Grouping Analyses in Beam
What is an Analysis?
An Analysis in Beam is an evaluation of a particular demand data set for a specific location and date range. This can be created on Control Center from any location you define or one of your saved locations in Location Insights. Alternatively, you can create Analyses at scale via the Beam API.
Why group Analyses?
While single Analyses are ideal, offering insights tailored to each store or location, grouping Analyses can present a more practical approach as it provides a manageable, aggregated view across multiple stores or locations. This can be useful for businesses, such as retail chains, that manage operations at a regional or state level. Group Analyses support these operations, like marketing campaigns, inventory management, and demand forecasting, by delivering aggregated insights consolidated across all stores.
How to effectively create Groups
Creating groups within Beam requires careful thought and consideration. Follow these guidelines for optimal results:
Group Size: Include at least two Analyses in a Group.
Location Replication: Prevent unwanted bias by ensuring each Analysis within a Group corresponds to a unique store or location.
Demand Consistency: Ensure all Analysis within a Group are based on the same demand and unit of measurement. For example, in-store pizza sales in $US.
Country Consistency: Ensure all stores or locations are within the same country to maintain a level of homogeneity.
Temporal Consistency: Aim for either identical, or at least, substantial overlap in date ranges across Analyses.
Group Membership: Define Groups based on meaningful criteria that offer strategic value. Common grouping parameters include:
Geographical Location: e.g. by state, region, or city.
Type of Area: e.g. urban or rural areas.
Store Characteristics: e.g. standalone or inside a mall.
Performance: e.g. high, medium or low sales.
How to create Group Analyses in Beam
At least two Analyses are required to create a meaningful Group. First create Analyses, then create Groups.
Analyses:
Analysis creation: Analyses can be created on Control Center or via the Beam API.
Analysis name: This user-defined field serves labeling and identification purposes. Incorporating group membership details into the Analysis name helps in identifying relevant Analyses, thereby speeding up the group creation step. For example, ‘groupA__store123__instore_pizza_sales__2022’.
Groups:
Group creation: Groups can be created on Control Center by selecting specific Analyses or via the Beam API by passing a list of Analysis IDs.
User-defined grouping: Users can define their own Groups based on business needs.
Flexible group membership: An Analysis can be part of more than one Group, providing greater flexibility.
How to use Category Importance
Category importance at the group-level follows a similar interpretation as that of single Analyses. It represents a weighted aggregation of the Category Importance from each contributing Analysis, where the weights are proportional to the average daily demand of each. This gives more influence to Analyses with a larger share of the overall group demand.
As with single Analyses, the important categories highlight key drivers of demand for your stores or locations, though with a more generalized view. Different strategic actions can be taken based on these insights, for example:
Improved demand forecasting accuracy:
Integrate Category Importance results using the Beam API and Features API by introducing specific event features to your models. See this notebook for an example of how to get features for all your stores or locations.
Informed marketing strategies:
Run targeted campaigns to capitalize on events occurring near your stores or locations, taking advantage of the buzz around events.
Run off-season promotions during quiet periods to stimulate demand and maintain customer engagement.
Partnerships and sponsorships:
Collaborate with event organizers or related businesses for cross-promotional opportunities.
Tailored product offerings:
Develop and offer products or services that align with the interests and needs of customers attending these events.
Considerations and Watchouts
Impact of demand share: The aggregation of Category Importance assigns weight to analyses based on their average daily demand, naturally giving more weight to those with a higher demand. This approach is designed to reflect the real-world impact of each analysis. If more balanced results are desired, consider grouping analyses with comparable levels of demand to prevent any single analysis dominating the overall results.
Integration with Features API: The Beam API provides a list of features for each important category, which can be incorporated into your models via the Features API. Ideally, these features should be tailored to the store or location of interest but using group-level features is a viable alternative if store or location-level modeling is impractical. However, this may lead to some stores or locations having features that are entirely zeros, which is something to consider and take into account.