Using Smart Insights: The Basics

  1. Smart Insights are applied at the KPI level. By looking at both the dashboard's filter definition and row level security, Birst ensures that each user gets results that are personalized and relevant to their own use cases. From your KPI, click the Smart Insights icon.
  2. Your dashboard filters are applied to your KPI. You can edit or remove the filters before you start your analysis.
  3. From the Configure Insights page, add additional columns for Birst to determine which measures and attributes are the main drivers behind your KPI. You can do this by clicking Browse and searching for your columns or by clicking in the search bar and typing.
    Note: Start with 5 - 10 attributes from different dimensions. Adding varied columns helps analyze the KPI in unique ways to get more options when looking for insights.
  4. Select your Time Interval of Day, Week, Month, Quarter, Year, or Custom. This field impacts how Birst will plot your data.
  5. By clicking the Saved Configurations icon, you can view, edit, or create your saved configurations for this KPI. For more information, see Managing Smart Insights.
  6. Click Show Smart Insights. Birst will identify the attributes and measures explaining your KPI then plot them for you using the Value Based Design
    .
  7. Save creates a new dashboard with your Smart Insights results. You will be able to access your new reports from your Collection Manager.
    Note: Not finding what you're looking for? There are several reasons why Smart Insights may not return results in your search.
    • Not enough data: First, make sure your dashboard filters are not too restrictive. In some cases, the attributes you included in the analysis may not have enough elements. For example, analyzing four years of Sales data with only the Year attribute and the World Continent attribute will result in 20 rows of data. Try adding more attributes to increase the grain of the data and run Smart Insights again.
    • KPI type: Smart Insights works best on "source" KPIs. These are the KPIs that are formed using a simple measure coming from your subject area.
      • Expression-based KPIs are currently not fully supported. Depending on the complexity of your expression, you may get fewer or no results.
      • Source KPIs using "Count" or "Count Distinct" as the aggregation type will not return results
    • Measures with NULL values: Measures having multiple NULL values are harder to analyze. Birst will intelligently decide whether to remove a measure because it has too many nulls at the granularity analyzed. If so, it will warn you in the results screen. In general, try filtering out the NULLs before running Smart Insights.
    • Large dimensions: Attributes that contain many different elements will have more natural variation and be less likely candidates to explain the behavior of a certain KPI. The more elements, the less common patterns can be found across these. For example, explaining your sales by the customer's email address is likely not valuable since you may have thousands of different records. It is also very challenging to effectively visualize how a KPI evolves across a large dimension. Birst automatically removes dimensions with more than 50 unique values from the result set and will notify you in the results screen. To include such an attribute in the analysis, filter it to the most important 50 or fewer elements you wish to analyze, and run Smart Insights again.

    For more information, see Understanding Smart Insights.

Related topics