Google is putting more money into its Marketing Mix Models (MMMs) to help marketers measure cross-channel media campaigns more accurately by developing an analytics tool that takes into account sales impact based on several factors such as response and reach. 

With tools such as this becoming more popular (one study said 60% of US advertisers are using MMMs, and 58% of those not using them are considering it in the future), Google has decided to launch Meridian – an open source MMM designed to help decision makers do what they do with more accuracy to drive better results. As an open sourced resource, anyone can investigate the code and methodology, and then change the code to meet their business needs.

That makes it an incredibly flexible, and potentially powerful resource, for those with enough knowledge to really deep dive the data.

What is Marketing Mix Modelling, and why is it becoming popular?

MMMs are becoming more popular as brands respond to changing regulatory landscape regarding privacy and the phasing out of tracking cookies, so they are required to gather more first-party data and reduce their reliability on trackers that are on the way out.

How does Meridian work?

Meridian is built to last in the ever changing privacy landscape and includes the type of advanced measurements that’ll make analytics nerds drool.

Harikesh Nair, Senior Director, Data Science

“Innovations at launch include calibration with incrementality experiments, reach and frequency incorporation to link outcomes with planning, and guidance on measuring search. These innovations will be applicable to all media channels that can provide the necessary inputs.”

Set up to help people answer questions such as ‘how did the marketing channels drive my revenue or other KPI’ or ‘what was my marketing return on investment’, Meridian can provide clear insights and visualisations to help with marketing budget and planning. To do so, marketers can access data points such as indexed Google query-volume data, YouTube reach and frequency.

Meridian users can also run scenario planning and budget optimisations to support future cross-channel media campaigns. Google also says that marketers will have plenty of resources available to them to help get to grips with Meridian.

Harikesh Nair, Senior Director, Data Science

“We will provide comprehensive technical documentation, including a list of technical FAQs for troubleshooting. For some questions not addressed by existing documentation, users may have the opportunity to connect with Google and partner support for answers.”

Research has backed Google’s claim that marketers using MMMs get better outcomes. A Deloitte measurement research found that C-Level users who placed high importance on MMMs were twice as likely to exceed revenue goals by 10% or more.

After all that introduction, it’s time to run through Meridian’s key features – there’s bound to be a few that will pique your interest!

Meridian’s key features? 

Meridian supports all major MMM use cases. It’s key features include:

1. Hierarchical geo-level marketing, which contains more information about marketing effectiveness than national-level data and can yield more credible information on metrics like ROI.
2. Incorporating prior knowledge about media performance through the use of ROI priors, which can take any prior distribution. The good thing is that no additional calculations are needed. Knowledge can be taken from any available source, such as past experiments, MMM results, industry expertise and benchmarks.
3. Accounting for media saturation and lagged effects using parametric transformation functions. Saturation is modeled using a Hill function, which captures diminishing marginal returns. Meridian utalises Bayesian Markov Chain Monte Carlo (MCMC) sampling to jointly estimate all model parameters.
4. Optional uses of reach and frequency data for additional insights. As well as using impressions, Meridian provides marketers with the option to use reach and frequency data as model inputs to dive deeper into insights and deliver better predictions of how each media channel may perform with a change in spending.
5. Modeling lower funnel channels such as paid search is supported as Meridian is designed to support rational decision-making efforts by giving the option to use Google Query Volume as a control variable when measuring paid search impact.
6. Media budget optimisation to help marketers understand the best budget per channel, based on an overall budget. Meridian can also suggest an optimal overall budget based on advertising goals and provides frequency optimisation for any channel with reach and frequency data.
7. Estimation using what-ifs, which lets marketers estimate what ROI would be under hypothetical media scenarios such as increases or decreased spend on specific channels.
8. Evaluation and reporting using model fit stats, using within-sample and out-of-sample models.