Data-Driven Forecasting

In this month’s comment I provide a further observation grounded in my background of working with large advertising budget clients, and with relevance for businesses operating at a range of scales.

Planning and forecasting are serious considerations whether working with big or small budget clients. Forecasts might be done a year in advance, and they can be reviewed a quarter ahead for Business As Usual (BAU) activities. They are carried out specifically for each one-off individual campaign and activation.

Working with different types of clients, it is clear that there is often variation in the expectation of what can be achieved from investment in this type of activity. Importantly, a forecast can’t be realistic and/or achievable if it is not data-driven. The historical performance of an advertising account is the key foundation of a realistic forecast which combined with a percentage of improvement YOY can create an achievable plan for future advertising campaigns.

If there are no historical data, or a significant change in business activity has occurred recently skewing the available data (e.g. the client’s website has changed completely), then such factors can make it difficult for the media agency to forecast delivery of realistic sale and return on advertising spend. Sometimes agencies might use their other clients’ data to create this forecast, but I would argue that customer interaction with any brand and their advertising creative are unique characteristics; furthermore, similar data obtained from a competing brand may not be available.

As a young brand entering the advertising landscape for the first time or else possessing unreliable historical data, the most effective approach is to allow some time for your advertising accounts to build-up relevant data and performance history before you expect your media agency to present you with realistic scenarios and useful results.

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: