This highly technical and advanced method of measurement may be perceived as a black box, but it delivers a higher magnitude of benefits than the cost of doing the analysis.
The market for effectiveness products has become increasingly more complex. Brands are now viewing tools such as digital attribution as less useful, particularly with the eventual phasing out of third-party cookies. As a result, marketers are gravitating back towards Marketing Mix Modelling (MMM), or econometrics, as the next new (old) shiny toy.
The fact is that MMM is a highly technical, advanced method of measurement. As such, it can often be perceived as a black box with limited opportunity for stakeholders to have input and offers a general lack of transparency for clients. This is not helped by the industry’s persistence in using multiple terms for essentially the same thing e.g., MMM vs Econometrics, Return on Investment (ROI) vs Return on Ad-Spend (ROAS).
As overall marketing budgets tighten and brands compete for their respective slice of the pie, there’s a lot of weighing up to be done on the measurement tools available and their respective costs.
For most businesses, MMM is undeniably worth paying for. MMM delivers a higher magnitude of benefits to a business than the cost of doing the analysis. And it’s important not to forget that the cost of MMM needs to be proportional to the potential for efficiency/effectiveness gains. This will depend on multiple factors, such as the size of the media budget, and the size of the brand.
Larger brands, with larger media investments, tend to have more to gain from measurement-produced media efficacy improvements. The opportunity for improvement will then inform the most optimal modelling frequency (quarterly, or annual), and whether the efficiencies gained by modelling at a granular breakdown e.g. splitting sales by category and retailer are worthwhile given the cost.
The next challenge is choosing the right MMM provider. Generally, there are three options: the media agency’s MMM solution; an independent MMM provider, or in-housing an MMM solution. Each option has different merits, which may be more suitable to different businesses depending on the existing planning tools and processes the models need to be integrated with, the size and complexity of the models and, of course, cost.
With MMM’s resurgence as the measurement technique of the moment, it is important to be cognisant of some key considerations when weighing up providers:
Although MMM models are complex, there’s no reason why the modelling itself should involve proprietary information. Aside from the modelling software, which is often licenced from a third-party or built using open-source platforms, the underlying modelling methodology is available from multiple open sources, and the data ingested largely belongs to the client.
A provider should be happy to demonstrate or show all the statistical tests conducted when building the models. As well as sharing content that effectively explains the decisions made around model structure or variable choice without getting bogged down in technical detail.
A strong MMM provider will also encourage the key stakeholders to be involved at every stage of the process, particularly for the first wave of modelling. Ensuring that the technical analysts are involved during data ingestion and inviting the finance director to examine the interim model outputs. Being able to ‘kick the tyres’ leads to better models that are a more accurate reflection of true business performance.
The default recommendation when designing the model structure shouldn’t be ‘individual models for each stock-keeping unit, across each retailer, across each region, across each sales channel’.
The focus should be given to the types of critical stakeholder questions the models need to answer to inform the modelling scope. For example, if the client is exclusively interested in using the models to inform the most optimal media mix for next year’s budget, without any intention of using the models to identify an optimal pricing strategy, then modelling at an aggregate sales level may suffice.
This does not mean certain effects are ignored, it still needs to account for variations in sales due to price fluctuations. Their function will be controlling factors in the model as opposed to being able to extrapolate learnings that the client is not primarily interested in.
The competitive advantage from MMM does not come from ‘better models’; it comes from more people using the models more frequently to make better planning decisions. Be careful of selecting providers whose main claim is to have the most rigorous approach in an industry where everyone broadly uses the same techniques.
Emphasis should instead be placed on ensuring the rich and powerful outputs from MMM are integrated with planning decision making & relevant tools. This means that the outputs of the models drive more effective plans for the future and are not just ‘marking homework’.
With an unwarranted reputation of being slow and labour-intensive, it’s all too tempting to turn to simple, and cheap automation efforts to accelerate the MMM process. However, purely adopting a cheap automation solution results in second-class models which generally contradict established economic and marketing theory.
If a very sophisticated AI trained on live data is being used that regularly updates the models until it reflects a true representation of the drivers of a business, this is worth pursuing. However, it generally comes with a similarly impressive price tag.
There is a far greater opportunity to accelerate MMM project timelines by focusing on the part that is most prone to breakdowns, data collection. This critical stage of the process is susceptive to significant delays, and human error. Automation of this part of the process could reduce typical timelines by three-four weeks and would improve model accuracy also.
The answer isn’t always MMM. The models should form a core part of your overall measurement framework but shouldn’t be the only tool used to inform marketing effectiveness. MMM providers should have a good understanding of how to calibrate outputs from multiple analyses.
This next new (old) shiny toy can certainly be an invaluable part of the measurement toolkit for brands and agencies alike, especially in the current climate. However, as with everything — due diligence and the right research and insight is needed when selecting the right tool and provider. The effort is worth it.
Article originally published in The Media Leader.