What is Marketing Mix Modelling? And what challenges does it pose for advertisers? Macy Edwards evaluates the effectiveness of MMM, and how it can be used to measure paid channels accurately.
The Modern Marketing Mix: Measuring Paid Channels Accurately
Marketing Mix Modelling (MMM) is a technique commonly used by advertisers to understand how marketing tactics affect sales. MMM has proven to be effective for us to glean accurate insights about traditional media, but how does this model work for today’s complex, ever-changing digital channels?
Facebook recently released a report outlining aspects of the MMM methodology that need to be reviewed, explaining how the impact of digital channels can be accurately captured if advertisers update their techniques and take these key approaches:
- Adjusting time frames. Search and social demand change so frequently that we should start analysing shorter time frames to ensure we understand how the marketing mix is changing.
- Utilising the right insights. We have more metrics available to us than ever before – the key is to use the right ones. For example, paid impressions that enable effective cross-channel comparisons are preferable to engagement metrics or clicks.
- Contextualising past performance. Assessing media and creative quality has never been more important. Advertisers are able to identify what worked well for them and test which approaches are most suitable. As a result, they can incorporate additional metrics, such as quality scores, into their marketing mix models.
- Calibrating with experiments. Continually testing your hypothesis is fundamental to understanding your digital marketing mix, while also allowing for improving accuracy.
In this article I want to explore these approaches in more detail, and to explain how they can be applied to measure your paid channels accurately.
What is Marketing Mix Modelling?
Marketing Mix Modelling is a methodology used for measuring marketing efficiency. Historically it has focused on print advertising, by using a regression model that determines the relationship between channel spend and business-level outcome (for example, cost per acquisition (CPA) or number of sales).
MMM takes on a functional form:
But what are the challenges of the model?
Most digital channels use dynamic, auction-based ad delivery systems with extensive audience targeting. As a result, there are a lot of variances that make MMM more challenging. If we focus on historical models that are weighted towards conversion, we will lose true visibility on how marketing channels support each other.
Let’s use the example of Facebook. If we were to report on a last click basis, we wouldn’t gain a true representation, as most of the data at a click level is lost along the journey due to cookie rejection between browsers and channels.
Another challenge posed to us is education. Education is the main barrier that must be overcome for business stakeholders to feel comfortable with the recommendations provided by the model (and how to execute them). As mentioned in the Facebook report, output is only as good as input. If spend is the main factor in the MMM, and we are making similar pushes for spend across all channels, it becomes difficult to understand the relationships and to evaluate recommendations.
How does granular data help us to evolve the model?
There are three elements to consider here:
- Click Attribution
- Media Mix Modelling
While there are pros and cons of all three of these elements, they can be used together to provide a holistically informed view of marketing efficiency:
|Click Attribution||Experimentation||Media Mix Modelling|
Very easy to calculate and understandEasy to spot trends over time
Gives a ground truth for the incremental value of marketingEasy to implement on many advertising platforms
Incorporates online and offline channels and can control for non-marketing factorsModel is prescriptive in that it provides an optimal media mix
Only applicable for online channelsPenalises ‘view-through’ channels and favours ‘demand capture’ channels
Not always possible for all channels
Opportunity cost of not marketing to the control groupResults dependent on a point in time media mix
Quality of the results is highly dependent on quality of inputs. Needs years of data.
Difficult to distinguish between correlation and causation
Model can appear as a black box to business stakeholders
Following in-depth research with various MMM partners (for example, Accenture), one of Facebook’s main recommendations is to focus on paid impressions being split in meaningful ways in line with strategy changes. For example, if a Facebook campaign was optimised for mid-funnel marketing to increase engagement, separating the impression variables to account for this shift would make sense. Taking this approach can reveal whether inaccurate data or variables are producing misleading results and insights.
Here at ClickThrough Marketing we use the programme RStudio, along with code dependent on business goals, to understand impression vs click attribution. As spoken about in my last article on marketing attribution and getting stuck on last click, we are working to encompass a broader mix in the context of MMM for our clients. In the future we will be able to use this method to accurately forecast against different variants and, as a result, implement more flexible optimisation between channels.
How do we run experiments for effective testing?
All digital platforms offer conversion lift and brand lift experiments to support us with understanding these relationships. For example, Facebook offers to run conversion lifts to help advertisers understand the effectiveness of Facebook, Instagram and Audience network ads in driving incremental sales and conversions. Similar lifts are available from Google and Amazon.
Facebook remains one of the most difficult channels to attribute in the path to conversion, which is why this methodology is important to adopt. MMM enables us to effectively measure and evolve our Facebook advertising strategies, especially when it’s more important than ever in this time of uncertainty.
Multi-Touch Attribution (MTA) explained
What this all brings us to is Multi-Touch Attribution (MTA). Let’s return to the example of Facebook advertising. If we assume that the impact of Facebook ads on revenue is 2.4%, then the MTA would credit the remaining 2.5% to paid search, organic and direct; however, this halo effect would not have occurred without Facebook.
We see these sorts of halo effects within key channels a lot, making it all the more important to evolve your multi-channel marketing mix to drive more revenue through acquisition growth. MTA gives credit to every touch point, whereas MMM weights credit to the main channel in the journey.
How can we help?
Our teams have strong partnerships with key digital channels, which means we are able to continually explore opportunities to improve measurement and development within digital strategies for our clients. If you would like to find out more about what we can do for your advertising strategy, get in touch with our digital experts today.