crystal Version 2.5 is now available. It comes with a wave of updates aimed to strengthen crystal’s usability, security and stability.
While most of this release package is focused on the behind-the-scenes enhancements, we added some quality-of-life improvements to make your data interaction even simpler.
Please find the updates below split by user type:
Hitting dead-ends in the conversation can be frustrating when you’re looking for an insight. That’s why we improved crystal’s disambiguation.
Now, in addition to this, if you ask for a topic that isn’t configured in your environment, crystal will suggest one that’s similar. So, let’s say you ask for “total sales” — a topic that doesn’t exist — but a similar topic like “average sales” does. crystal will then explain to you that “total sales” isn’t configured, and suggest “average sales” (or any other similar options) as an alternative.
- Filter management improvements in the Dashboard
- Filter grouping: Before, if you wanted to apply filters to the topics, you would have to sift through a long list of filters that apply to all sections of your Dashboard. Now, when you open the filter menu, you’ll only see filters that apply to the particular section you’re on. This means you can get to the filter you need faster, and avoid the risk of selecting an inapplicable filter.
- Filter persistence: Before, if you applied a filter/some filters in the Dashboard, they would disappear when you ended the sessions. Now, filters selected in previous sessions will be saved, and remain in future ones.
- Applied filters shown inside topic: Before, applied filters weren’t shown inside the topic on the Dashboard, raising the risk of the data being misinterpreted. Now, when a filter’s applied to a topic, it appears under the topic title.
Before, if a project had its own time-related filters (ones custom to their use case), they might clash with the default time ranges in crystal (in the last month, in the past two weeks, etc) and lead to errors in the conversation. Now, crystal fully supports any configured dynamic filters that relate to time (i.e., period, season, etc.) — resulting in less misclassification errors for members.
Before, training the NLP in the ‘Train’ step could block the admins’ progress in publishing the topic — causing delays, especially for more complex topics. This process now happens in the background, as part of the topic publishing process so that set up is smoother and less time-intensive.