Navigating the complex maze of media mix modeling in today's multifaceted market can be a daunting endeavor.

My journey at Autodesk, where the landscape consists of a bifurcated business model pivoting on distributors and online sales, has been a testament to the importance of leveraging data and optimizing media mixes to drive tangible results.

I’m Siara Nazir, and one of my greatest strengths is analyzing and unearthing the granularities of data to ensure our conversion strategies not only resonate with the macro-perspective but also trickle down effectively to the specifics, ensuring a robust strategy that withstands the undulating tides of the global market.

In this article, I’ll delve into:

The origins and evolution of media mix modeling

The origins of media mix modeling can be traced back to the 1970s, starting with a certain statistician professor at MIT.

Despite becoming a topic of more common discussion only in the past decade or so, this strategy has been embedded in the marketing world for quite a substantial period. The MIT professor crafted a mixed model, aiding marketers in deciphering viable investment options.

Slide of the Media Mix historical timeline beginning with 1970s with John DC Little, then 1989 with Marketing Management Analytics, then 2005 with MarketShare and finally 2023 with a screenshot from a list of hundreds of marketing tools.
Courtesy of Siara Nazir and Autodesk

As we catapulted into the 1990s, media mix modeling began gaining momentum when various companies formulated offerings that facilitated other businesses in accelerating their marketing ventures.

It was a significant era – the burgeoning of the internet altered the media landscape, shifting focus from traditional mass media (such as television, print, and radio) to a novel medium, enriched with behavioral marketing and new segmentation available to marketers.

Companies like MarketShare emerged, delving deep into experimenting with media mix modeling – companies with which I've had the pleasure of working.

Fast-forwarding to today’s Martech landscape, media mix modeling has nestled itself snugly within the framework of data analytics, media, and, of course, modeling.

The red box outlined in our current martech landscape highlights where media mix modeling resides – not isolated in its own segment but rolled into the myriad offerings companies now have surrounding data analytics and media. This conflation into a cohesive section was non-existent just a few years ago.

Peering into the future, I foresee artificial intelligence (AI) and machine learning (ML) not only amplifying the number of companies in this sector but likely birthing a new sector concentrated around predictive modeling. This, I believe, represents the forthcoming evolution where marketing is headed.

Diving deep into media mix modeling and its potency

Let's demystify what media mix modeling truly is. At its core, it's an econometric model deeply rooted in statistical methodologies, and to truly understand its prowess, it’s essential to distinguish it from conventional attribution methods like last-touch or multi-touch attribution.

What is media mix modeling? A method rooted in statistical analysis that helps businesses understand how well their marketing investments are doing.
Courtesy of Siara Nazir and Autodesk


While the latter are principally predicated upon a customer action – such as viewing a banner ad or clicking a link, and accordingly, credit is assigned in the attribution system – media mix modeling operates a tad differently.

It thrives on correlation: the association of variables with another variable.

Consider this analogy: if it rains, people are likely to carry umbrellas. The presence of rain correlates highly with the emergence of umbrellas.

So, how does this translate into a powerful marketing strategy? Media mix modeling allows us to scrutinize non-action-oriented variables like brand equity.

Taking companies like LifeLock or Stitch Fix as examples, ponder the worth of the brand and the equity it hauls in. A media mix model has the capability to quantify this, establishing what we term a 'baseline'.

Upon establishing this baseline, the enthralling journey of experimentation commences. This involves exploring queries such as:

  • What happens to the incremental revenue above the baseline if the display or social inputs are escalated?
  • Is it effective?

This realm of experimentation is not merely intriguing but is a hotbed for innovation and strategic fine-tuning in marketing endeavors.

Visually conceptualizing it: you have all of your channels – and it’s noteworthy that many media mix models can incorporate offline data too.

If your product is retailing at Staples or other outlets, that data can be synergized along with your online and television data. The data funnel through, purchases are executed, and the model regurgitates what it perceives as the attribution.

Therein lies the playground where marketers delve in, optimizing their mix, and that's essentially the back-end mechanism of how it all intricately works.

Global use cases: Tailoring media mix modeling to unique market dynamics

Embarking upon our journey with media mix modeling, we initiated experimentation both with a small company and in-house, mindful of the budgetary constraints that many entities navigate.

What transpired was enlightening: our in-house experimental findings closely mirrored what the external media mix modeling produced, bolstering our confidence in our strategic maneuvers. Yet, like all ventures, it was, and always is, a cycle of testing and learning, with myriad factors influencing outcomes.

