Demand Prediction

The Forecasting-Engine is a Machine Learning system which trains, tests and runs predictive models using external and client data.

Predict the unpredictable

The Forecasting-Engine analyses a vast range of structured and unstructured data to connect seemingly unrelated information and reveal correlations and models for prediction. It produces accurate and reliable results that can drive decisions about everything from demand forecasting to programmatic advertising efficiency.

Forecasting cold & flu outbreaks for GSK

case study - forecasts events with science

Predictions using our Forecasting-Engine can massively boost efficiency in managing supply chains. For example, GSK knows that when the pollen count rises or flu season approaches, a spike in drug sales will occur.

But failing to forecast exactly when that is means missed opportunities to increase stock in the distribution network or activate marketing activities to maximise sales.

Early Warning System

case study - warning system uses the collected data to inform clients

By combining GSK’s sales data with external data-sets such as search behaviour, weather and social media, we were able to discover unique triggers that signalled a potential outbreak at postcode level

Through our prediction algorithm we created an ‘Early Warning System’ model that integrated directly with GSK’s retail and digital media partners to automatically activate geo-targeted advertising and in-store activation.

Hyper-local accuracy

case study - predict the future location of infected areas with data science

We predicted high-risk areas for the spread of cold and flu with hyper-local accuracy, a full four days ahead of any local authority advice. The client’s geo-targeted media reached 9 million consumers, with a 200% uplift on previous click-through rates and a 12% rise in sales.

Retail: A forensic business

For retailers, reliable forecasting is a critical and increasingly forensic business function. Getting it right drives supply efficiencies and maximises revenues and customer satisfaction. Getting it wrong means stock shortages, lost sales and wastage. Our Forecast-Engine can automate modelling thousands of Stock Keeping Units across stores and regions, achieving accuracies of up to 95%.

Discover here, how we helped Tesco predict BBQ weekend sales more accurately.

Find out more

Other examples from the field

Consumer goods

We improved product demand forecasting accuracy by 10% for Mitsui, a leading Japanese FMCG supplier, by combining sales, promotions and store information unified with external weather and environmental data.

Read more

Entertainment

For Disney, we forecast opening weekend box office sales to 74% accuracy three weeks in advance of a new film’s release, by using a combination of internal and external sources including social media and trailer views.

Read more

Unlock new possibilities from data

We work collaboratively with our clients to create powerful, predictive models unique to their business model and needs. Here’s how we do it:

Client Engagement Process

client engagement process