Free tool
CSAT drop checker. Is the change bigger than chance?
Enter satisfied and total responses for two periods, by category or contact reason. The checker tests whether the change is bigger than chance, gives the 95% interval of the change, and shows which categories moved and whether category mix explains it.
One row per category, contact reason, channel or region. Enter the number of satisfied responses and the total number of responses. One row is enough to test the overall change.
Your result
Each category's contribution to the overall change, in CSAT points: its average share of responses times its change in CSAT.
p-value: two-proportion z-test with a pooled standard error. 95% interval: normal approximation with each period's own standard error. Mix and within use average shares of the two periods, so they add up to the total change.
Everything is calculated in your browser. The calculator makes no network calls and saves nothing: close the tab and your numbers are gone. The site's usual analytics record that you visited this page, but the numbers you type and the results are masked from them.
Run this on all your data
Tag every ticket, chat and survey comment with its contact reason and root cause, and see which issue moved CSAT, not just whether it moved.
Plans from $299 a month. 14 days free. Card at checkout. Cancel before day 14 and pay nothing.
What the checker does
CSAT is the share of responses that count as satisfied. When it falls from one period to the next, two questions come before any root-cause work. Is the drop bigger than the normal wobble you get from sampling? And did customers in some category get less satisfied, or did more responses come from a category that always scores lower? The checker answers both from counts you already have.
The formula in plain words
- Is it bigger than chance? A two-proportion z-test. Pool both periods to get the CSAT you would expect if nothing changed, work out how much the gap between two samples of these sizes would vary by chance, and divide the real gap by that. The p-value is the chance of a gap at least this large if CSAT had not really changed.
- How big is the change? The 95% interval is the change plus or minus 1.96 standard errors, using each period's own CSAT and response count. If it does not include zero, the p-value is below about 0.05.
- Which categories moved? Each category's CSAT change gets its own test, and its contribution to the total is its average share of responses times its change in CSAT.
- Does category mix explain it? The mix effect is each category's change in share times its average CSAT. Mix plus within adds up to the total change, the same shift-share split as in our NPS mix-shift post.
With the example data loaded above (made up for this page), CSAT falls from 79.4% to 76.4%, a change of -3.0 points with a p-value of 0.003. Almost all of it is Delivery, whose CSAT went from 76% to 67%. Category mix moved the total by -0.1 points.
When to use it
- Before you escalate a CSAT drop in a weekly or monthly review.
- When a team or category has few responses and you are not sure the move is real.
- When the routing, survey trigger or channel mix changed between the two periods.
Limitations
- The test assumes each response is independent. Many responses from the same customer, or surveys sent in bursts, make the real uncertainty larger than shown.
- Testing many categories at once means some will look significant by chance. Treat a single category with p near 0.05 as a lead, not a finding.
- The normal approximation is rough when a category has very few responses or CSAT near 0% or 100%.
- A significant change says the number moved, not why. The reasons are in the comments and tickets behind it.
Common questions
How do I know if a CSAT drop is significant?
Compare the two periods with a two-proportion z-test: the share of satisfied responses before and after, and how many responses each share is based on. A p-value below 0.05 means a gap this large would rarely appear by chance if CSAT had not really changed. The checker also shows the 95% interval of the change.
What counts as satisfied?
Use your survey's own definition, then keep it the same in both periods. On a 1 to 5 scale, a common choice is to count 4 and 5 as satisfied.
Can I use weeks instead of categories?
Yes, with a caveat. To test the overall change, add up each period's weeks and enter one row per period total, or one row named All. To see which part of the business moved, use rows that exist in both periods, such as contact reason, channel or region.
What does category mix mean here?
CSAT is a weighted average of each category's CSAT. If more responses come from a category that always scores lower, overall CSAT falls even if no category changed. The mix effect measures that part.
Is my data sent anywhere?
No. The checker runs in your browser and makes no network calls. What you type is not sent or saved, and your inputs and results are masked from the site's analytics and session recordings.
Ask why the number moved. On every response.
Start with one file. Ask why a number moved. Plans from $299 a month, no demo and no contract on monthly plans.