Navigating Media Mix Models, Marketing Mix Models and Multi-touch Attribution Models

Media Mix Modelling, Marketing Mix modelling and Multi-Touch Attribution modelling have all been widely discussed and written about in recent years.

As CMOs face increasing pressure on their budgets, the need to deliver marketing efficiency is paramount.

Modelling has the potential to provide definitive answers to the age-old question of what is working within the mix and where wastage lies.

It provides an independent 3rd party assessment of what is working and how to take advantage of this within the mix.

It is the ultimate performance measurement of activities undertaken with agency providers against the only metrics that matter  – Sales and other marketing ROIs.

The different models available self-classify into three categories:

  1. Media Mix
  2. Marketing Mix
  3. Multi-Touch Attribution (MTA)

Media Mix and Marketing mix modelling are often blurred together and, for some, are considered interchangeable terms to describe the same system.

However, from a purist viewpoint, there is a clear distinction.

Media Mix models only measure and report the effects of paid media within the marketing mix.

Marketing Mix models measure, track and report the effects of all marketing mix elements across the full spectrum of Owned, Earned and Shared assets as well as Paid.

Given the increasing influence of Owned, Earned and Shared in the mix, media mix modelling in its purist form lacks the capacity to validate all marketing mix elements and how they can be optimised holistically.

Media Mix and Marketing Mix models – how they work.

Media Mix and Marketing Mix models provide a top-down macro-level view of marketing effects across channels.

They ingest the detail of channels used and any other external factors that can influence business performance – i.e. weather, competitive activity, and economic factors.

Systems then analyse all data inputs to identify relationships between marketing assets and results against external factors.

By quantifying these relationships, the contribution of each action to overall sales or KPIs can be assessed for efficiency and effect.

Both models use regression analyses or similar statistical models to determine the impact of channel mix, weightings and usage against defined business outcomes.

They provide an ongoing opportunity to test and refine the marketing mix for long-term benefit and efficiencies with the capacity to understand the effect of transitory influence factors.

Over time an increasingly detailed pattern of insights emerges to provide both client and agency with the ability to drive marketing efficiency and effectiveness.

Traditionally these systems require at least two years of data for reliable patterns and insights with any degree of certainty; however, the latest systems using more refined algorithms and AI promise findings in a much shorter timescale to high degrees of reliability.

We review the new entrants and the veracity of their claims later in this article.

Multi-Touch Attribution modelling is very different.

It is a bottom-up approach used for measuring marketing efficacy.

Systems identify and assess the value of each marketing initiative by looking at the user’s actions before conversion. Individual customer journeys are tracked to find successful combinations and patterns of contact to provide learnings specific to each product based on their marketing mix and messaging.

Attribution modelling has been highly valuable in optimizing conversion through the sales funnel by identifying the contribution and role of different digital touchpoints within the mix.

It relies on more granular and detailed data on customer interactions.

Marketing Mix modelling and Multi-Touch Attribution modelling aren’t an either/ or choice.

They both contribute to effectiveness and efficiency in their own right and can be used to significant effect in combination. Marketing Mix modelling to present a holistic view of the efficiency of the broader marketing mix. Multi-Touch Attribution to enhance and improve the selection, usage and dovetailing of digital activities within the mix.

To use Multi-Touch Attribution in isolation is refining elements of the mix to work better without a clear understanding and insight into the effectiveness of the broader mix and the influence of external variables.

It assumes a definitive understanding and optimization of the marketing framework within which the customer journey takes effect.

Multi-Touch Attribution systems moving forward.

MTA systems have provided valuable insight and support to advertisers as digital options have grown and digital spending increased within the marketing mix. Selection and usage of digital streams have been rendered  significantly more effective; however, moving forward two factors cloud their potential use and value:

  1. The Death of the Cookie and increased scrutiny of Privacy.

Touchpoints on the journey will be more challenging to track without the cookie that’s supported by these systems. Issues of privacy and permission present a considerable challenge.

  1. They have improved the sensitivity and reliability of Market Mix modelling systems.

The new generation of Market Mix modelling systems is more sensitive to more readily identify nuances of usage within the media mix.

As the capacity of MTA models to identify efficiency diminishes, the capacity of MMM systems to provide insight is increasing to ‘close the gap’.

Market Mix modelling will never replicate the understanding MTA systems provide of the customer journey through the funnel but will increasingly be able to identify the usage of marketing streams most effective at courting the consumer.

Which advertisers should be investigating the use of Marketing Mix models?

Marketing Mix Modelling used to be the exclusive domain of larger advertisers, spending $10 million plus.

Today with improved AI and greater sensitivity, they offer value to any client with a spend of over $5 million.

At a spend of $5 million and Market Mix Modelling entry price of $150k to $200k, you need to achieve improved effectiveness of only 4% for the system to pay for itself.

The mix of options across Paid, Owned, Shared and Earned constantly evolves and presents many possibilities. The task of identifying and truly understanding what works has become increasingly blurred.

What worked in the past is unlikely to represent still what works best today and tomorrow.

Media Agencies have a swag of tools that can be directive strategically and deliver improved value in channel mix and usage. Still, these systems are founded primarily on audience consumption behaviours of media rather than product purchase behaviours.

Certainty on the optimum mix of activities is beyond the capacity of the media agency’s standard suite of tools.

The larger agencies have their marketing mix models, which invariably come at a price.

They vary in quality across the large Media Groups, but it remains questionable as to whether any of them are now the best in the market.

A further frustration for many clients is the independence and validity of the systems’ proposed media usage. When managed by the agency, there is no independent 3rd party validation of the directions and decisions provided. It’s the agency marking their homework.

An independent Market Mix model paid for directly by the advertiser provides genuine accountability and oversight for strategic recommendations undertaken by the agency.

The new generation of available systems undertakes training to allow clients to be equally involved in analysing results and evaluating insights the system produces.

Media Agency owned models are increasingly coming up short.

As one Media Agency CEO recently acknowledged, ‘ Good modelling is hard. Really, Hard. Building bespoke models of any merit from the ground up for each advertiser will not be cheap.’

The new entrants in the market have built Saas-based foundational models that can support fast data refresh and learn on the go and enable clients to get insights quickly and make decisions.

The best indicator of what the agencies think about their systems is demonstrated by the actions of several of the major groups. They’ve recognised their shortcomings and the benefits of a best-in-market system with client autonomy that provides an independent assessment of them as a media service provider.

The deal between Mutinex and Dentsu is the first of several to come.

The use of a modelling system and the most appropriate for an advertiser spend on their particular needs, but the key to any selection process would be the following seven factors:

  1. The quality and range of outputs/insights
  2. Sensitivity of the system to interpret findings and provide insights
  3. Reliability of insights – appropriately balancing speed and accuracy
  4. Speed/ timeliness of reporting – Real-time capability versus scheduled reporting
  5. Ease of use:
    • When ingesting and updating data
    • In the ready interpretation of findings and delivery of insights through dashboards
  6. Adaptability
    • To take additional feeds now
    • To swiftly absorb changes at both a product and market level to provide learnings in the future.
  7. Cost

No universal solution or provider works for every client. Still, several exciting new products have taken the modelling options into territory that offers swifter returns and improved certainty for most advertisers.

You can read more on how we can advise on marketing and media mix modelling as part of our Media Operational Reviews or how we can assist with the selection of the right MMM provider. Or for a confidential discussion on your needs, contact us here.