Real Time MMM for CPGs: fantasy, fallacy or entirely feasible?

While Marketing Mix Modeling (MMM) is once again in the spotlight, for some, it is still mired in pre-digital, traditional media preconceptions. It takes too long. It’s done annually at best. It’s backward looking and self-congratulatory. It doesn’t mean anything after 11 months. Fair? We think not.

However, many MMMs are slow, lack comprehensive coverage, suffer from access to walled garden, siloed and missing digital data, and are mismatched across media types and timeframes. Especially for expected seasonal fluctuations.

Of course, timely, consistent data can be difficult to collate across a breadth of sources, platforms and partners. But given things change frequently for CPGs, a traditional annual MMM just isn’t going to cut it.

It’s also not necessarily the case that real-time MMM is the answer. Rather that it’s a case of matching measurement frequency and granularity to the cadence and detail of decision-making and associated action-taking. in some cases, you may need to aim for near real-time, and for others, the cost of doing so would outweigh the benefit. Or the information overload would hinder rather than help. There is no one-size-fits-all.

But what does that mean exactly?

Let’s start from a simple premise of a monthly MMM. Which is at least six times more frequent than most.

And let’s consider three different measurement requirements – all of which are genuine Ekimetrics implementations that have user-friendly, real-time dashboards:

1.      A Coffee Chain that sells hot and cold drinks, baked goods, hot food and more.

2.     A Global CPG company with thousands of brands and products.

3.     A brand’s need for weekly incrementality but where a monthly MMM would be overkill.

‍

The Coffee Chain example

This national coffee chain prefers to stay flexible in its budget allocation, making decisions week-to-week so that it can respond with agility to changing market conditions with a changing marketing calendar. A traditional MMM would struggle to support this, but other measurement methodologies aren’t geared up to support a holistic view of marketing activity.

So, we implemented monthly MMM cycles to reduce the time to insight to match their desire for week-to-week flexibility over campaigns or menu segment promotion. This has generated a large database of campaign reporting that allows them to see how campaigns are working, but more importantly, how campaigns are interlinked, and which campaign attributes are performing across 60+ campaigns annually. And in turn, this allows for regular course corrections, optimizing across multiple dayparts, to drive better future results.

‍

The Global CPG Company

With hundreds of brands, the majority of which are FMCG, the number of moving parts across product sales and marketing activity is mind-boggling. As is the speed at which things can change.

In this environment, speed of response is everything when it comes to maintaining the competitive edge. Which is why we implemented a near-real-time approach to MMM, performing daily promotional computations and refreshing models every four weeks with a two-day turnaround. The models are then deployed directly into user-friendly dashboards with insights across media efficiency, ROI, promotion evaluation, and annual planning and optimization runs.

‍

Weekly incrementality

In some circumstances a monthly MMM would be overkill, but a brand still needs to look at weekly incrementality figures as they come in.

Here, we have implemented Digital Monitoring designed to overcome an Attribution blind spot. This ingests weekly Digital performance and engagement data using MMM results to transform this data for enhanced insights, where the digital data is contextualized in the wider marketing ecosystem. The weekly figures are pushed to an MMM dashboard to ensure all critical KPI metrics, including digital metrics such as impressions, are monitored in a ‘one-stop-shop’, with a clear view of incrementality.

So how did we do it?

‍

Delivering Real-Time MMM

The reality is that the MMM itself is not the limiting factor.

More often than not, it is challenging those preconceptions and securing buy-in, especially to meet some of the specific data and technical requirements to achieve increased MMM frequency. The level of granularity in insights is often limited by media spend and data, rather than modeling capabilities.

How come the MMM itself isn’t the problem? In today’s world of computational power, AI and advanced techniques, Hierarchical Bayesian modeling – which is used to capture multi-level data dependencies and uncertainties and produce more robust results – hugely reduces modeling time. For both the coffee chain and the CPG, this is how we achieve a two-day turnaround.

Where things almost invariably get tricky is data. But it’s not insurmountable; strong buy-in, relationships and engineering need to come to the fore.

Strong buy-in from the brand’s senior leadership and their media agencies will ensure there is the impetus for timely data provision. Strong relationships are needed with major platforms to ensure you can establish direct data feeds from the likes of Google, Facebook and Amazon. And strong engineering ensures the data is available when needed.

In the case of the global CPG company, that meant building the entire solution in the client’s IT environment to ensure speed to insight and security, with direct links to the client’s data lake for continuous updates and the integration of sell-out data, Nielsen reporting and various digital partners.

‍

MMM Transformation: Five Key Considerations

It’s essential when conceiving projects looking to achieve real-time or near-real-time measurement, to think of them as transformation programs - where organizational considerations are as important as data, technical and modeling ones. Be sure to include in your thinking:

• What measurement approaches suit your industry. Consider the specific needs of your business and its key     stakeholders, including how frequently and at what granularity you need your measurement for different applications.
• How you will balance the trade-offs between the speed and depth of analysis. And how and when you will supplement those trade-offs with deeper dives.
• The integration of multiple data sources. To ensure you’re able to deliver a holistic view of media performance.‍
• How you will streamline data collection and processing. To deliver faster insights.‍
• How you will achieve multi-stakeholder buy-in. Including engagement and program management needed to secure project backing, data provision, insight delivery and adoption. And whether you have the skills you need to cover all of the bases.

Real-time or near-real-time MMM is eminently possible, and highly desirable for brands with rapidly changing landscapes. But it must be the right version of it for the brand in question and it must be approached with commitment.

As for anyone who tells you MMM is too slow, we beg to differ. And have already delivered on the exceptions that prove the rule. The key question for other CPG brands is how long will they wait until they become the exception.