When building a statistical forecast, the best foundation to start from is with your company’s own transactional data. However, once you’ve progressed to evaluating variations in seasonality and explaining unusual errors in your forecast, you may find that external factors are impacting your business and bottom line. For example, weather such as floods and dramatic storms can drastically affect how you use your data, to make future business decisions.
Understanding these factors and finding data to support your theories is critical especially with strategic and sophisticated Revenue Management systems where the scope of influence may span far into the future.
When evaluating whether a given external data source is relevant to your analysis, consider these essential aspects of your forecast:
- Data Granularity – the better the granularity, the better you can explain the low-level ebbs and flows in your data, whether geographical or temporal or demographic;
- Frequency of Update – a quarterly data source may suffice for annual decisions, but daily is best for micro-transactions. Choose the data source that fits best with the scale of the decisions you are making;
- Consistency – how volatile is the data? Do you often observe gaps? Some of this can be handled through outlier controls and fill-in logic. However, the more assumptions you build into your analysis, there becomes a point of diminishing returns. You may choose to use the data to validate your assumptions then simply parameterize your predictive analytics and avoid undue noise;
- Accessibility – is the data source freely available or is there a license fee to gain access? Or are there technical hurdles to bringing in a full historical load into your analysis? Having accessibility to the data source will ensure a true assessment of insights.
Within the United States, there is a wealth of publicly available macroeconomic data recorded at the zip code level and measured on a monthly basis. In addition, financial institutions record and report on industry-specific indexes on a frequent basis. Depending on your organization, it’s essential to determine if these metrics can infer and explain unusual behavior in historical transactions, or uncover potential leading indicators that can help to increase forecast accuracy. For example, we find that by leveraging predictive analytics, weather is a factor and explains automotive sales, building permits and/or retail sales, which can impede your organic revenue growth. Using these factors to help filter down which external factors are essential and will help you the most.