InsightsMeasurement

The Art & Science of Paid Media Forecasting: How Confident Are You In Your Forecasting? 

Written by Emmeline Constanti Crosby

Posted on: November 26, 2023
Bio

It is safe to say that we would all like a little more certainty right now, especially when forecasting performance for next year. Considering the range of external factors that can impact a forecast, approaching your upcoming media plan with confidence is a methodological skill, and one that Andy Blewden, a co-founder of DemandMore, discussed in a recent webinar.  

Andy has spent the last ten years running paid media for some of the UK’s biggest brands, including Which? and JD Sports. He sets out three crucial steps of building a reliable forecast, and managing the multitude of ‘what if’ questions that usually occur during this process: ‘what happens if we increase spend by X?’ or ‘how many conversions can we get for X budget?’ 

 Granularity

The first key step to an accurate forecast is granularity, through preparation and segmentation. Data must be prepped in three stages: (1) start with last year’s data, (2) split it by month and (3) ensure that it is full funnel. 

Divisions by month enable you to identify any anomalies and take into account seasonality, whilst a full-funnel picture is essential if you are a lead-generation business tracking MQLs, SQLs and revenue.

After the initial prepping of the data, it must be segmented by (1) channel, such as Google or Bing Ads, (2) sub-channel, such as Search or Shopping and (3) intention, such as Brand and Competitors within the Search channel. This is to ensure that certain groupings don’t skew your scale expectations, for example, grouping Brand Impression Share with Search might lead to overly optimistic expectations on Generic Search. 

The Art & Science

The second step to forecast accuracy lies in understanding the ‘science’ of what to expect for each channel, in addition to the limitations that each channel is exposed to, or the ‘art’.

The Science:

There are three elements of forecasting across the channels of Search & Shopping and Display & Video respectively.

Search and Shopping

  1. Forecast using a few base metrics and calculate the rest: Andy suggests forecasting impressions, CTR, CPC, AOV and CVR and calculating the rest. Metrics like CTR, CVR and AOV can be forecasted based on past performance, even if you’re looking to scale your spend. CPC and Impressions can be forecasted using tools such as Google Performance Planner, giving projections for those metrics at different spend levels. All other metrics like CPA, revenue and ROAS can be calculated based of these 5 input metrics. We have put together a complimentary template with these formulas built in to help save you time.
  2. Use Performance Planner: Google’s native Performance Planner can be effective at forecasting diminishing returns at different spend levels. Andy suggests using this tool to forecast CPCs and impressions at different spend levels.
  3. Review Impression Share: Finally, when forecasting impressions, be careful to review your Impression Share. For example, if you are already achieving a 80% IS, you can only scale 25% more, or if you have a 95% IS, your budget will not be particularly scalable. 

Display and Video

  1. Use the forecasting template: Firstly, use our complimentary template to forecast impressions, CTR, CPC, AOV and CVR and to help you scale more linearly. 
  2. Fill the funnel from the top: If you are looking to scale remarketing, ensure that you are scaling prospecting to fill the audiences, as it will not be possible if the size of the remarketing audiences doesn’t increase. 
  3. Be aware of the halo effect: Finally, beyond driving direct conversions, these channels drive lift across Brand and direct, which needs to be considered. The incremental impact of top-of-funnel can also be understood through geo-uplift tests.

In a geo-uplift test, you run a study comparing location group A vs B, preferably choosing two groups with similar search volumes. 

You would increase spend by 50% in one location, and not the other, for at least a month, and then evaluate the geo-uplift effect on Brand. 

For example, you might spend more on top-of-funnel in one city, and achieve 20% more Brand searches or conversions. By understanding this correlation, you can then apply the multiplier to your bottom-of funnel forecast. 

The Art:

Despite following the above channel-specific methodologies, you will encounter certain limitations:

  1. Controllables: These include plans for new products, campaigns or websites, which will affect performance and, in some cases, will require extra budget. 
  2. Uncontrollables: Economics, press, and stock shortages will influence performance. 
  3. Mismatched promotions: Running sales and promotions in different months year-on-year will impact performance. 

This is where the art comes in. you will have to draw by eye here and take an educated guess of the impact these changes will have. Make sure to trust your intuition here.

Managing uncertainty

Despite extensive forecasting, we would recommend building two forecasts, one ‘ touch of optimism’ and the other ‘cautious’, to manage uncertainty and unexpected limitations. 

The ‘touch of optimism’ forecast should be your working model. It is a cautiously optimistic forecast, but still fairly neutral.

The ‘cautious’ forecast is more of a stress test, considering if you can adapt if your other forecast doesn’t quite go to plan. 

It also helps set expectations of the board, taking into account a wider range of external factors that could affect performance, whether you’re forecasting yourself, or alongside an agency partner.