Creating the media mix commenced with a critical step: aligning it closely with the nuances of the market or country in focus. Tailoring our mix to resonate with the demographic and socioeconomic dynamics elevated our conversions by an impressive 23%.

This success was birthed from our investment in understanding the customer first, and establishing our goals clearly – be it to amplify awareness, drive conversions, or propel trials – before crafting the mix.

For instance, during our experimentation in Germany, we discerned that customers harbored a penchant for data and extended web page content, a stark contrast to the US market, where brevity and concise choices are paramount to maintaining customer engagement.

Consequently, our media mix was meticulously tailored to each country: the US, Japan, Brazil, and Germany, each benefiting from a unique, custom-crafted media mix strategy.

Socioeconomic data was instrumental in enabling us to comprehend customer learning and decision-making processes. In Germany, the proclivity of customers to turn to groups for information and decision-making was noteworthy, marking a departure from the more individualistic tendencies observed in the US.

Thus, in Germany, strategic partnerships, co-branding, and group engagements were remarkably effective, juxtaposed with the US strategy, which honed in on ensuring that customers felt self-sufficient and empowered in their decision-making journey.

In our exploration of media mix modeling, the Japanese market surfaced as a particularly interesting case study due to its distinctive social media usage patterns.

Let me share a fascinating journey that unfolded over five quarters, demonstrating not only the intricacies of media mix management but also the patience required to observe tangible results.

Customizing your mix to the region - two tables that show mobile phone usage across different countries. It says, since 2020, mobile phone usage has more than doubled in countries like Japan and social media usage is one of the highest.
Courtesy of Siara Nazir and Autodesk

Historically, with our marketing approach somewhat generalized, employing a somewhat universal strategy across different markets. In the US, for instance, we primarily concentrated our efforts on paid search, dedicating approximately 70% of our resources there, with the rest being distributed amongst various other channels.

This strategy was, to an extent, replicated in other markets. It was both a logical and convenient approach, yet it lacked a tailored specificity to each geographical locale and their inherent consumer behaviors.

When we shifted our focus to the Japanese market, we recognized the necessity for a more nuanced and experimental approach. The first quarter mimicked the US-centric strategy, heavily favoring paid search.

However, as we moved through subsequent quarters, we started meticulously diversifying and fine-tuning our strategy.

Undertaking this endeavor manually, without a predefined model, posed its unique set of challenges. For instance, uncontrollable variables like unexpected campaigns from other internal teams or unanticipated market fluctuations could introduce an element of 'noise' into our data, potentially skewing results.

Manually navigating through this demanded a level of vigilance and continuous adjustment to preserve data integrity and reliable insights.

As our strategy in Japan evolved, we introduced digital out-of-home into our media mix, responding to the country's substantial engagement with social media and technological integrations within their daily life.

This adaptation not only substantially expanded our reach but also significantly heightened our brand awareness in the Japanese market. This, in turn, enabled us to leverage this amplified visibility to generate new leads and further establish our presence within the market.

In retrospect, this journey through evolving our media mix strategy, particularly within the Japanese market, provided enlightening insights and learnings into the importance of customized, geography-specific approaches in global marketing strategies.

Analyzing success and challenges through test results in global markets

Embarking on our media mix modeling journey, we engaged with varied Key Performance Indicators (KPIs) such as conversions (primary KPI), traffic, and market penetration as secondary KPIs to measure and understand the impact of our strategies across different markets.

Autodesk, with its business model, bifurcated between distributors (encompassing 70% of total revenue) and online sales, offered an intriguing platform for our media mix experimentation.

The aim was to gauge the growth of online business against the wholesale market and scrutinize if our media mix manipulations were facilitating an enlargement of our slice of the revenue pie.

Media mix testing across APJ - a table that is the penetration of revenue in-market across ANZ, Japan, Korea, Singapore, Malaysia and India. ANZ is significantly higher and reaches a steep incline from Q1 of 2022.
Courtesy of Siara Nazir and Autodesk


In the Asia-Pacific region, introducing a new media mix started demonstrating perceptible growth in our online business. However, it's critical to note that not every alteration in the media mix guarantees an uptick in growth.

This was especially evident in our ventures within Europe, where various factors introduced a heightened level of volatility and 'noise' in the data.

For instance, the United Kingdom, in particular, did not observe the expected positive uptrend, necessitating a reevaluation and iteration of our strategy therein.

