SOLUTION

An incrementality experiment was designed using the following methodology:

• Step 1: Collected data - Sales numbers of healthcare devices on the webshop for the past 2 years
• Step 2: Created a model - Use the sales data to build a model that predicts sales.
• Step 3: Make predictions - we predict what the sales would have been if there was no YouTube exposure happened (e.g., the model might predict sales would have stayed around 100 units per day).
• Step 4: validate the accuracy of the model by predicting previous timeperiods to actuals. We achieved a margin of error of ~7%, which was more than good.
• Step 5: Compare actual vs. predicted - Compare the actual sales (150 units - example) with the predicted sales (100 units - example)

In addition, we leveraged in-platform Surveys by Google on YouTube to understand impact on brand awareness and consideration.

Impact

We understood that versus control (the prediction) the total sales volume with the YouTube exposure increased by 17%. From this, we could calculate the total number of incremental sales. These incremental sales were then divided by the total spend of the campaign that we ran. What follows, was the incremental cost per sale: around ~$55.

Given that there was only 1 product with same price, we could straightforwardly calculate incremental return on profit and through that we discovered that the incremental efficiency of YouTube was almost on par of that with Search -- the golden standard of direct response performance.

In addition, we were able to calculate the cost per lifted user in awareness and consideration. While these results were not as strong as the direct response results, they did show a significant lift that was an additional bonus.

Accordingly, spend on YouTube was scaled dramatically using the in-platform data and this experiment as callibration point for in-platform target performance.