MMM delivers a true holistic view of the performance of all marketing & media tactics.
It is a popular analytics methodology as it gives you the confidence to make future media investment decisions based on statistically robust models, allowing investment to be reallocated in a more efficient way
It is based on econometrics - a statistical method designed to measure the relationship between key drivers (e.g. advertising, economics, weather, promotions) and the KPI - such as sales, web visits, search or brand metrics.
The most intuitive MMM results come from the perfect blend of tools/technology & human brain power. MMM shouldn't be the only piece of the jigsaw - it works best when used alongside other analytics & research
If you want to understand the impact your marketing activity is having on your business (especially through MMM) a great starting point is to consider:
What are your key business KPIs? And hence what are you tracking (or should you be as a business) to determine success?
Are the objectives of your marketing activities designed to drive these business KPIs or do you have separate KPIs for marketing? If separate do you have the data to track your marketing KPIs?
MMM can be an incredibly powerful tool, but only if it is pointed at measuring the impact of your marketing activities in the right way on the appropriate KPI!
I can’t emphasise enough the importance of reviewing the data before commencing an analytics project. Charting and analysing the data is vital!
1. You can instantly spot issues in the data – whether these are missing data points, misaligned data, odd data, or differences to previously supplied data
2. Visualising the data helps when you get to the modelling phase – you can start to see potential relationships and thus know which variables to test
3. Charting the data helps you refine your list of potential new variables to create & test. Why create 100s of multiplicative media variables for example when looking at the data clearly shows you that most of them are not viable – not enough spend, not enough variation etc…
4. You know up front which of the business questions you are likely to be able to answer vs not answer by knowing your data – thus you can set client expectations
The better the analyst knows the data the better the models!
Does my media activity drive my business?
If so does it pay back?
Where would my business have been if I hadn’t invested in media?
How are the individual media channels working?
Which drives the highest volume and which deliver the best ROI?
Are we spending too much or too little in any channel?
How can I optimise my media budgets to drive more business?
How much should I be spending?
What is the optimal mix of channels?
How are other elements of my marketing strategy working at driving my business?
How effective are price & promotions?
What impact is my CRM having?
How do external factors influence my business?
Is there an underlying seasonal pattern to my sales, or do sales fluctuate with weather?
What impact has the economy had on my business?
Automation can be quite a controversial area in MMM and there are many different views on its use – to be fair there isn’t necessarily a right or a wrong … in my view its about the how, when & how much - which will depend on the needs of each project and client
One way or another automation plays a vital role in MMM. It can, and often is, used to enhance efficiency - speeding up data collection, data processing and data validation, building basic models (to further) enhance manually, or automating outputs
Although I do mostly sit in this ‘automation for efficiency’ camp, I do also recognise that there are cases where automation becomes far more important - such as the need for speedy & regular model updates, and projects that require a vast quantity of models to be built to name a couple … automation in these cases becomes a necessary and integral part of project delivery
My experience over the years has involved using automation to help improve the efficiency of my projects. In these instances I am a firm believer in the fundamental benefit the human brain can bring to the process of the modelling and the drawing out of insights
Our understanding of our clients and their markets, our ability to analyse the data and follow it down various paths to explore the questions we need to answer can lead to some of the most impactful insights that ultimately come from having been up to our knees in the nitty gritty of the models – you can not only see the findings come to life but also refine the hypotheses to test
These kind of insights add real depth and impact to a results deck. Of course I wouldn’t want to go back to matrix algebra approach of my university days (#showing my age) and automation definitely plays a vital role for me, but I am at hearts a hands on modeller