Sentiment Analysis
Scoring every incoming resident message from negative to positive with Snowflake's built-in AI, and turning the results into real-time alerts when a resident is particularly upset.
What I did
I looked at all the incoming conversations across our multifamily properties and used Snowflake’s native AI sentiment scoring to rate the content of each message, from negative to positive.
What worked, and what didn’t
This project was interesting because it used an essentially “out of the box” generative AI tool that Snowflake built.
- The extremes were spot on. Very negative messages were picked up well, and so were very positive ones.
- The middle ground was weaker. Scoring was less reliable for messages that weren’t clearly positive or negative.
Putting it to use
Even with that limitation, there are useful ways to apply it:
- Track negative messages against building performance. Count how many negative messages each property receives and see how that relates to how the building performs.
- Flag problems in real time. Use the scores in a Power BI report to flag a particularly angry resident as it happens, and notify everyone who needs to know.