Proof Analytics Blog

Ajay Ahuja, one of the world’s top MMM champions - Proof Analytics

Written by Niklas Karlsson | Jan 13, 2020 8:54:00 PM

Ajay Ahuja, one of the world’s top MMM champions is a Senior Business & Marketing Analytics Manager at Kellogg. Five years ago he didn’t even know what MMM was and today he is one of the best in his field, improving the business value for Kellogg with the help of Marketing Mix Modeling.

 

How did you get into Marketing Mixed Modeling (MMM)?

While working at Nielsen, I could see the value that data & analytics could provide to a company. This motivated me to explore areas other than sales.

Call it my Curiosity for more or sheer good luck, I was offered a role at Kellogg to work with MMM and later manage the MMM department. Believe me when I say this – I didn’t even know that MMM existed 5 years ago. It just happened to me.

 

Which analytics method would you say works best to prove the value of marketing?

Before I talk about methodology, which I am sure you are very keen to know, there are certain prerequisites to ensure the correct measurement is in place. Define what success looks like, set your targets, and commit the effort required ahead of the research.

The vastness of data available can easily make you go bonkers. However, the beauty is in triangulating multiple data sources to find the golden nuggets. These mini-gold mines provide valuable insights that provide a unique pathway, an ACTION plan to follow that will help achieve your success criteria (KPI).

Personally, that is why I am a big fan of MMM; combining several data feeds from different sources to develop a holistic model & generate powerful learning. You can use any type of data – From paid media to owned media, offline to digital marketing, internal factors, and competitor information.

All of these work in one beautifully crafted model to answer questions like – What is driving business growth, where should we invest? How much should we invest? When should we stop investing? Pretty cool, isn’t it? or is it just the geek inside me talking?

In my opinion, context is everything. My viewpoint above is limited to the physical retail store model but imagine if you are in the business of selling goods or services online. Then the analytics method choice would be very different. Also, the speed of incoming data and results output with the recommendation would differ.

You don’t expect one shoe to fit all ballerinas, right?

 

One thing I strongly believe in – Do not create massive data lakes without the results view:

  • Know your audience: Who will swim/sink in this lake?
 
  • Analytics approach: Which analytics will run in the lake? Moreover, the cost of setting up and regular upkeep.
 
  • What success looks like: Agree on KPIs in advance & keep a check on performance & share it too. Be proud of your achievements but also learn from your failures.
 
  • Change in organizational behavior and work style: This is the recipe for success. You need leaders of change by your side. Internal pilot champions from various areas of business for your data lake with analytics to be successful.
 

Being very selfish, I like MMM but Context is everything. What is your business (online/offline)? Which media channels should you invest in (Offline/Online)? Can you make changes to recommendations on the fly? All of these factors should dictate your analytics journey.

 

What would you say are the challenges with MMM right now? 

There are quite a few, to be honest. There are several problems with traditional MMM that pop-up during the modeling process. The first one is the input. A considerable amount of time is spent on collecting, cleaning, processing and mapping the data.

In an ideal world, media agencies, internal finance teams, external scanning sales database suppliers and others should all come together as ONE team that reduces the data collection process, simplifies analytics steps, drives business revenue and lowers the cost of non-working analytics.

“Too much time is being spent on data collection, cleaning up, processing and mapping the data before you can take action.”

Two other problems occur during the modeling and output. Let’s say that you have the input (data), and you give it to two different data modeling experts. What often happens is that due to the variety of variable transformation & selection processes, the results might look a little different. This leads to a lack of standardization in the modeling process. This is a big pain point for modelers because the results look a lot less credible.

I firmly believe that a certain benchmark or standardized process should be inculcated by a regulatory body just like in medicine or law. I think a high level of credibility in the modeling process is necessary considering the stakes are staggeringly high.

“Overall, there is a lack of standardization in marketing mix modeling”

MMM analysis and recommendations are difficult to interpret and only accessible to a certain few keen enthusiasts; the others are just devoid of the benefits of the model.

All this information should be easily accessible with benchmarking from own and competitive industry. I wish this was a possibility, but I wish for many things, I also wished for a Ferrari, not sure, when that is happening.

 

Is there any other analytics method that can measure time lag effects and offline sales, to get the whole picture? 

You could run some extremely basic analysis to measure the impact of media on offline store sales. You can also embed seasonality if budget is an issue. It is a tricky one, I experience this personally for our smaller markets where investments are small but running a full-blown marketing mix model. It might not be worth the money at all.

Again, context is the king here. But, do not use partial methods, like digital attribution for holistic challenges. Always think of Online to Offline attribution and vice-a-versa. Our business invests behind offline, online, in-store, out of the store & various other marketing levers to build mental awareness & drive profitable growth. For us, Market Mix Modelling is the right tool to measure the impact of our array of marketing activities.

“Do not use partial methods, like digital attribution for holistic challenges”

What do you think CMOs need to change in regard to analytics in 2020? 

In the past, CMO’s decisions could have been based on pure intuition and experience rather than concrete facts and numbers. Many CMOs executed stand-alone campaigns on branding and marketing alone. However, there is a wave of change now.

The CMO’s these days are more focused on customer experience and driving sales with the help of marketing. The need of the hour is for the Head of Marketing to work jointly with the Head of Sales. The increased focus on sales and customer experience has enriched the appeal of MMM in the business world.

It has truly elevated the interest in data and analytics. I believe going forward many more CMOs will have MMM as a part of their go-to toolkit.

“Listen to the recommendations and alter your actions if the data says that you are going too heavy on media and GRPs.”

What is your biggest goal in 2020?

I genuinely believe the Omni Channel is going to be a major trend for us. This is not going to be limited to just e-commerce and high-frequency stores.

It goes far beyond vending machines, coffee shops, gyms, etc. It would be beneficial for us to start measuring and analyzing these channels. We are ready to drive our Omnichannel strategy forward and make these a part of the growth pillars.

I believe that revolutionizing Omni channels measurement & expanding MMM beyond Hyper/Supermarket will account for a major part of generating profits and growing our share in this highly competitive environment.

Written by: Maziar Nodehi, Proof Analytics