Background
A leading US–based online marketplace with an annual media budget of $800m tasked Gain Theory with navigating the weekly impact of changing weather conditions on their business. The company knew from existing models that 32% of weekly sales were driven by changing weather conditions. However, these models didn’t provide forward-looking insights. Specifically, the business wanted to know how it should optimize marketing and media budgets on a weekly basis at a city level as a result of changing weather conditions to drive growth.
Solution
Gain Theory built a predictive model using a combination of algorithms that were tuned in an initial training phase before deployment. Business metrics, media budgets, and weather forecast data from every weather station in the US, for example, were used to ensure the model was as accurate as possible.
Every Monday, the business was able to preview the projected deviation from target sales in all the cities it operated in for the following week, the next two weeks, and one quarter out.
In combination with the scenario planning and optimization capabilities of Gain Theory Interactive, the company was able to reallocate existing budget or seek new investment to offset any projected shortfall in sales. It was also able to provide the rest of the business – particularly finance, marketing, and media – with concrete actions to take across channels, tactics, and campaigns.
Results
Our forecasts were in a range of 0.2-0.7% accuracy of actual sales each week, which gave the business the confidence to act. The predictions quickly became the company’s main source of future–facing information – the CFO used them in weekly shareholder calls.
Accuracy range of actual sales forecasts