As data-driven marketers, we are often faced with challenging ROAS targets, whilst trying to run media marketing activity as efficiently as possible. In order to do so, we need to hit targets without reaching diminishing returns. But how can we know how much budget to spend on each marketing channel or a campaign?
Back in 2014, our beloved Liz wrote a great blog about additional budget planning, where she shared the Dos and Dont’s on how to efficiently allocate digital spend on a large scale. Specifically, when you've got a bit more in your pot than expected and need to spend it quickly.
The article was based on the media side of things but today, I am going to talk about the same topic from the analytics or media science point of view.
Keep reading this blog if you want to learn a simple analysis that will help you answer it without having to use a finger in the air.
Finding an alternative to regression models
You won’t always have the opportunity but, imagine you could spend as much as you want on a channel or a campaign, as long as you meet your targets. How can you know how much budget to spend while keeping the best ROAS possible?
Building a regression model would probably be your initial approach to answer that question - and I am sure you have come across tons of articles where they explain how to achieve it - but what if this isn’t a feasible option? You may not have the resources required e.g. enough data to get a robust model, time to implement the analysis, etc. What can you do then?
The solution is simple: carry on a correlation analysis
How to implement a correlation analysis
Step 1: Collecting historical data
One of the most common questions that clients and media teams want to answer is how much revenue they would get when spending X budget on a certain piece of media for a media channel Y.
To find this out, all you need to do is to collect the historical spend on the channel Y and revenue figures for the last months/years (as much as the rule: the more the merrier is true, always keep in mind that if the strategy has dramatically changed over time, you may want to work with more recent data, so your conclusions are reliable and the revenue truly due to the spend on the channel).
Step 2: Finding the relation between the variables
To see the impact/relation between the variables in your model, (in this case, media spend and revenue), all you need to do is plotting them in a scatter plot and get the relation between them from the model equation.
Step 3: Diminishing returns and predicting values
By using the coefficients from the model equation, you got in step 2, you can get the maximum spend that reaches your target ROAS without hitting diminishing returns.
This methodology will allow you to play with different budgets and the ROAS you want/need to hit. It doesn’t require years of historical data and you can decide the granularity to apply the insights (channel vs campaign level, for example).
Keep in mind though, that you won’t be able to determine the curve accurately if there are no fluctuations in the historical spend data. For instance, if your budgets have always been low, the relationship between spend and revenue will be linear and you won’t be able to provide an accurate prediction.
Now you know another way to produce predictions about budget reallocation to avoid surprises in the outcomes from your media activity.
If you’d like to know more about allocating budget and maximising ROAS you can contact the Media Science team at Merkle|Periscopix.