• A leading grocery retailer wanted to automate assortment management by store and identify key customer segment drivers to optimize shelf space allocation.
• The objective was to get the right product to the right shopper at the right price.

• Identified common purchase behaviors of customer segments across store locations
• Correlated those patterns to optimized assortment segments
• Created new assortment packs tailored to each store’s demographic profile

• €15M margin impact achieved through optimized assortment allocation
• Constant updates with fresh transaction and external data to refine segments

• A premium grocery retailer needed to understand the lifetime value of its customer base.
• However, it lacked a robust analytical approach to build a value-based customer segmentation.
• The goal was to identify and target its most valuable customer profiles.

• Identified common behaviors of the highest-value customers.
• Most valuable customers of “tomorrow” bought value products and preferred seafood.
• Discontinuing household product purchases was a leading churn indicator.
• Defined behavioral segments and tailored marketing strategies accordingly.

• Defined six customer segments using behavioral dimensions like purchase categories, channel, loyalty, promotion spend, and private label usage.
• Analyzed 3 years of transaction-level data within one week to uncover trends and customer behaviors.

• A retail chain wanted a more accurate long-range sales forecast to better inform its strategic positioning.
• SparkBeyond partnered with the client to build a monthly sales forecasting model by store segment, 15 months rolling forward.

• Incorporated complex dimensions:
• Diverse store segments characterized by differing categories (e.g., fresh vs. non-fresh).
• Considered the impact of new store openings and closures.
• Included store maturity curve effects.
• Accounted for Sunday closure bans (e.g., 25% of franchisees closed).

• Built a suite of models predicting monthly sales up to 15 months ahead for 4500+ locations.
• Achieved 90%+ accuracy in 80% of 1-month revenue forecasts.
• Used to set franchisee sales targets and support long-term planning (budgeting & promotions).

• A top Japanese convenience store retailer aimed to open 2,000 new stores in Tokyo.
• However, the first 200 openings underperformed, triggering an urgent need to uncover profitability patterns and select better locations.

• Combined internal data and external sources (e.g., Google Maps backend, store networks) to identify drivers of store profitability.
• Built a model to predict store revenue and applied optimization algorithms to select the best 1,300+ sites out of 5,000+ possible location

Identified 1,300+ potential high-profit locations based on non-obvious factors, such as proximity to:

• Laundromats
• Fast food chains
• Mobile phone shops
• Gas stations