I realised that this title is a bit ambiguous. There are of course lots of factors that determine if a marketing mix modelling project tastes nice, not least the experience of the team conducting the modelling. Both in terms of their selection of the best techniques but also their ability to engage the right people in a business.
But those are broad questions and what I mean to write about today is far simpler. Of all the inputs, variables, data streams – call them what you will – that are included in a marketing mix model, which group is the most important?
Is it, like the name of the technique suggests, the set of inputs that captures marketing investment – the marketing mix? Or is it a trick question? Perhaps marketing AND all the other drivers should be considered equally.
Econometricians will often rightly say that you cannot accurately measure the impact of marketing without stripping out everything else.
To clarify, the most important doesn’t necessarily mean the biggest effect nor the one that your business happens to be the most hung up on, for whatever reason.
As an analyst, I mean which input is the one to really get the measurement right for because it is the model’s anchor that determines the fate (accuracy) of all the rest. If you want a refresher on what marketing mix modelling is have a read of this first.
Sorry, forgot the spoiler alert. If I am modelling your business, then we will have selected one or more KPIs that reflect an aspect of your own business performance. Maybe your sales, subscriptions, or website visits.
But would-be consumers or prospective customers when they identify a need or desire probably won’t automatically think of your brand unless you’re operate as a monopoly. That kind of loyalty takes years and is less certain in price sensitive times.
No, they think in terms of the category of products that will satisfy their gap, then draw up a list of possible brands to whittle down as their research progresses. Google lets all of us see searches indices. Searching indicates at least some intent.
People need to decide to buy the thing before they decide if that’s your thing. The number who are ‘in market’ week by week also probably fluctuates. According to what the thing is – think ice cream versus red roses. And what else is going on – is there a recession or other distractions to make them put off a purchase?
One implication is that your marketing job may be two-fold. Remind them it’s time to buy the thing and wave the flag about your thing being the best. If the category you operate in is mature and flat the game is winning share from rivals.
Or if it’s a new category, then the task is about growing it, convincing people in growing numbers that it’s worth a punt, will make their lives better. If you’re first or second in that may be tough but the spoils could be big too.
Regardless of which scenario is you, the type of marketing that you need to run, and the ceiling impact of that marketing are both determined by the category that you operate in – its size, cycles, and underling trends.
Category demand is therefore the most important ingredient in a marketing mix model (with a few technical exceptions before any analysts wade in.) Leave it out when you shouldn’t, and you could be attributing way too much credit to marketing.
You don’t necessarily need to have the historic category-level sales as inputs, in fact unless you are an FMCG or CPG brand you probably don’t have these anyway.
In the UK, ONS is also a great starting point. For example, you can pull GDP by categories of goods and services. ONS also publishes survey results on a range of topics that can be used as proxies for likely behavioural changes over time.
There are international equivalents for many other nations, just google government statistics offices to ensure reliable sources.
I mentioned Google’s search data earlier and it’s publicly available here. There’s a bit of a knack to using it in models and you need to be careful if you’re working outside the UK to use native language terms but for a free resource it’s great.
The trick is to put yourself in the place of the consumer over the past few years and jot down the factors that have made them more or less likely to buy into the category. Now go find robust data streams that reflect these influences.
For example, we’re way beyond this point now but I remember using the proportion of people who had access to the internet at home as a proxy for the likely future growth trends in online shopping for a forecasting project once. Yep, I’m that old.
The above doesn’t mean that you should only go to town on the category inputs, merely sketch out the others and hope for the best. The same depth of thought needs to be applied to all model inputs if you want to do big things with the outputs.
But marketing mix modelling can be overwhelming. And overly focused on the marketing. I blame the name. It’s never a bad idea to break stuff down into its parts and focus on one job at a time.
If you need a hand with your next marketing mix modelling job or are just curious about what businesses use it for, don’t hesitate to get in touch.
© Jo Gordon Consulting Ltd 2022