Gartner: How to build a best-in-class marketing analytics strategy

Analyst firm talks through key lessons from Avon, and Polaris on lifting your data-driven marketing decision making game

Bold test-and-learn, commitment to data-driven cultural change and employing a predictive and prescriptive approach are key to making marketing analytics truly impactful on the brands it’s being applied to serve, says Gartner.

During the analyst firm’s Gartner 2020 Marketing Symposium, Gartner senior director and analyst, Jason McNellis, dived into marketing analytics as a growing priority investment and what it takes to make it sing within an organisation.

According to Gartner research, 86 per cent of marketing leaders want to increase their marketing analytics maturity, yet only 7 per cent say they’re operating at master level. What’s more, despite growing investment into marketing analytics as a capability, only 37.7 per cent of CMOs surveyed by Gartner are using available or requested marketing analytics before a project decision is being made. It’s a figure that’s only nominally increased in the last eight years.

Yet harnessing marketing analytics pays commercially, McNellis said. Gartner figures show having one or two types of marketing decisions significantly influenced by analytics results in 40 per cent financial overperformance, while three or more marketing decisions influenced by analytics results in a 50 per cent financial overperformance.

So what can CMOs do in order to improve their marketing analytics game? Here’s Gartner’s three key steps.

Step 1: Embrace test and learn

McNellis pointed to accommodation platform giant,, as a great example of how test-and-learn has been embraced to drive positive growth. What many don’t know is that the majority of the tests the brand does ‘fail’, with what was proposed not outperforming what the business is doing today.

“It’s not just because the team is throwing crazy ideas against the wall – it’s the opposite. The vast majority of tests are designed by cross-functional teams with skills from UX to customer strategy. But 90 per cent still fail,” McNellis said. “How does deal with this? It performs 25,000 tests in any given year.”  

To make this a reality,’s key tenet is anyone can test without management’s permission.

“When you have this freedom, people can test all kinds of things - pricing decisions, how results are presented, details about the digital customer experience,” McNellis explained. “That is very important.”  

What often stops this kind of culture from flourishing is the highest paid person’s opinion, or ‘Hippo’, outweighing test-and-learn and more data-driven decision making.

“In our 2020 marketing analytics survey we asked what the number one reason for not using marketing analytics is in any given decision. The result was that data findings conflicted with their intended course of action,” McNellis said.

The key learning from “If we want a great best-in-class marketing analytics strategy, we need to figure out how we can test more and excise opinion from our data-driven decisions,” he said.

Step 2: Measure to change

As an example of how marketing analytics can be used to change, rather than tinker around the edges of marketing decision making, McNellis pointed to recreational sports vehicle manufacturer, Polaris. The group is in its fourth year of helping the marketing organisation deliver better bets. It’s work that’s seen the marketing deliver seven-figure bottom-line incremental profit year-over-year.

McNellis said the key has been changing the cultural mindset around data from pure reporting, to a question of how to evolve. Instead of ‘can I have a dashboard?’, Polaris’ head of data science has worked to change the request to: ‘How can I solve this business challenge?’.

“At Polaris, the data science team was being asked for dashboard with trends and impressions over time, or a dashboard to show marketing performance. These are useful for descriptive analytics and diagnostics. And that’s important – you do need good dashboards. But while they are necessary, they are not sufficient,” McNellis said.

Instead, Polaris built a marketing mix model combining both predictive and prescriptive analytics in order to work out how to make the business outcome it wants more likely. To do this, it tested 400 variables, from weather to dealer variables, competitive activity and units shipped.

“This meant Polaris was able to understand some correlations and trends and predict future marketing performance,” McNellis said. “It could perform bunch of simulations. For instance, if snow falls early, should we market snow mobiles earlier, or will that drive demand sufficiently and we hold back on marketing? Or if we doubled down on marketing in the spring for our 4WDs, does that mean we can have less rebates in the fall? If we fund a certain brand in the north east less, and one in the south east more, will it drive more overall sales?

“The team tested dozens of these hypotheses and reviewed which were the most important ones to put in the field. It then tested these rigorously, with dealer match testing. This gave Polaris a great way to know when they were winning or losing and to know with confidence it was generating those seven-figure profit improvements.”

Again, not all tests work, and Polaris keeps score on a portfolio basis, rather than via each test. This allows teams to be more innovative and take more risk, McNellis said.

