Are we just as happy to shop on sunny days as on rainy days? Which weather variables are the correct ones to differentiate between good and bad weather? How can we use the knowledge of the current weather and the weather forecast to optimize the placement of advertisement?
Typically – just as in our weather project – the first challenges for analysts are the availability of data, the elimination of data errors, the merging of data of different origin, and the correct aggregation of relevant information. Our first goal was to reduce the variety of quantitative information about the historic, current, and forecast weather that was provided to us by Wetter.com to the most essential parameters. Thereafter we included geographical, socio-economical, and periodical data in our analysis in order to be able to explain the sales of our client’s product – which can be purchased online exclusively – in the best way possible. The result is a spatiotemporal, explicit model with high resolution which is easily scalable and can be transferred to many other products at a low effort.
The resulting regression model has very high explanatory power (with a coefficient of determination of 0.98) and therefore is suitable to explain the various influences on the purchase decision of certain products. Furthermore we discovered that the spurring factors aren’t the absolute weather, but the relative changes, or rather the deviation from the expected values.
The essential advantage of our model is a precise basis for decision-making for a weather-based targeting of advertisement. Sales can be increased by weather-dependent placement of advertisement and at the same time spendings for advertisement can be used more efficiently, resulting in increased gains.