[Beta] Anomaly detection
- We have launched a beta version of anomaly detection for business metrics! Using anomaly detection, customers can quickly identify unexpected increases or decreases in a metric.
- Anomalies are determined by first calculating the range of expected values for a given metric for each day (or week). The metric value is flagged anomalous for data points where the metric value was outside of the range of expected values.
- Anomaly Detection is included as part of our Trend Detection analyses. No special configurations needed.
- Multiple anomalies in a row may mark the beginning of a new trend that may take a few more data points (days or weeks) to be confirmed as a true trend change vs. just an anomaly.
- To learn more, see the following video: https://www.loom.com/embed/ba79c41d4b4b4d869cddca0b364ba6d6
[Beta] Unified Workflows for Key Driver Analyses
Customers using the “High Accuracy” Top-Driver model can now see statistically-selected “top driver” subgroups alongside all other subgroups, without needing to run two separate analyses to do so. i.e. a unified view of analyses!
Subgroups identified as statistically significant Top Drivers are highlighted with a blue indicator, where-as all other subgroups (with lower statistical significance, but some impact to the metric) are shown without this indicator.
This feature is presently in Beta. If you would like to give it a try, reach out to your Sisu Customer Success contact.
Global Search — We are introducing Global Search, a simple search bar that lets you search across projects, metrics, dashboards, and analysis from anywhere in the product. Just click on the “search” icon on the left navigation, or press “cmd/ctrl + k” to initiate, and navigate to what you are looking for in one search. To learn more, see the following video: https://www.loom.com/share/3cad750bf655404fbd24ea7e4db90d51
- Trend Detection
- We now automatically default the time filter window to the available date range in your underlying dataset.
- Any subgroups with a size (i.e., number of rows) greater than 80% are now included in the evaluation of trends. Previously we only looked at those below 80%. This change ensures that even large / obvious subgroups are present in your results for you to evaluate trends and anomalies.
- For Sum/Count metrics, when clicking on “why did this trend change happen” we now set the data-range for the Key Driver (”Why”) Analyses to be of equal number of days on each side of the trend change date. This ensures that the resulting analyses has even days for each comparison period and is thus more accurate is helping you identify the drivers of the trend change. (Note this issue does not exist for Average/Rate type metrics)
- Product Experience enhancements
- Introducing Right Click to Open: Project list, analysis list, and dashboard lists now have a standard right click so you can right click and open the analysis or project or in a new window or a tab, without leaving the original page you are on.
- Introducing Bulk Delete: You can now delete multiple analysis together and keep your projects clean with the new “bulk delete” feature.
- Improved Sorting: Project lists sort in the same default order across the app so that you can view the same list and order everywhere.
- Improved error messages for various data types in the application to give you more context on errors.
- We have centralized the “metric deletion” to the “metric library” only, so that you do not accidentally delete a metric by mistake from other pages.
- 90% faster load times of all pages that have a list of tables, queries, or dimensions in them from your data-warehouse!
- Metric Units & Decimal Places
- You can now set the units and decimal places for any Metric on the Metric setup page. The units show up on Key Driver Analyses, Trend Detection, and Dashboards, making it easier to consume and understand analyses results.
- Key Driver Analyses APIs
- Analyses results now include “impact” for each subgroup in the API response for Key Driver Analyses. With this update, all statistics in the “Top Drivers” table are now included in the API response
- Updates to pysisu
- Our python library, pysisu, has been updated. Now you can run analysis with the run() method, allowing customers to embed KDA executions directly within a data pipeline. Impact score has also been added to KDA results to complete the export of all fact table information
- Predictions [Beta, Invite only]
- Improved predictions table formatting based on the type of metric the predictions are for.
- Key Driver Analyses
- Fixed SQL Server, Big Query and SAP Hana Date Truncations not bucketing days from Monday to Sunday.
- Gracefully handle connection errors to customer data warehouse for Filter drop-down values
- Fixed an issue where the “Christmas tree” chart in the Fact Detail Page was broken for percentile bin columns
- Trend Detection — Fixed handling of timestamps in Redshift
- SQL Editor — Fixed bug that wasn’t saving new queries in the query editor.
- Dashboard Filters [Invite only Beta]
- Fixed an issue where our latest workflow results were not correctly assigned to filtered KDA
- Fixed exploration filter when there's only one default expression
- Fixed series and result loading for non-board sliced explorations