“Polaris shows how analytics can be used as a coach, not scorekeeper, to improve marketing performance over time,” he said. “The lesson here is we have an opportunity to push backs confirmatory measurement and instead into measured predictions that drive growth.”  

Step 3: Predict and prescribe

McNellis’ third must for best-in-class marketing analytics is to orient towards predicting and prescribing over reporting and diagnostics. He used Avon as an example, which saw an opportunity for its sales force to be engaged more consistently with customers and prospects. This opportunity was estimated to be worth $2bn in sales per year.

“Lots of hypotheses are generated as to what can help realise that opportunity. As the executive team started assessing these, it needed a lot of descriptive and diagnostics analytics coming from a variety of data sources to figure out how each correlates to the opportunity,” McNellis explained.

Multiple causes were found to be contributing to this challenge, from online competition to marketing problems, gig economy, changing customer preferences and cultural change. These also varied by product, geography and other factors, making an answer complex.

“The other problem was many data sources were not designed or managed to back up to salesperson engagement. So there were lots of questions on what data means and if it could be trusted,” McNellis continued.  

“What Avon did was a simple pivot. It took this why question – why are salespeople showing less active participation than we think they should – and pivoted to a question perfect for machine learning: Which salespeople are most likely to be inactive?”

This was tightened up to: What is the probability a salesperson will be inactive over the next three sales campaigns? To answer this, Avon tapped historic data plus hundreds of potential predicters for behaviour. It then created salesperson segments based on the predictions to help with scenario planning. Avon also performed rigorous tests to know if the effort is working and if it’s generating sufficient returns.

Segments were scored on predicted level of activity over next three campaigns, cross tabbed historically with historic average order value. “The segments created allowed Avon to not only target salespeople but tailor the content and message to their specific needs,” McNellis said.

Pilot group figures across potential reps resulted in a + 7 per cent uplift in sales, while with lapsed reps, Avon lifted active reps by 11 per cent resulting in a 15 per cent increase in sales.

The lesson for McNellis is to identify reporting requests better served by predictions, then predict. “As great marketing analytics teams, we will do prescriptive and diagnostic reporting, but part of our plan has to be focused on the future,” he concluded. “And part of our analytics has to be focused on the future to make the results we need more likely to happen.”

Follow CMO on Twitter: @CMOAustralia, take part in the CMO conversation on LinkedIn: CMO ANZ, follow our regular updates via CMO Australia's Linkedin company page, or join us on Facebook:



Join the newsletter!


Sign up to gain exclusive access to email subscriptions, event invitations, competitions, giveaways, and much more.

Membership is free, and your security and privacy remain protected. View our privacy policy before signing up.

Error: Please check your email address.
Show Comments

Latest Videos

More Videos

Well, the conversion can be increased by just using marketing, but in general if you are considering an example with Magento, then it is ...


How Remedy is using digital marketing and commerce to drive conversion

Read more

yo nice article


6 Ways to ramp up Social Media to Your Web Design

Read more

Thanks for sharing with us. I just loved your way of presentation. I enjoyed reading this .Thanks for sharing and keep writing. It is goo...


Cancer Council: Finding the physical-virtual engagement balance post-COVID

Read more

yes AI should be a course so many People Use AI

M Abdullah Khan

Is AI on course to take over human creativity? - Modern creative - CMO Australia

Read more

Extremely informative. One should definitely go through the blog in order to know different aspects of the top retail technology.

Pooja Gupta

Donut King takes in-store marketing to the next digital level

Read more

Blog Posts

Brand storytelling lessons from Singapore’s iconic Fullerton hotel

In early 2020, I had the pleasure of staying at the newly opened Fullerton Hotel in Sydney. It was on this trip I first became aware of the Fullerton’s commitment to brand storytelling.

Gabrielle Dolan

Business storytelling leader

You’re doing it wrong: Emotion doesn’t mean emotional

If you’ve been around advertising long enough, you’ve probably seen (or written) a slide which says: “They won’t remember what you say, they’ll remember how you made them feel.” But it’s wrong. Our understanding of how emotion is used in advertising has been ill informed and poorly applied.

Zac Martin

Senior planner, Ogilvy Melbourne

Why does brand execution often kill creativity?

The launch of a new brand, or indeed a rebrand, is a transformation to be greeted with fanfare. So why is it that once the brand has launched, the brand execution phase can also be the moment at which you kill its creativity?

Rich Curtis

CEO, FutureBrand A/NZ

Sign in