Marketing Spend Optimization by Market Mix Modeling & MROIs
Introduction:
‘Half the money I spend on advertising is wasted; the trouble is I don’t know which half’ — John Wanamaker
Companies are spending a lot of money on digital marketing for increasing their revenues. This digital marketing spend is becoming a hot spot for companies to focus on optimizing their resources and improving their Key Performance Indicators (KPIs). Return on Ad Spend (ROAS), Click Through Rate (CTR), Revenue, Cost Per Acquisition (CPA) are the major KPIs that companies monitor daily for measuring the incrementality.
Companies try a variety of marketing channels including Affiliate Marketing, Search Engine Marketing (SEM), Organic, Social, etc., for marketing their product/services. They plan the budget for each of these channels based on the performance and forecast.
Here comes the hardest part which is measuring the performance of marketing channel. When a conversion happens, it gets attributed to a channel based on attribution logics in place. But, the customer journey plays a key role in his/her decision-making in purchasing the product which is not captured by the single channel attribution. Customers can see an ad on Facebook and then get exposed to another ad of the same product on YouTube and then decide on purchasing. In this scenario, there is an influence of both ads in the purchase. But we don’t know how much impact FB alone has on the customers decision making.
We want to measure this type of interaction and the impact of one channel on the other and ultimately on revenues. This measurement can be achieved by developing the Marketing Mix Modeling.
Marketing Mix Model:
Marketing mix modeling (MMM) is statistical analysis such as multivariate regressions on sales and marketing time series data to estimate the impact of various marketing tactics (marketing mix) on sales and then forecast the impact of future sets of tactics— Wikipedia
Required Data: (An Example)
- Spend for each marketing channel by day
- Adstock for each channel by using decays
- Total Revenues by day
- Seasonality factor
- Price change factor
Data Transformations:
Once we get the Adstock model for each marketing channel by using the decays factor, create a scatter plot for all variables Vs Revenues. This gives us the idea of how each variable is related to the dependent variable. We are good to go if all the relationships were linear. But, if there is any non-linear relationship existing, we need to transform the data by using the required transformations. This can be logarithmic, exponential, etc.
Regression:
After making the required data transformations, the regression equation with the interaction terms looks like,
Interaction terms help us to measure the impact of one marketing channel on the other. Pick either Python/R-Programming to run this multivariate regression. After running the regression, we need to analyze the summary statistics to identify the variables and the interactions which are statistically significant to influence the dependent variable which is our total revenue.
Now, remove the variables which are statistically insignificant to our analysis. Now, it’s time to concentrate more on beta values which are variable coefficients. The interpretation of the coefficient for affiliate spend is, by increasing $1 in affiliate spend and keeping all other factors constant, total revenue increases by the value of the coefficient. This is how we can calculate the Marginal Return On Investment (MROI).
As we have interaction terms included in our analysis, we need to take a partial derivative for getting the MROIs for each channel. MROI for affiliate channel from the above revenue regression equation is,
Similarly, we can calculate MROIs for all other channels.
Marketing revenues follows diminishing returns principle which is Advertising Saturation.
“Increasing the amount of advertising increases the percent of the audience reached by the advertising, hence increases demand, but a linear increase in the advertising exposure doesn’t have a similar linear effect on demand. Typically each incremental amount of advertising causes a progressively lesser effect on demand increase” — Wikipedia
This non-linearity can be taken care of with data transformations.
Conclusion:
So, from these marketing channels MROIs, we can measure the saturation points and can gauge the returns expected. Generally, the marketing channel which has high MROI is optimal to invest our very next $. This is how we can optimize our marketing spend for getting optimal Return On Ad Spend (ROAS).
