What is Marketing Mix Modeling?

Let’s start at the beginning — which, by the way, isn’t “What is Marketing Mix Modeling?”. No, it’s “Why should you care?”.

The truth is, nowadays most of us would struggle to name a brand that isn’t promoting itself via multiple channels; social media, display advertising, direct mail, email marketing, the choices sometimes feel endless. And that’s a great thing for marketers — you can reach more people in more ways than ever before.

But do you know what it also does? Make things more complicated.

With today’s consumers being bombarded with numerous ads, promotions and messages on a daily basis, good media coverage alone isn’t enough — you need to reach your customers via channels they prefer, with items that will appeal to them the most, priced according the customers’ price perceptions, promoted correctly, and so on. You know, the whole customer orientation approach in marketing.

And on top of that, you want to track and analyse your efforts too.

It’s almost enough to make you consider a career switch — we hear gardening is quite therapeutic.

But wait, there is an upside!

Get it right and your brand can go further than ever before. That’s right, as the world of marketing becomes more vivid, if not chaotic, recognizing and exploiting the difference between effective and efficient marketing by going for that extra mile with your marketing mix modeling can provide a substantial bottom-line growth plus that competitive edge that makes your marketing so impactful compared to competitors.

Reminder: What is the marketing mix?‍

The key to becoming (and remaining) relevant is to get the correct marketing mix. Your marketing mix, in case you wanted a reminder, is the cross-section between product, place, price and promotion.

Product = What you’re selling

Place = Where you are going to sell it

Price = The amount of money you can (and should) sell it for

Promotion = How you are going to get the word out to potential customers ‍

That is why you should care about Marketing Mix Modeling (MMM) since it is the most effective way of dealing with these seemingly never-ending issues.

And now we can talk a little more about what MMM is…

MMM hinges on statistical analysis and is a method through which you can understand the success of multiple marketing channels simultaneously, as well as better understanding all the factors that play into their success.

The most common method for doing that involves linear regression, although at Sellforte we utilize something bigger, bolder and more advanced: the Bayesian Model (which is also used by other trailblazers such as Google). It’s worth noting that not all MMM is created equal and how effective your analysis is, depends on the features that your model employs. Below are some of the most important features that you should consider while selecting the right method for your company.

Frequency

Have fast can you adapt to changes and align your activities with the current context? In most cases monthly updates are sufficient interval to review performance, but for some staying at the top of their game require weekly or even daily checkups, especially if agile marketing and growth hacking are on the to-do list. Try not to settle for quarterly (not to mention yearly) modeling projects – you’ll end up living in the past while the life happens, well, you know, today.

Data size

This part is more about the amount of specific data, not about the number of independent variables you have in your model to explain the results. More data (sales & marketing data) from a longer period of time enables the number cruncher to recognize how individual activities drive up the results in different scenarios (which eases the sales attribution), what kind of seasonal effects there are (spoiler alert: there’s more than just the Christmas season), what’s the base sales (the amount of turnover the company would have made without any marketing activities) etc.

Moreover, go for as granular sales and marketing data as possible. We’re talking about receipt row data, campaign flagging,media items and so on. The more granular data you have, the more accurate results you’re able to derive from it.

Modeling dimensions

Attributing sales uplifts is great, but it’s only one side story. Including margin data in your modeling reveals the true business impact for the various marketing activities and makes the results much more appealing to your CEO (which tends to care about the bottom-line growth above all else). And calculating the ROI based on margin uplifts is the ultimate source of truth when it comes to the question “how profitable our marketing is?”

Modeling both the sales and margin uplifts also enables you to categorize media channels and promotions based on their impact to e.g. traffic or margin drivers, or both if you’re hitting the jackpot, which is also crucial for crafting a sophisticated marketing plan.

Access & Application

Lastly, it is important to determine who are utilizing the results and are the insights implemented into everyday operations. Involving more people, both vertically and horizontally, ensures bigger impact than assigning few power users and trusting them to continuously share information on top of all the other tasks they have at hand. Moreover,breaking the silos and building bridges across marketing, sales and merchandising departments with common metrics will not only improve the communication between different functions, but can also introduce new synergies and business opportunities that might have been otherwise missed or lost in translation.

Whatever the methodology, the goal is to recognize how different activities drive uplifts over time so that you can see bigger uplifts in the future.

Take, for example, our client Gigantti -- they were receiving a lot of traffic and sales from search engines. It would be quite natural to deem this to be thanks to their SEO efforts or paid search marketing, through MMM, however, it was possible to see (primarily from the timing) that the surges in traffic from search engines were actually originally the result of their TV adverts.

See, by employing MMM they were able to understand the relationship between these two channels rather than viewing them as standalone efforts. If these two channels hadn’t been analysed together, then it is possible they would have attributed too much of their success to search and overspent on that channel, rather than putting more time and effort into TV.

And it doesn’t stop there — MMM can be used to help you decide which products are performing best (and where), what price point they should be sold at and what discount percentage should be given (if any). Marketing Mix Modeling also takes into account the difference between markets as, let’s face it, each one has its unique opportunities and challenges.

This translates into insights such as:

• What your star promotions are—those deals that drive customers to the store without sacrificing net profit (bonus points for customers who buy large, high-profit baskets).
• How each media channel performs within different customer segments.
• Your most-loved products based on each segment’s preferences.
• How much €1 in direct mail/print/TV/digital advertising brings in incremental sales and profit margin.
• Which campaigns and media channels (offline and online) are delivering the best bottom-line impact.
• Which products should/shouldn't be promoted in each campaign (and why).
• What the sales & margin forecast is for upcoming months based on your current media plan.

This may all sound like marketing magic, but it’s simply a case of analyzing all of your receipt and marketing data, adding in external econometric factors (for example, weather data) and then using advanced machine learning and AI algorithms to make sense of the complex relationships in your masses of data. Oh, and then turning these into actionable insights for you to use every day.

So, what are you waiting for?

Want to get a head start in MMM for your company? Check out Sellforte Marketing Mix Modeling as a continuous service

Curious to learn more? Book a demo.