[ayanb]

In this case, the primary category for Intercom is `User onboarding and Engagement` which can be broken down separately to `User Onboarding` and `In-app Support` or `In-app Chat`. That will suddenly spike up the sanity of both categories (maybe we can even normalize further - siftery has more than 700 categories/tags, a first for any such platform).

Categorisation of a data set like this can be non-trivial and will false positives. We can cherry-pick such a false positive and ignore all the actual positives. That said, thanks for this feedback - this is exactly what is needed to improve status quo. We are on it and will report back in a few hours with the category split I suggested in the previous paragraph. Do let me know if you think it can be broken down into a different type of granularity?