I have worked in digital marketing for over ten years and always searched for opportunities to improve my marketing analysis and reporting. Media Mix Modeling (MMM) is of the utmost importance in my work.
I co-founded a SEO and PPC agency with Leigh Buttrey, and we are devoting ourselves to multi-channel marketing. She leads PPC, I lead SEO, and we both know that our work influences the others. We also work closely with internal teams that have many other channels in their marketing media kit.
When it comes to marketing, we can all agree that an omnichannel approach is best.
The challenge? Well, that’s in the analysis. We all know that buyers rarely convert thanks to a media, but what media results?
Well, here Media Mix modeling comes into play.
In this article I research the media mixing model (MMM), starting with its definition, framework and examples for MMM in action. I turned to marketers, sales professionals and business owners who use MMM in their marketing analyzes. They share real anecdotes and tips that help you to feel confident about MMM.
Let’s start.
Table of contents
What is Media Mix modeling?
Media Mix Modeling (MMM) is a statistical analysis technique with which the effects of marketing efforts across channels are measured. His main purpose is to use marketers from the analysis of results in silos and instead to individually understand the combined efforts of different channels in their media mix and the actual inputs of channels.
If you understand how your channels “play” together and how effective every channel is, you can optimize your Media planning. For example, you will better understand which channels should invest more in and for what purpose.
If you stay on it, I will share some real examples of how MMM has changed a company’s approach to marketing.
Media mix modeling in marketing
Media Mix Modeling is a best friend of a marketer – although it is not a little task to set up. You want a lot of clean data, ideally for years.
MMM uses historical data to identify and quantify the relationship between marketing channels and their effects on business goals such as conversion or income. Here it differs greatly from other models.
Last-touch attribution models, for example, only look at the last marketing channel that led to a sale, while the first touch attribution does the opposite.
I believe that most marketers are happy to say that sales rarely come from a channel and play a role.
When modeling Medienmix, marketers can predict the future performance that will help you make marketing decisions such as the assignment of budgets.
With MMM, markers can:
- Collect historical data to things such as marketing editions, sales, trends, etc.
- Develop a statistical model that explains the relationship between marketing activities and business results.
- Interpret results to understand the effectiveness of every marketing channel.
- Use findings to refer budget and resources for the maximum ROI.
- Predicts future performance based on various marketing scenarios.
In my opinion, the main value of MMM is in the data. Instead of making decisions based on intestinal feelings, you can quantify the role of a channel in marketing. In addition, move away from the key figures from the New Channel to make wider data -controlled decisions with the full image of the performance of marketing.
Media Mix modeling framework
The media mixing model framework consists of six steps.
- Data acquisition is based on high quality, longitudinal data from various sources and marketing channels. As shown above, this can include sales, marketing editions, consumers, product, economic and competitor data.
- Data hygiene It’s as simple as it sounds, but it is time -consuming and a very important step. It includes cleaning and pulling the data into a uniform data record that is ready for the analysis. If you do not do this correctly, you will not receive precise data output. Spend your time here.
- Model development In general, it depends on machine learning models with which you can understand the relationship between marketing entries and business results.
- analysis is best carried out with some human interventions. AI can carry out a lot of analyzes and is astonishing for the analysis of large data records, but marketing is very nuanced, and a human overview of AI results is essential.
- optimization Varizes largely from the knowledge gained, but with your new data-controlled knowledge you can optimize your marketing and budget allocation for future campaigns.
- forecast is as it sounds. After you have data, you can predict the potential results of different marketing scenarios, create hypotheses, test and repeat until your marketing leads to your desired result.
Examples of media mix modeling
The best way to hear MMM and its effects are examples in real life. I was thrilled with the amazing findings that the marketers had received below.
Spot synergies between channels
Aaron Whittaker Is the Vice President of Demand Generation at Thrive internet marketing. When Whittaker was asked about the value of MMM, he “changed the way we assign marketing budgets and measure the effects of the cross channel”.
A particularly valuable application that Whittaker found was to discover synergies between channels. In this solid application, Whittaker explains: “When analyzing the holiday campaign of a retail customer, instead of looking at the channels isolated, our MMM revealed unexpected synergies between radio advertising and social media.
We have found that radio advertising during the morning pendulum times increased the commitment of social media by 25% in the following hours – an insight that would not have been visible through traditional attribution models. “
What I like about it: Marketing attribution is a massive challenge for every company, and without MMM it is very easy to overlook the value that has added radio in this case. It would be easy to assume that social media visits, sequence, commitment, etc. have expired. What happens often is that the efforts are fully attributed to social media, but the reality is that radio plays a role here.
