In today’s competitive landscape, defining the promotional calendar and the flyer are among the most critical processes in Retail.
Current software tools available to Marketing and Commercial departments fall short in supporting decisions on which products to include and at what price.
The main solutions on the market have significant gaps in the features that are fundamental for correct adoption.
Basing on a set of products eligible for promotion, GPromo® selects the product subset, the mechanic to propose, and the placement within the flyer.
The developed solution includes a sales forecasting module (mechanism-dependent promotional lift) focused on calculating the metrics underlying the choices. In building the input data model, all the various qualifying elements of the promotional process (POS clusters, products, features, exogenous variables, contextual data, etc.) are analyzed and incorporated.
To train the model, the system compiles a table in which the various exogenous factors for calculating the promotional lift are calculated and attributed for each promotion (intended as a product/store pair).
Definition of the main strategic goals and the weighting to be assigned to key KPIs (sales, margin, profit), the pages, and the number and type of products to include in the flyer, integrating Marketing and Sales recommendations.
Once the basic mechanics to be applied have been defined, GPromo® proceeds with the calculation of the promotional lift and the expected sales quantities, automatically adjusting the turnover and margin metrics.
Identification of the best item/mechanical option based on the defined strategic rules.
GPromo® proceeds with the population of the flyer by integrating the rules and constraints defined by Marketing and the optimization process.
GPromo® is developed entirely on the Google Cloud Platform, which in addition to AI and ML cloud tools and services allows you to create and manage web and mobile applications.
The forecast model aims to predict the promotional lift of an item.
The forecast model is trained based on at least one year of store sales.
A separate forecast model is typically built for each product macro-category.
The forecast model takes into account various exogenous variables and item attributes that can be parameterized based on the specific customer.
The mechanical management module applies the defined commercial logic and constraints, generating a forecast of sales and margins.
We developed the optimization model using OR-Tools, the world's most widely used open-source operations research tool, and integrated it into the Google Cloud platform, making it resilient.
The model identifies the best item/mechanical option by integrating several steps like profit-to-sales metrics optimization based on defined strategic rules, integration of defining any lower bounds for non-optimized metrics, management of product crowding constraints at various levels and control of incompatible brand mixes.
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus amet
