Are we just as keen to buy on sunny days as on rainy days? Which weather variables are the right ones to differentiate between good and bad weather?
MOTIVATION
Does the weather influence retailing?
Can the targeting of advertising be optimised based on the weather?
PROCEDURE / MODELLING
Automated download of temporally and locally localised weather information for the entire federal territory.
Methodical selection of "relevant" weather variables with an influence on viewing figures.
Consideration of geographical, socio-economic and cyclical effects in order to be able to explain viewing figures as precisely as possible.
Clustering and quantification of weather influences.
RESULT
The driving factors are not the absolute weather, but relative changes or deviations from expectations.
If the advertising budget is optimised according to the weather and redistributed geographically, a sales uplift of up to 30% can be achieved.
Are we just as keen to buy on sunny days as on rainy days? Which weather variables are the right ones to differentiate between good and bad weather? How can we use knowledge of the current weather and forecasts of future weather to optimise advertising? We addressed these and similar questions in a project together with Wetter.com.
Traditionally - and also in our weather project - the first challenges for us analysts are the provision of data, the correction of data errors, the merging of data from different sources and the correct aggregation of the relevant information. Our first analysis objective was to reduce the large amount of quantitative information available to us about the historical, current and forecast weather from Wetter.com's extensive database to the "essential" factors. We then included geographical, socio-economic and cyclical data in the analysis in order to be able to explain purchase rates for our client's product, which is offered exclusively online, as well as possible. The result is a temporally and spatially explicit, high-resolution, yet easily scalable model that can be transferred to many other products with comparatively little effort. Due to its high explanatory power (coefficient of determination of 0.98), the regression model found is ideally suited to explaining the complex influences on the purchase decision of certain products. Furthermore, we found that it is not the absolute weather but relative changes or deviations from the expected state that are the driving factors.
However, the main benefit of the model is that it provides a precise decision-making basis for weather-based targeting of advertising. Weather-dependent targeting can increase sales figures, but at the same time reduce costs for (online) advertising - and thus increase profits.