CX Score

This was the segment the show kept circling back to, and it is the one worth your time this week.

Start with Madelyn's reframe, because it is what turned the conversation around. A post-ticket survey only hears from people motivated enough to answer, and in practice that means the unhappy ones. It is the same reason online reviews cluster at five stars and one star with nothing in between. A score calculated automatically on every closed ticket reads the whole population instead of the self-selecting slice. That is a real improvement over a health score you configured yourself, and it carries less of your own bias than one you would have built.

Now the caution, and this is the part to act on. The signal list is lopsided. There are a couple of positive factors against roughly a dozen negative ones. An AI reading tone against a list shaped like that will miss positives and over index on negatives, the same way people do. There is a related asymmetry underneath it: a good ticket is often just one that got resolved fast, which reads as the absence of a negative rather than the presence of a positive. Some of the classifications look off on their face, too. Positive future intent and advocacy landing as neutral factors is a strange call.

So do not take the number and run with it. Put your own business context into the portal, the terms and moments that actually signal something for your business, before you let this score drive a decision. That kind of transcript management is already available to you, and it is what makes this worth having.

One more thing worth saying plainly. This is easier for support than it will ever be for lead scoring. Tickets are conversations, so there is tone and language to read. Traditional lead scoring has none of that, because the whole job is deciding who to start a conversation with, before any conversation exists. Intent signals and unstructured social data are starting to close that gap, but it is not closed yet.

What to do: if you run support on Service Hub Professional or Enterprise, get into the beta. Then set your exclusions and your business context before you read a single score as truth.