Forecasting is the backbone of every supply chain. For years, the same models have allowed businesses to minimise waste, meet demand and maximise revenue. Exceptional, unplanned events can render these models obsolete, requiring businesses to rapidly adjust to the new normal.
flux is a service that calibrates your demand forecasting by intelligently adjusting for the black swan event. It helps you find the 'new normal', with limited data. Best of all, it's simple to use, easy to integrate into your existing processes and quick to get started.
What we did
✔︎ Built a simple portal for staff to upload time series data from before and after the black swan event in Excel format.
✔︎ Built the flux forecasting module to return a future prediction for the time series, that intelligently incorporates both patterns observed before the event and new patterns observed afterwards.
✔︎ Validated forecasts against hold-out sets to determine accuracy.
flux provides accurate forecasts from after the black swan event, easily integrated into your existing forecasting pipelines. flux incorporates weather data and any other data streams provided, deployed either through an API or batch process.
Whether you're a cafe, clothes store, car showroom or any other business, we're offering free demonstrations of flux on your data. Let us show you how flux predicts your ‘new normal’ with limited data.
average reduction in mean-squared error (MSE) by using flux
How we did it
Typical time series forecasting assumes that the underlying model is the same through the time series, but flux is different. We’ve designed flux to intelligently recalibrate its forecasts given sparse information after a black swan event.
Using ideas from Bayesian statistics and dynamic modelling, we’ve developed a new technique that finds a way to adjust to the new normal, whilst not throwing away all of the relevant information the model has learnt about the fluctuations in demand from before the event.
The flux service includes implementation of the modelling into your existing forecasting processes, incorporation of additional data sources such as weather data and social media streams and AI generated text commentary to accompany the forecasts.
Start a conversation
Take the first step by speaking with one of our data experts today.