gA launches demand forecasting platform based on neural networks

gA launches demand forecasting platform based on neural networks

A new Demand Sensing platform which uses predictive customer data science to sharply increase short-term forecasting accuracy

gA presents “Custonomix”, a demand planning platform based on the smart use of massive volumes of virtually real-time customer data. The platform is designed to reduce uncertainty and increase accuracy in the demand forecasting systems, in particular for multi-brand Consumer Goods companies.

Custonomix uses mathematical predictive models to run large volumes of customer transactional data and external events such as weather forecasts, competitive pricing or promotion campaigns, to automatically synchronize into the demand planning systems, which in turn adjusts and optimizes the supply chain execution platforms.

“The use of Custonomix increases the accuracy of short-term demand planning by 30% compared to traditional historical time series systems”, says Pablo Rodríguez, Supply Chain Practice Director at gA, and formerly the Supply Chain Head at Unilever.

Distribution of snacks and soft drinks in Latin America is highly fragmented, where Consumer Goods companies serve tens and even hundreds of thousands of stores (“changarros” or kiosks in urban and suburban areas) on a daily basis, as well as the conventional supermarket chains. Custonomix captures the daily transaction data on the mobile platforms of the delivery trucks, stored in cloud-based repositories, and then filtered and processed using a predictive analytics based on probabilistic methods and prescriptive algorithms.

Demand sensing increases the efficiency of supply chain execution at SKU level on a weekly basis: it reduces inventory levels and shelf-life risk, increases distribution and replenishment efficiency and optimizes category management programs, it reduces the capital tied up in trade promotion and allows for closer assessment of new product launches.

At the core of Custonomix is Forecastia®, a quantitative forecasting engine based on the use of neural networks, developed by Continente Siete, a predictive algorithms firm based in Buenos Aires. Forecastia has a machine learning model that is capable of understanding the complex relationships between the forecasting variables, incrementing its accuracy power.

gA and Continente Siete have partnered to develop Supply Chain demand planning and simulation models for clients with complex logistics organizations in Latin America.


About gA: at gA (Grupo Assa) we are recognized as a leader in Digital Business Transformation. With facilities and operations in the US and Latin America, we serve clients in 80 cities around the world. We bring 23 years of experience to our clients across 5 vertical markets, and more than 450 transformation projects. Gartner named gA “Cool Vendor in Latin America” for our work in Digital Business Transformation. With a forward-thinking culture, a proven track record and our relentless commitment to  deliver results, gA is the right partner for you. We are gA: discover us (


About Continente Siete: Data Science for Supply Chain. We are a team of academic + business-oriented enthusiasts with an industrial hint. Our background is centered on analytical methodologies ranging from statistics up to artificial intelligence, including simulation, data mining, and machine learning. We have over 7 years of experience working in Supply Chain solutions for different industries such as Consumer Goods, Oil & Gas, Telecommunications, Metalurgy, etc through our 2 main business units: Forecastia and Simcastia. (


You might also like