Based on how well a campaign’s target audience is familiar with the product/brand, marketing campaigns can be roughly divided into two groups:
- Acquisition campaigns – targeting potential customers with no prior interactions with the brand,
- Retargeting campaigns – targeting people who are already familiar with the brand to some extent.
In most cases, retargeting campaigns outperform the acquisition campaigns in terms of customer acquisition cost (referred to as CAC), as the audience of retargeting campaigns is by definition more familiar with the brand and its products. So if we compare acquisition and retargeting campaigns solely based on their respective CAC figures, retargeting campaigns always win. On the other hand, it could be argued that acquisition campaigns serve a different purpose than just bringing conversions; they also raise brand awareness. In this article, we would like to see if there is a way to test whether or not acquisition campaigns increase brand awareness, and if so, how to measure the effect.
We start by formalising the problem, after which we present some relevant theory needed in order to analyze the data. After this exposition, we link to a Google Sheet document where the theory is implemented on some sample data. This Google Sheet document is ready to be used on new data, so feel free to explore with your own campaign performance data from Facebook, Google or any other channel. Information about how to use the document and how to interpret the results is included in the document itself.
The problem
We assume the account has only one acquisition campaign, for which we want to test if it increases brand awareness. For the purposes of this article, “brand awareness” is defined as the number of Google searches for that brand. The reasoning is that if there are 1000 searches for a specific brand in January, and 2000 searches for the same brand in February, then it’s reasonable to assume that brand awareness has increased. Similarly, if the number of searches drops in February to 500, we would say that brand awareness is lower. This gives us an idea of what would be the independent variable in the experiment (the variable for which we have control over), and what would be the dependent variable:
- Independent variable: daily spend on the acquisition campaign (CS for campaign spend),
- Dependent variable: number of daily searches for the brand, on Google (GS for Google Searches).
Some theory
We assume the simple linear regression model:
Where b0,b1 are model parameters we need to calculate from historical data. Data for the number of daily brand searches can be acquired from the Google Search Console, and daily spend for the acquisition campaign can be collected from whichever advertising platform is used. Using the least squares method, the values of parameters can easily be found from historical data (the formula and its implementation can be found in the provided spreadsheet). The interpretation of the parameters is very important, so we advise getting familiar with their meaning:
- b1 – the slope of the fitted linear curve: how many additional Google search queries does one additional dollar spend on the acquisition campaign bring,
- b0 – the y-intercept of the fitted linear curve: the number of daily brand Google searches with no spend on the acquisition campaign in question.
Once these parameter values are calculated, we are interested in the following:
- We would like to test if there is any correlation between the independent and dependent variable.
- If we prove that the relationship exists, how can we use it to make predictions?
Proving the correlation
Proving the correlation between two numerical variables means proving that changes in the independent variable (campaign spend) will cause changes in the dependent variable (number of searches). If we take a look at the model once again,
We see that changes in spend will change the number of Google searches if and only if the value of the parameter a is different from zero. In the formal language of statistics, we want to test the following hypothesis:
Numerical example
Download the spreadsheet to run the analysis on your own data.
Measuring brand awareness
Suppose the sample provided was sufficient to prove that the slope of the line is not equal to zero, and let’s say that the slope was calculated to be b1=0.1.
Once more, the interpretation of this figure is that 1000 additional dollars spent on the campaign will bring 100 additional searches on Google. Scaling this up, we get that $1000 additional dollars spent on the campaign will bring 100 additional Google searches. To get from Google searches to purchases, two metrics are required: CTR for the brand keywords and the conversion rate for the same keywords. CTR can be found in the Google Search Console, and CR in Google Analytics. If these two tools are linked together, then all of the data can be read in Google Analytics. Assuming that the CTR for the brand keywords is around 45%, and that the conversion rate for the brand keyword is 10%, these 100 additional Google Searches amount to 45 additional website visitors and 4.5 additional organic purchases.
Conclusion
The above described method represents a starting point for analyzing the impact of individual acquisition channels on the amount of organic traffic. There are a number of ways the model can be improved upon:
- Selecting more than one organic keyword,
- Analyzing more campaigns at the same time, as it’s not a reasonable assumption that only one acquisition campaign/channel is active (this brings up the topic of multilinear regression),
- Using time series to analyze data. In our calculations, the temporal component of the data was lost, and we treated each data point as independent from other data points, which may not be the case.
This article serves as a way of starting to think about how different traffic sources are related, and in the next article, we will start to fix the mentioned model deficiencies, and start to build a better understanding of how brand awareness is spread.
For more analysis like this, feel free to try our app which helps marketers like you make sense of advertising data: RAPP.
Note: If you found this interesting and would like to discuss further, feel free to reach out to us at “[email protected]”.
No comment yet, add your voice below!