Every support ticket we close ends with the same email: "How did we do?" followed by a thumbs-up and a thumbs-down button. This survey, called CSAT, is treated internally as a direct measure of support quality. We'd like to walk through, honestly, what it's actually measuring, because it is rarely the thing anyone presenting the number means.
The survey fires on reply, not on resolution
A support agent replies to your ticket — "try clearing your cache and let us know" — and the CSAT survey fires immediately, attached to that reply, before anyone knows whether clearing the cache worked. If it didn't work, and you have to write back a second time, the second reply gets its own survey. Your problem can remain completely unsolved across four replies and four surveys, and the dashboard will show four separate interactions it can rate, none of which were asked to rate "is my problem fixed," because that was never the question on the button.
The thumbs-up is rated on tone, not outcome
Most people click thumbs-up when the agent was warm, fast, and apologetic, regardless of whether the actual issue got resolved. A kind, well-written "I'm so sorry, I don't have a fix for this yet, but I'm escalating it personally" reliably scores higher than a terse "fixed, see attached." CSAT, as actually clicked by actual humans in actual inboxes, is measuring the agent's bedside manner far more than the thing bedside manner is supposed to be in service of.
The incentive structure built on top of it
Agents are measured on CSAT score, individually, which means agents are incentivized to write replies optimized to be liked in the sixty seconds after they're sent — warm, apologetic, personally invested — rather than replies optimized to actually close the loop, which sometimes requires saying something less pleasant, like "this is a known limitation and won't be fixed this quarter." The honest reply scores worse. The agent knows this. The agent writes the warmer reply instead, and the ticket stays open a little longer so the warmth has somewhere to land.
The non-response bias nobody corrects for
Response rates on these surveys hover around 5-10%, and the people who respond are disproportionately the extremely happy and the extremely furious — the two groups motivated enough to click anything. The large, silent majority whose problem was fine, not great, just fine, never responds, which means the final number is an average of two shouting minorities, reported upward with the confidence of a representative sample.
What a high CSAT score actually certifies
Mostly, that your support team is pleasant to interact with, which is genuinely valuable and genuinely worth measuring — just not the same thing as "your product has fewer problems" or "your problems get solved faster." A team with constant, serious, unresolved bugs can run a very high CSAT score indefinitely, as long as every agent is kind about it, which should worry anyone using the number as a proxy for product health rather than for interpersonal warmth.
What we're not saying
We're not saying CSAT is meaningless — tone and responsiveness genuinely matter to how a bad experience feels, and a team that's warm under pressure is doing something real and valuable. We're saying a single thumbs-up/thumbs-down button, fired per-reply instead of per-resolution, measures a much narrower thing than "customer satisfaction" as a phrase implies, and treating it as a stand-in for product quality lets real, unresolved problems hide behind a genuinely nice tone.
Our recommendation
Fire the survey on ticket close, not on every reply, and ask two questions instead of one: "was your issue resolved" and "how was the experience," scored separately. The gap between those two numbers is the real story — and right now, that gap is exactly what the single button is built to hide.
If your support dashboard currently shows a 94% CSAT next to a bug that's been open for six months, email us at contact@b2bsaas.guru. We'll help you build a better survey. We can't promise the number it produces will look as good in the board deck.