Bollinger Bands and Paid Media
The concept of Bollinger Bands comes from the financial market, especially used in asset analysis and forecasting, as a volatility indicator. But have you ever imagined how fantastic it would be to apply them to analyze your performance in Paid Media? Then find out all about how they work, how they are calculated and practical examples in this article.
Anyone who works in media knows that the market fluctuates a lot and these fluctuations can have a direct impact on account performance. Imagine your client asking you why the CPA has dropped by 10% when no major changes have been made to the account. Or even your own e-commerce is showing this drop.
You look at the whole structure and nothing seems different. How can you explain this change?
This is the kind of situation in which Bollinger Bands can help you in a simple way and with a very solid statistical basis.
What are these bands?
First of all, let's clarify what these "bands" are, which have a rather difficult name. Bollinger Bands are a widespread tool in the financial market worldwide. They help investors analyze market volatility over a given period.
These bands consist of a graph with three linesone upper, one lower, drawn at a distance from a moving average, the third line. As shown in the image below:
How is the calculation made to find these bands?
It's time to get a better understanding of how Bollinger Bands are calculated. They are calculated as follows:
- A middle band corresponds to a simple moving average, based on a 20-day period.
What do you mean? An average based on a 20-day period?
Unlike what we traditionally learn at school, in which the result is just a static number, because it's movingthis average is always moving, adding new values and eliminating old data.
For example: In the moving average on day 21 we have the average of days 1 to 20. Every time a new day is added to the calculation, the oldest day is deleted. Therefore, on day 22, we will have the average of days 2 to 21.- The upper band can be defined as the sum of the moving average and 2 times the standard deviation of the asset being analyzed. The standard deviation is a measure of how uniform the data set is. When the deviation is low, it means that the data set is closer to the average.
Upper band=(Moving Average)+2*(Asset Standard Deviation)- Finally, the lower band is given as the moving average minus 2 times the standard deviation of the analyzed asset.
Lower Band=(Moving Average)-2*(Asset Standard Deviation)
There are various ways of observing and interpreting bands. Here at Pareto we use them to analyze variations in certain metrics related to paid media over time.
How about an example?
In the graph below, we analyze the revenue of an e-commerce business using Bollinger Bands. The data is for 6 months, from January to June 2021.
It can be seen that at the beginning of April 2021, revenue began to show a little volatility, which we can see with the wider spacing of the bands. However, over time we can see a narrowing of the bands, which represents a period of lower volatility for the asset.
Shall we take a closer look at this recipe?
Considering March to April 2021 as a period of high volatility, we can see the following information in the Revenue graph.
In this 2-month period, revenue varied by around 215% between the high and low, going from R$167k to approximately R$53k. This confirms the period of volatility shown in the band graph.
During this period, revenue exceeded the bands on two occasions shown in the graph.
The first occurred on April 5, with revenues of R$167k, surpassing the upper band and showing an upward momentum in revenues.
Then, on April 23rd, we see a point on the graph where revenue exceeds the lower band, reaching close to R$87k. Thus, we see a period of real decline in revenue, as it is not within the range predicted by the bands.
However, when we look at the period from May to June, given as one of low volatility on the band chart, we notice something quite interesting.
In this new 2-month period, revenue varied between highs and lows by 154%. Around 40% less than in the period of high volatility. Revenue fluctuated between R$130k and R$51k in the period analyzed.
Within these two months of lower volatility, we see revenue points outside the band range, above the upper band. Showing positive days and representing growth for e-commerce.
Addendum - Uses
It's worth noting that we also use Bollinger Bands when we want to understand whether a certain oscillation in a variable is within expectations or whether we need to do a more precise investigation of the account.
For example, suppose a client asks for an explanation of a 10% drop in CPA over a short period of time. In this case, if the moving average has broken through the lower band, it means that the drop is greater than a natural oscillation in CPA and should be studied more closely. The same goes for a rise in the variable with the moving average crossing the upper band.
For analysis based on an indicator such as Bollinger Bands, it is recommended to use it in conjunction with others. So, for example, a slight but steady drop in e-commerce revenue indicates that something is wrong. Even if the values don't go outside the bands.
So their purpose is to facilitate analysis by indicating sudden changes that go beyond simple normal variation. But don't overlook cases of constant change within the bands, as they also deserve attention! Combine the use of Bollinger Bands with other in-depth analyses.
Conclusion on Bollinger Bands and Paid Media
In short, the further apart the bands are from the moving average, the higher the volatility of what we are analyzing. A narrower band shows that the metric has low volatility.
Therefore, monitoring this graph periodically is of the utmost importance if you don't want to be caught by surprise by unwanted variations or moments of major change in the market. So be prepared for the next challenges and actions in your business!
Now we've listed some other topics that might interest you:
Winning Email Marketing Method
Market research: using Google Trends to predict demand and trends
