Weather plays a factor in many of the subjects in our case studies. Although the weather is often not a major component of the analysis, weather data has helped us in the following examples of our work:
- Retail sales forecasting
- Demand planning
- Marketing mix modeling
- Predictive maintenance (IoT)
- Employee retention
- Fuel conservation
- Animal health
- Food safety
- Food manufacturing quality
- As part of our research of the domain for which we are building models, we always first try to understand how the weather can play a factor, and then go and acquire the data.
- We have standard routines to extract data from public sources. However, we find that special transformations work best in models, rather than using simple absolute raw measures like temperature and precipitation. Often the variables need to be recast in relative terms, expressed as moving averages, or put into classification buckets.
- Though many companies prefer to use free sources, there is often value in using the data that third parties provide, as they can supply these value-added transformations. Additionally, they can provide future values of the variables, something that is not available from government and free sources.
Examples of some of the ways weather has led to insights include:
- The discovery that ambient humidity was responsible for much of the food quality production issues a manufacturer was experiencing, in conjunction with such controllable factors as oven temperature set-points.
- A retailer was able to learn, by category and region, how upcoming weather could help them improve their financial forecasts.
- Several companies were able to deploy safety risk identification systems to alert managers of risky conditions for employees at the start of each workday and location.
- Food producers identified the weather conditions that lead to an elevated risk of pathogen infection.