Supercharge Your Cross-channel Customer Acquisition
This article was also published on Towards Data Science.
Offline marketing and sales channels are returning, and you need to adapt.
These issues are costing brands millions of dollars per year. With the current economic uncertainty, preventing such a waste of money is more critical than ever. Therefore, brand executives are tasking their marketing and data teams to explore potential solutions and determine the most effective ones. In this article, I will discuss how you can tackle these dilemmas with marketing mix modeling (MMM) and gain an edge in today’s challenging economy.
Measuring marketing signals with MMM
A classic approach to measuring whether a marketing initiative is driving return is through regression. We can discover whether the variable brings a return by evaluating the correlation between each independent variable and return.
MMM is based on complex regressions that process many inputs, including ad spend, macro economy and other external factors, and revenue, and then attribute the conversion credit to each input. Typical MMM inputs include the following.
And MMM showcases factor contributions as below.
MMM in practice
Let’s look at a real-life example. A beauty brand sells products across its storefront, Amazon, Sephora, and convenience stores, and it spends sizable marketing budgets on Google, Meta, Amazon, TV, podcast, and in-store. Every month, it wants to understand how its ad spend affects sales across channels. The beauty brand is also interested in learning the best marketing budget allocation for the following month.
From experience, many beauty brand customers see ads online and purchase offline later, and vice versa. Therefore, it’s unrealistic to separate online and offline channels' performance.
Here’s how MMM can help. By inputting daily ad spend, beauty market trends, and daily sales data into the model, the brand can see how each variable contributes to sales over time.
Better yet, regression models are easy to visualize and explain. Business stakeholders from the beauty brand can evaluate how well the models fit their business and decide whether to accept the model results.
On the flip side, by inputting different ad spend scenarios into the trained models, the beauty brand will get the associated revenue forecasts in those scenarios.
Connect the dots across marketing and sales channels
MMM ingests aggregated data like daily ad spend and sales data instead of granular user-level data like clickstream. Ad spend and sales data are usually accessible from major marketing and sales platforms, and each data attribute generally follows similar measures. Therefore, brands can expect comparable marketing measurement across ad platforms. Without such comparability, marketing measurement will be unreliable.
Similarly, MMM provides a reliable measurement for view-through ads. If brands try to measure view-through ad performance with attribution models, the measurement is likely underestimated. Since attribution models rely on clickstream data, if a user sees an ad but doesn't click on it, the models won't know whether a user has viewed it. In such cases, MMM becomes an excellent complement to attribution models.
MMM has its limitations
However, MMM is not perfect and has a lot of limitations.
Since MMM takes aggregated data, it requires dynamic and long-term marketing data to detect enough market signals. Therefore, only brands investing heavily in marketing can utilize MMM. Further, if a brand wants to measure a particular marketing channel, the brand must have active marketing activities on that channel for an extended period. Otherwise, MMM can't generate meaningful results due to the lack of data.
Also, MMM generally can only measure on the ad platform level at best, and it is unable to measure on a campaign level. That is because most brands won't generate enough data points on a particular campaign.
Finally, from time to time, brands need to run experiments on campaigns to create dynamic marketing movements for MMM to work. Not all teams will have the bandwidth to operate campaigns properly to make the best of MMM.
Marketing measurement should serve your revenue goal
To summarize, here are the use cases in which MMM can add value to brands uniquely:
Investing diversely across multiple marketing channels
Selling both online and offline
Online marketing is intended to increase awareness of offline sales, and vice versa
Spending heavily on view-through ads like video and some paid social ads.
However, there are certain scenarios in which MMM may not be as beneficial for brands:
If specific channels have short marketing or sales history, MMM won't have an accurate measurement for these channels
The measurement goal is set on a campaign level
The brand has limited bandwidth to run marketing experiments, so there are not enough marketing signals for MMM.
It's important to remember that no marketing measurement approach is perfect. Marketers and data teams should look for a combination of techniques that best fit their use cases and help them make more profits through marketing. In the current economy, it's essential to operate with less, and marketing measurement approaches can help us achieve that.
I discuss how to use data science to level up your business and optimize your marketing in my articles. If you want to discuss marketing measurement or other data science topics, please follow me on LinkedIn. Until next time.
