Increasingly, success is tied to tangible metrics like bottom line results. CMOs need to show how various advertising and marketing tactics got someone to buy their company’s product, as well as whether they led to softer measures like raising people’s awareness of a brand. This evidence must convince CFOs – who still view marketing as a “cost center” despite the best efforts of CMOs – to maintain marketing budgets.
One approach is to use marketing mix modeling, which allows CMOs to show company leadership how their efforts contribute to the bottom line. “CFOs love it because a lot of analysis is done in silos,” said Jon Turner, global director of analytics at Mediahub, adding that those silos can add gaps in reporting. “With marketing mix modeling, you’re looking at it holistically, so it can’t explain more than what your sales actually are. It explains all sales and allocates them to different marketing levers. »
Sure, but what is marketing mix modeling?
It’s a way of using statistical analysis as a tool to look at sales over a period of time to determine exactly what caused those sales. Essentially, it’s a way to help marketers and agency executives contextualize what works and what doesn’t. For example, suppose a marketer who typically spends the majority of his advertising dollars on television reallocates that spending to digital channels and offers a reduced product price. If that approach represented higher sales numbers, then that marketer could take that analysis, modify their approach, and optimize it to spend more of their budget on what works and less on what doesn’t.
It seems like an obvious thing to do. How it works?
Marketers and agency executives enter data into analytics based not only on the marketing tactics they use, but also on each activity a brand may deploy or encounter. Thus, they not only count digital, television, outdoor, radio, podcast and social media advertising, but also the price of a product and various ongoing promotions. Of course, that’s not all. It would be too easy. They also take into account factors such as inventory levels, seasonality, even changes in weather conditions – basically anything that could impact sales. This data is then compared to previous sales data, often at least three years old, to show how sales have changed and give a reason why they have changed. It is correlation rather than causation.
If that sounds like a vague synopsis, well, that’s because it is. The model is specified for each brand and should take into account anything that would cause sales spikes for valleys.
OK so that’s just another attribution method. Big phew.
Well, yes and no. While it’s a way for marketers to indicate a reason for selling, it’s also a predictive model to help marketers make decisions for the months ahead. Marketers will use the analysis – often on a quarterly basis – to see what changes are happening and move the dollars around to hopefully continue the positive trends. If the model shows that a particular channel is performing better, it will likely move more marketing money there. Take away, for example. As people started traveling and commuting again after the shutdowns, it’s become a more useful channel again, so marketers are spending more on it.
But you just mentioned the pandemic. Doesn’t that throw a wrench in the whole thing?
In some ways but not really. This is why marketers use a few years of data for marketing mix modeling. “When you have a system shock like Covid, having years of data becomes even more important,” explained Larry Davis-Swing, vice president of advanced analytics at Spark Foundry. “By having a lot of data before and a lot of data after, you can start to understand and isolate everything that you saw happening during Covid.”
Davis-Swing continued, “When the markets closed, we saw consumer behavior change. People have moved from restaurants to take-out and delivery. We have seen delivery explode. So we can explain this initial explosion not because of advertising or marketing, but because consumers had to change their behavior. »
So yeah, the data from mid-March 2020 through late 2020 — maybe even summer 2021 — is a little fuzzy because consumer behavior has changed significantly, making it harder to make predictions. However, as people leave their homes and resume their pre-pandemic activities, marketers can then weigh 2019 data on the upside and factor in more normal behaviors to help future predictions be more accurate. .
That’s why you need to make sure the entries are correct.
Exactly. Marketers and agency executives need to think about anything that could explain the variation in sales so the model can work well and help predict how they should split their marketing mix. If you have a model that tries to explain the variation in champagne sales, you’ll want to capture a spike on New Year’s Day and Valentine’s Day, explained Trisha Pascale, analytics group director at The Many. If you ignore this, the model could be inaccurate and its predictor useless.
It is also important to consider changes in marketing and advertising strategies. With the turnover from one CMO to another, which tends to happen every 18 months or so, there is often a change in strategy. If you haven’t factored in more digital advertising or whatever change in marketing mix modeling, it won’t show how that change is working.
OK, but aren’t you using a bunch of data. What about cookie death? Won’t that be a problem?
Unlike multi-touch attribution, marketing mix modeling is not performed at the consumer level, so the more personalized data that might disappear with the death of the third-party cookie is not as important to mix modeling. marketing.
“We’re talking about really big trends, and we’re not building those patterns at the consumer level,” said Michael Salemme, vice president of analytics at Zenith. “There are ways to run aggregate data to keep working [marketing] mixed modeling. We try to explain changes in sales usually at a national or regional level, so we just need to know the approximate exposures.