Analysis performed by Michael Pearson

NOTE: The following piece is an analysis done by a young graduate with a BS in Applied Mathematics, and a certificate in Data Science. I provided a small bit of guidance for this project, and was absolutely blown away by the results. Not only is it very well done, but it is so timely for our industry. I hope you enjoy it and find it as illuminating as I do. — Michael Caroff

Step 1: The Goal

Using the data provided by the IAB (Internet Advertising Bureau) and other sources (noted below), I am attempting to predict overall revenue for 2020 and 2021.

Step 2: Methodology

To begin with, I spent some time analyzing the existing data. After a bit of experimentation, I determined that using a system similar to a Linear Regression model would yield the most likely results.

The equations I had come up with were:

h1 + h2 = Net
h1 = Net(0.47)
h2 = Net(0.53)
h1/0.47 = h2/0.53
h1 = (0.47h2)/0.53 h2 = (0.53h1)/0.47
(h1 = first half year revenue, h2 = second half year revenue, and Net = annual net revenue).

Once I predicted the missing data for 2016, I needed one more piece of information to help accurately predict the next coming years. So I had taken the average of the annual growth rates, about 19%, and used the equation:

x(n) = 0.19*x(n-1) + x(n-1)

to predict the preceding years annual net revenue (x(n) = net revenue, x(n-1) = previous year net revenue).

Verifying the Predictions

To test the methodology, I extended the revenue figures out to 2019 and then checked my model’s predictions against additional known data that I found. (See PDF file.)

chart of predicted vs actual revenue

As shown by the chart above, this model does a credible job of predicting revenue trends.

Step 3: Extending into the Future

Assuming COVID-19 Never Happens

In order to create a control model, I extended my previous model to the years 2020 and 2021.

Factoring in COVID-19

Using my previous model as a base, I added a reduction rate for the year 2020. Through my research I found that digital CPMs were down 16% by the end of May 2020. (See previously referenced PDF.) Based on that finding and previous CPM trends, I set the yearly reduction rate at 16%. This changed the equation to:

x(n) = 0.19x(n-1) + x(n-1) – 0.16x(n-1)

Using similar methodology from 2020, I then extended into the year 2021. However, I introduced several additional factors.

Assumption 1: People would adjust to the new way of life, which would become the new normal.
Assumption 2: The negative annual net growth rate in 2009 was attributed to the Swine Flu pandemic that had broken out in that year. Remembering what life was like in 2009, I thought about the differences in lifestyle between then and now.
Conclusion: By 2021 the net revenue will begin to bounce back, but not fully recover.
Assumption 3: The reduction rate will decrease by 5% which made the new reduction rate 11%. Applying the new reduction rate produced the following results:

chart of covid vs no covid predictions

Step 4: Further Suggested Analyses

Using these models and predictions as a guide, it is suggested that this project be extended by looking at individual industry verticals: Travel, eCommerce, Health, etc.


About the author:

As a data analyst with a B.S. in Applied Mathematics from Sonoma State and a Certificate in Data Science from UCSD, I worked part time assisting customers in a restaurant while completing the UCSD Data Science program. Coding skills include Python, Pandas, Flask, PostgreSQL, MongoDB, JavaScript (with experience in D3, Plotly, and Leaflet), HTML, Tensorflow, Excel, VBA, Tableau. I work well with others and always enjoy collaboration. I work diligently to communicate with my teammates on what I am doing as well as how to most effectively achieve our plan of action as a group. I hope to use my diverse skill set to help present data in a relevant, user-friendly format.

Michael Pearson | LinkedIn Profile

associations and accreditations

Learn More
Loader