Why we publish our survey thresholds
Every employee survey has a margin of error, a point below which a group is too small to report, and a test of whether its questions hang together. Most never say what those are. Culture Crest publishes them, and here is what each one protects.
By WCRC Culture Crest Research Desk
Method note · Evidence · 10 October 2026 · 3 min read
Key findings
- Survey items in each pillar must reach a Cronbach’s alpha of at least 0.70 before a pillar score is reported.
- Samples are sized for a precision of ±3 points at 95% confidence.
- No result is shown for a group of fewer than 10 respondents.
- Every reported gap and change is tested for significance at 95% confidence, with a correction for multiple comparisons.
A CHRO who signs off an employee survey is making three promises: that the results are accurate, that they are fair to every group compared, and that no individual can be identified. Each promise depends on a threshold. Culture Crest publishes its thresholds so that a CHRO can check them rather than take them on trust.
Accuracy: reliability, validity and precision
Reliability asks whether the questions meant to measure one thing actually do. Before a pillar score is reported, its items must reach a Cronbach’s alpha of at least 0.70. Validity asks whether the five-pillar structure holds in the organisation’s own data; it is tested by confirmatory factor analysis, with a CFI of at least 0.90 and an RMSEA of no more than 0.08. Precision asks whether enough people answered. Samples are sized so that a score is accurate to within ±3 points at 95% confidence.
Precision has a practical consequence leaders should know. A pillar score of 68 is best read as "between 65 and 71". A team that moves from 68 to 70 may not have changed at all. That is why every Culture Crest score is shown with its confidence interval.
Fairness: comparing groups properly
Comparing two groups is only fair if they read the questions the same way. Before any group comparison, Culture Crest tests for measurement invariance. Rushed, patterned and straight-line responses are screened out, and results are weighted where a level, function or location is under-represented. Every reported gap is tested for statistical significance, with a correction for the number of comparisons being made, and is reported with its effect size.
Anonymity: the minimum of ten
No result is shown for any group with fewer than 10 respondents. This is the threshold that makes honest answers possible. A survey that employees do not trust measures their caution, not their experience. Signing in to a survey only prevents duplicate responses; it is never attached to the answers. Consent notices and data processing agreements under the Digital Personal Data Protection Act, 2023 are in place before any data is collected.
| Standard | Threshold |
|---|---|
| Reliability | Cronbach's α ≥ 0.70 |
| Structural validity | CFI ≥ 0.90, RMSEA ≤ 0.08 |
| Fair comparison | Measurement invariance required |
| Precision | ±3 points at 95% confidence |
| Response quality | Flagged responses excluded |
| Significance | 95% confidence, multiple-comparison correction, effect sizes reported |
| Anonymity | Minimum 10 respondents |
| Qualitative coding | Cohen's κ ≥ 0.70 |
| Audit scoring | ICC ≥ 0.75 |
The full standards in the Workplace Analysis
How this was measured
The thresholds are those published for the Culture Crest Workplace Analysis and the Workplace Value Index. The example of a score of 68 read as 65 to 71 illustrates a ±3-point interval and describes no organisation.
Questions this story answers
What is a good reliability score for an employee survey?
Culture Crest requires a Cronbach’s alpha of at least 0.70 for the items in each pillar before a score is reported.
What is the minimum group size for anonymous employee survey results?
Culture Crest shows no result for any group with fewer than 10 respondents.
How precise are employee survey scores?
Culture Crest sizes samples for ±3 points at 95% confidence and reports every score with its confidence interval.
Cite as: WCRC Culture Crest Research Desk, "Why we publish our survey thresholds", WCRC Culture Crest, 10 October 2026. https://www.culturecrest.work/insights/why-we-publish-our-survey-thresholds
More from Culture Crest
Report · The Founding Edition · 10 October 2026 · 3 min read
The Founding Edition in numbers: what the public evidence says about 250 of India's largest employers
Equitable Opportunity scores highest and Psychological Safety lowest. Multinationals lead on four pillars; Indian companies lead on the fifth.
ReadAnalysis · The Five Pillars · 10 October 2026 · 3 min read
Psychological safety is the weakest pillar for 162 of India's 250 most loved workplaces
Across the Founding Edition, the ability to speak up without cost is the least evidenced part of workplace equity. Here is what the data shows and what changes it.
ReadAnalysis · The Founding Edition · 10 October 2026 · 3 min read
Indian companies lead on trust. Multinationals lead on transparency.
Trust & Belonging is the strongest pillar for 59% of the Indian companies in the 250, and for 3 of the 101 multinationals. The pattern on Transparency & Accountability runs the other way.
Read