In my prior blog, I introduced a multitude of inventory trade-off decisions that TV and radio broadcasters must make on an ongoing basis to drive business. While there are many unknowns in the media industry, there is one common element that helps inform almost every other decision: ratings.
An accurate ratings forecast could improve your pricing, help inform your Upfront and Scatter deals, mitigate your liability concerns, and influence your promo placement decisions. To be successful, you need to systematically blend predictive analytics with business insights to drive adoption. At its core, it can be broken into the following components:
1.Evaluate historical patterns
An accurate ratings forecast shouldn’t only leverage historical patterns to predict the future, however, it’s an essential building block towards the final forecast. The critical element is a suite of time series models that are dynamically selected by segment to ensure each segment of your business uses the ‘best fit’ model. Do you want a steady-state show that’s been around for 10 years to use the same model as your hit show in its second season? Probably not.
2. Proactively react to changing show conditions
The next component in an accurate ratings forecast is the ability to proactively react to the ever-changing show conditions. For example, there could be a steep decline in ratings due to a lead actor leaving the show, or a drastic increase in ratings due to a plot twist. Having the ability to place a large emphasis on recent trends allows the model to quickly calibrate based on the latest ratings, either through systematically re-running the model selection process or through business user interaction. By updating your ratings forecast, your Sales and Revenue Management teams can proactively manage the ADU allocation process for existing orders and realign guaranteed ratings expectations for future deals.
3. Account for internal and external factors
Another key element in your ratings forecast is understanding the impact of other internal and external (i.e., competitor show) factors on your show’s ratings. For internal factors, this could include whether a specific episode of your show is a season finale, premiere (i.e., the first time it’s airing), or a rerun. For external factors, this could include whether or not your competitor has a special event during the time of your show (e.g., a college national championship game). Using a forecasting approach that accounts for both situations will ensure your forecast can properly adjust up or down to dynamic conditions.
4. Seamlessly blend statistical forecasts with research estimates / business insights
The last and most powerful component in driving user adoption of your forecast is blending the statistical forecast with research estimates (or another internal forecast available in your organization). Rather than simply averaging the two forecasts or using gut-based approaches to pick which one ‘feels right’, the forecasts can be blended using a Bayesian approach. A Bayesian approach will put a larger ‘weight’ on the forecast with a tighter confidence band to produce the final ratings forecast.
In order to have an effective ratings forecast, it is essential to leverage historical patterns, proactively react to changing show conditions, account for internal and external factors, and seamlessly blend statistical forecasts with research estimates / business insights. Using this strategy, Revenue Analytics demonstrated how a cable TV company struggling with declining ratings could systematically improve their ratings estimates. The result? Improved forecast accuracy of 40-70 percent for some of their segments.
What could your company do with a more accurate ratings forecast?