With this information you can better rationalize why radio is part of the media mix. Better, you can appeal to the radiomedia at the right time (in the morning, if it turns out to be effective).
Pro tip: If you work on your ads, strive to achieve the desired results, or have evidence (thanks to MMM!) That advertisements work for you, take a look at lifting spots. Paid media template. It does the organization of your media planning and media buying easier.
Quantify long -term brand formation
Whittaker provided many examples of MMM. The choice that should be included in this article was a challenge! I had to include the value of the long -term brand structure and how MMM can contribute to quantifying its role.
When it comes to branding, Whittaker says: “What is fascinating is how MMM contributes to quantifying long -term activities for branding. We found that podcast sponsorship showed a minimal immediate ROI, but our modeling showed that you have made a significant contribution to reducing customer acquisition costs across other channels over six months. This finding has contributed to justifying the continued investments in brand awareness channels. “
What I like about it: Similar to the above, I really like to justify MMM to justify marketing efforts that may otherwise remain unnoticed. It is true that a channel that does not lead to an immediate ROI is “not collapsed” using the last touch attribution model. With MMM you can see how these media work for your company.
Understand the crossover between online and offline media activities
Peter O’CallaghanMarketing manager at Scrapingbeefound the greatest value using MMM to uncover regional trends. O’Callaghan describes MMM as a transformation.
He says it helps “assign budgets, refine messages and identify growth opportunities. It is a powerful tool to predict where you invest and where to withdraw. “
When O’Callaghan was asked for an example, it says: “It helped us to determine California and Texas as hotspots for scratching API demand and to contribute 40% to our leads. By redistribating $ 5,000 for geo-targeted campaigns in these states, we have increased regional engagement by 30%, while we shorten our sales cycle by almost two hours per ladder. This regional focus continues to shape how we approach the campaign strategy. “
Tips for using media mix modeling
Tip 1: Start your analysis with lots of data.
In addition to his victorious MMM examples, Aaron Whittaker advised that anyone who started with the MMM analysis should start with “data at least 18 to 24 months”.
The more data, the easier it is to recognize trends. Whittaker explains: “18-24 months data (helps) make seasonal patterns and long-term effects. We have found that shorter periods often lead to misleading conclusions about the effectiveness of the channel. “
Whittaker has another MMM example that shows the value of data wonderfully.
“A surprising discovery occurred in the modeling of seasonal effects. Our analysis showed that the effectiveness of different channels varied dramatically after the season. E -Mail marketing culminated in the winter months, while the advertising outdoors delivered the highest ROI in summer. As a result, we develop dynamic budget allocation strategies that postpone the expenses due to seasonal effectiveness. “
What I like about it: I am sure that many marketers who read this nod in agreement. We all know that we need data – and the more of it, the better – to carry out a suitable analysis.
Tip 2: Make sure your data is clean.
Peter O’Callaghan advises: “MMM works best if you have clear, measurable goals. Without defined results, it is easy to interpret the findings incorrectly and to react to incomplete information. “
It is easy to overlook where your data work, but O’Callaghan also has some tips for it.
- Pay attention to poor segmentation. O’Callaghan explains that the poor segmentation hides valuable patterns. He says: “If the data is too generalized, important trends that distinguish user groups can be lost. The division of data into smaller, meaningful segments enables you to understand the unique behavior and preferences of different target groups. “
- Imagine your assessment of short -term trends or seasonal spikes. O’Callaghan warns of short-term trends and seasonal spikes and explains: “Mmm outputs can occasionally mislead if they weigh short-term trends too much. Seasonal spikes or external factors can distort the results if they do not take them into account. For example, a unique traffic boost led us to an overvalued email performance. Now we are reviewing MMM results with long-term metrics to ensure a balanced view. “
What I like about it: This is a feeling that we used to hear in this article. I like that O’Callaghan Cross-checing results with long-term metrics recommend. This tip corresponds perfectly to Aaron Whittaker’s tip when you start MMM with long -term data.
Start with Media Mix Modeling
As a marketer and mainly as SEO, I know the value of the media mix modeling. Nevertheless, I write this article and speak to other marketers, I can see how MMM companies helps to make better marketing decisions. Instead of Feeling As with a certain marketing medium, MMM helps you to prove this.
So if you want to start modeling media mix, do it. Remember to collect this long -term data and to exceed short -term results with long -term trends.