Marketing mix modelling (MMM) is supposed to have fallen out of favour a long time ago. Starcom’s Laura Desmond wanted to nix the practice entirely back in 2011, when she went as far as to say “it will become obsolete”. The problem? Backward-looking data isn’t an accurate predictor of future performance of an ad, and the explosion of data points and media formats makes it increasingly difficult to actually do. Then there are those, such as Mindshare’s Chief Data Officer Rolf Olsen, who believe it still has a place in marketing; it’s just a case of combining models for a bottom-up and top-down approach. (A comprehensive outline of MMMs flaws can be read here.)
I spoke to Lenovo.com’s Vice President and General Manager, Ajit Sivadasan (pictured above) about how Lenovo.com uses media-mix modelling and what difficulties he counters with it.
(Sivadasan will be speaking at the Seamless 2017 conference, happening 19 – 20 April in Singapore. For more information, click here.)
How difficult is it to do marketing mix modelling in real time?
The exercise of creating a mix modelling format is easy enough; however, the process of getting it done effectively is very challenging. The challenge is not the mathematics but legacy environments, accurate mapping of customer data, first party and third party, mobile cross over, DMPs etc. So most people have models and they use it to make directional decisions at best. Of course the longer the models run, the better they become.
How expensive is it to do marketing mix modelling in real time?
The process of optimization can be expensive as it is quite data intensive and requires continued monitoring over a long period of time. The real success is the ability for organizations to make decisions in near real time and that’s always tricky and complicated, depending on markets they serve and what products and services they sell. The expense is highly correlated to complexity of the solution and intensity of the computational model and how real time it needs to be. Sorry, not a clear cut answer on this. We can run a very cheap, simple solution or a very complex, very expensive solution— it all depends on what ROI you are looking for.
Do you use marketing mix modelling at Lenovo?
We use mix media modelling to make directional decisions. The big ah ha moment we had from the exercise is that the model changes by country and geography. So how we deploy the model matters a lot when it comes to allocating marketing dollars for our demand generation activities.
If not marketing mix modelling, what would you do to monitor how well you’re spending your marketing budget?
We’d use experience to drive trial and error models using industry experts, agencies etc. Nothing beats a pilot—the issue is, it takes time and many times outcomes change when you try to scale the business. So while a common sense approach can be used, as the business complexity increases and the amount of data you need to work through gets large, it almost becomes overwhelming not to be using some kind of a data driven approach to deal with the challenge.
Sivadasan is set to join more than 200 other e-commerce, retail and payments professionals speaking at Seamless Asia. For more information, download our brochure here, or go right ahead and register here.
For speaking opportunities, please write to Oliver Arscott at email@example.com or call  6322 2727.
For sponsorship opportunities, please write to Audrey Leong at firstname.lastname@example.org or call  6322 2786.