Media mix testing across APJ - a table that is the penetration of revenue in-market across UK, France, Germany, Italy, and Spain. France is significantly higher and reaches a steep incline from Q1 of 2021. And while the other countries do continue to increase, the UK drops and there is a note to say "UK didn't do as well as other countries".
Courtesy of Siara Nazir and Autodesk


Navigating through the variances of each country within Europe posed its own set of challenges, primarily owing to the distinctiveness and inherent variabilities within each market.

Despite sharing geographical proximity, the intrinsic market dynamics within each nation demanded a more precisely tailored approach, requiring constant adaptation and adjustment of our media mix strategies.

Key outcomes, surprises, and learnings

Reflecting on our results, several observations emerged: our revenue penetration did, in general, witness record highs (barring the UK), and a satisfying 23% lift in conversions was noted.

Traffic, albeit a secondary KPI, also saw an uptick. It is pivotal for us due to our intention to captivate the 'unknown audience' and foster a sustained engagement with them.

Given that Autodesk’s purchase cycle is notably extended, often spanning several quarters, sustaining traffic and subsequently nurturing these potential leads toward a conversion becomes paramount.

Furthermore, the implementation of our new media mix, particularly the incorporation of digital out-of-home, enhanced our email-to-traffic capture rate. This was crucial in enabling us to secure and nurture leads at a more elevated rate than before the new media mix implementation.

This journey, interspersed with both successes and challenges, underscored the paramount importance of perpetual iteration and adaptability within our strategies.

From it, we’ve learned that while one can enter the arena armed with copious data and strategic foresight, the true mastery in utilizing media mix models lies in the ongoing process of learning, adapting, and optimizing in response to the ever-shifting global market landscapes.

In the pursuit of effective marketing strategies, especially when working within budget constraints that prevent collaboration with prominent media mix modeling software companies, adopting a do-it-yourself approach to media mix modeling can be both enlightening and economically viable.

Here, I present a five-step guideline to maneuver through this intricate process, based on my experiences at Autodesk.

1. Define your goal and primary KPI

It’s imperative to commence with a concrete goal. Your end objective significantly influences your media mix – a strategy focused on revenue generation will diverge considerably from one aimed solely at raising awareness. Understand what you're trying to achieve and let that guide your strategy.

2. Understand your market

Equally crucial is a comprehensive understanding of your target market. How do they make decisions? Where do they seek information? Your media mix should be meticulously tailored to align with your market’s decision-making and information-seeking habits.

3. Adopt agile marketing

During my tenure at Autodesk, establishing an agile marketing function proved invaluable. It facilitated quicker market entry and allowed us to undertake various tests in two-week sprints, enabling us to 'fail fast', learn, and then pivot to more promising strategies without additional resource allocation.

Regular stand-ups and a collaborative approach across paid search, content, and display teams, among others, were key in this agile marketing methodology.

4. Ensure meticulous measurement

As I often advocate, clear, goal-aligned measurement is paramount. If your measuring parameters are not in sync with your initial objectives, gauging success or executing effectively becomes a challenge.

5. Evade potential pitfalls

When manually managing your media mix, avoiding overly 'noisy' markets or those experiencing economic turmoil is advisable, as these can distort your results. Moreover, being mindful of potential disruptors, such as new company initiatives or creative outputs during a media mix test, is crucial to avoid confounding your data.

Key considerations and potential hurdles in media mix modeling

A few pivotal factors to keep in the forefront:

  • Data significance: Even if early data suggests a 'win', resisting declaring it as such until it achieves statistical significance is crucial.
  • Understanding halo periods: Different media mediums have varied halo periods – the delay between implementation and observable impact. For example, our display campaigns often resulted in increased trials, which have a 30-day duration.

    Thus, understanding and accounting for the actions generated by each channel and incorporating the entire buying period into your analysis is essential.
  • Continuous iteration: No strategy will be perfect from the outset. The introduction of agile marketing within my team was specifically aimed at constant optimization. Whether results are good or disappointing, the process of refining, back-testing, and deciding whether to persist or pivot is crucial.

While no path in media mix modeling is devoid of challenges and learning curves, adopting a structured, informed, and agile approach – one that is perpetually open to iteration and refinement – is intrinsic to navigating and harnessing the potential within this complex, dynamic marketing landscape.

This article is based on Siara's presentation at the Revenue Marketing Summit in Las Vegas in 2023 when Siara was the Head of Growth Marketing at Autodesk. She has since changed roles.

Catch up on this presentation, and others, using our OnDemand service. For more exclusive content, visit your membership dashboard.