combine several questions into a single score (for reporting or analysis) and flags items that may be confusing or off-topic. For multilingual surveys, it is especially useful: a change in alpha after translation can reveal wording that breaks consistency in a particular language, so you can target reviews and edits rather than retranslate everything. Use alpha alongside item diagnostics (item-total correlations, alpha-if-item-deleted) and simple factor checks — it guides practical decisions about survey quality, not a definitive judgment of validity.
FAQ
What value of Cronbach's alpha is considered 'good'?
There are no fixed rules, but typical benchmarks are: ~0.7 acceptable for basic research or group comparisons, 0.8+ good, and >0.9 excellent — while values above ~0.95 may mean items are nearly identical and redundant. Context matters: shorter scales and exploratory work often tolerate lower values.
If Cronbach's alpha is low, what should I do?
Check item statistics: look at item-total correlations and the 'alpha if item deleted' values to find problematic questions. Review wording and translation for those items, consider whether items are measuring different concepts, and possibly remove or rewrite items. Running a simple factor analysis can help detect multiple underlying constructs.
Can I use Cronbach's alpha with Likert (ordinal) responses?
Yes — Cronbach's alpha is commonly used with Likert items. Technically it assumes interval-level data, so for short scales or strongly ordinal data you might prefer ordinal methods (e.g. alpha based on polychoric correlations) or complement alpha with other checks.
Do I need to recompute alpha for each language version of a survey?
Yes. Translation or cultural differences can change how items relate to each other. Recomputing alpha per language helps identify language-specific issues so you can target wording fixes rather than assuming one set of items behaves the same everywhere.