written by
Gottfried Häuserer
Gottfried Häuserer

Media Mix Modelling – what it is and why you should be using it

Media Mix Modelling – what it is and why you should be using it

CMOs often have a basic perception of the effectiveness of their media mix based on experience but rarely do they have a coherent overall view. With Forbes Insights finding that over 80% marketers intend to increase the role of predictive analytics in marketing over the next 12 months, Media Mix Modelling (or MMM) has re-entered the spotlight as an established approach enabling CMOs to understand how media spend has impacted performance-driven KPIs and can also offer predictions on how future tactics will impact ROI.

How does your company currently determine its media mix? I still hear too many stories of decisions based on gut instinct, tales of “we’ve always done it that way” and relying too much on agency recommendations. In recent years this has become a problem: media types and sales channels are constantly diversifying and customer journeys are becoming ever more complex. Many companies do a decent job analysing their media mix but the results remain static, filed away in a drawer never to resurface.

Enter Media Mix Modelling, which Forbes estimates can help to optimise media spend by up to 25%. This smart-data approach collects various data sources (sales and marketing data from the advertising company, of their competitors, of the market, macro-economics, weather etc.), then uses statistical models to link historic media spend to business performance (classic KPIs are sales or revenue). The analysis then informs a set of recommendations which help to optimise media spend in the future – including ideal campaign period, budget/media pressure and the right media mix. It forms a complete view of marketing performance at the strategic and campaign level, can help to reallocate budgets between channels, tactics and media placements, models the impact of future marketing and media investments and helps to drive the highest ROI across channels and products.

MMM is a particularly useful tool for companies which invest in both on- and offline media and particularly in TV. TV advertising has a huge impact on a company, both on sales and branding. But the true ROI can only be seen over longer time periods. With MMM it is possible – in contrast to most other analytical tools which focus purely on the short term – to see the impact of TV advertising over a longer-term basis.

Alt-Text: Media Mix Modelling, predictive analytics, media channel optimisation, media performance

What are the top factors for success?

Alt-Text: Media Mix Modelling, predictive analytics, media channel optimisation, media performance

The whole organisation from the top down needs to be firmly committed to MMM to ensure its success. There will be winners and losers from the model so all departments, from sales through marketing, media, product and communications, must actively live the MMM process.

The project would ideally be sponsored by a c-level executive and led by a director with a mind for analytics and a reputation for objectivity. The team needs to start small and aim for proof of concept, building limited scope models which achieve early wins. A lead time of at least one to two years is required for this until the full model is in operation. These models should be tested aggressively and the results fed back into the model. It is crucial that model insights are shared and discussed continuously with all relevant stakeholders on company- and agency side. In reality the ideal media mix will always be a compromise of advertiser driven limitations, agency constraints (commitments) and recommendations as well as the model output.

It is also vital that data is provided in a fully automated format to ensure that the whole modelling process can handle the data in a timely manner. This is fairly straight forward process for online channels, but less so for offline channels. Manual entry of data just isn’t realistic given the amount of data required for a valid MMM.

What are the limitations of Media Mix Modelling?

Media Mix Modelling is an effective way to help companies to optimise their media spendings, but there is no crystal ball in the marketing world. Take examples such as Dieselgate in the car industry, volcanic eruptions in the travel industry and PR crises in any industry – these all have a massive impact on marketing campaigns but cannot be predicted. No algorithm will ever be able to predict these.

All insights are based on data and a model can only ever be as effective as the information that has originally been submitted. An effective MMM requires extensive and accurate data inputs over a statistically-valid time period and a one-off data sample simply does not offer the complexity of information required to build an effective model for the future. If there are limited data points or cases then companies must always consider whether the prognosis is valid. Furthermore gathering enough data from across the organisation, and from partner agencies and external service providers and ensuring data collection and provision is built into these contracts, is an onerous task.

What’s next for Media Mix Modelling?

These are exciting times for MMM. Modelsare becoming ever more granular, with prognoses on a daily rather than a weekly level. Models are looking to go down to creative and publisher-level with recommendations based, for example, not only on TV or broadcaster level but on daytime or programme level. We will also likely see the marriage of different attribution methodologies in the immediate future.