In 2019, most marketing decisions were not based on analytics. This is despite the fact that CMOs report that marketing decisions influenced by analytics increased 39 per cent from 2013 to 2019. During that same time period, marketing analytic investments increased 20 per cent.*
Regardless of how analytic output is being used, marketers continue to spend a lot of money on it. Gartner’s latest CMO Spend Survey shows that marketers are spending big on marketing analytics–16 per cent of their budget. That is more than they are spending on activities like content creation, marketing operations or building their brands. This raises a big question: are all those dollars producing enough quality insights? (Gartner clients can access the CMO Spend Survey 2019-2020 here; see figure 9).
To answer this, I often suggest marketing leaders divide their insights into three types. The figure below summarises these three types of insights and the analytic methodologies most likely to produce them.
- Hindsight: Know what has happened. How many impressions did you buy? What was their viewability? How much did you pay for them? Volume and efficiency metrics are common hindsight metrics. Brands often invest significant resources pulling together hindsight metrics from disparate sources. Of course, it is hard to manage what is not comparably measured, so this investment in pulling together data regarding past results often pays off. Data integration also supports and enhances the next two types of insights.
- Insight: Understand patterns in performance. How does click through rate vary by audience type? How does offer performance vary by value tier? Often, insight is produced by cross-tabbing a hindsight metric by another factor. There is a huge list of potential factors. Customer segment, geography, product purchase, usage patterns, response history and surveyed attitudes–how do you find the right subset for your brand? Drivers of results (e.g. conversions) can be a good place to start. One powerful way to identify these drivers is through supervised machine learning, which brings us to our third group.
- Foresight: Predict what is most likely to happen. Which customers are most likely to buy in the near future? Which customers are most influenced by discounts? These types of predictive analytics are becoming increasingly common. Personalisation engines, multichannel marketing hubs, and even a few marketing reporting platforms are coming with these types of predictions built in. What is rarer and arguably a larger opportunity is to use these predictions to help with planning. Prescriptive analytics also generates foresight. Prescriptive analytics can provide powerful what-if simulations, allowing you to consider marketing plans for many possible futures. It can also highlight changes when plans are optimised for different objectives (e.g. incremental sales vs. incremental shoppers).
Are your marketing plans informed by all three types of insights? Do you believe your marketing insights should be more impactful? Do your insights tell where you should invest more? I have two recommendations to improve your insights:
- If you are early in your analytic journey, make sure you are investing sufficiently in diagnostic analytics. You should have a handful of ways to consistently crosstab results that suggest new tests and can guide allocation of resources. You should also have some way to forecast expected outcomes and that forecast should be included for your KPI’s. After all, part of diagnosing what has happened is definitively answering the question, “did what we expect to happen actually occur?” This is a good reporting practice and you can grow in sophistication over time. You may start simple with year over year trending, then grow into sophisticated forecasts, perhaps leveraging open source frameworks like Facebook’s Prophet.
(Gartner clients who want some practical guidance on generating diagnostic analytics can start with Drive More Value From Your Customer Data With Segment Migration Analysis and Use Benchmarks to Take Your Marketing Analytics and Dashboards to the Next Level).
- If your marketing analytics support is more mature, prescriptive analytics is the area of future competitive advantage. Imagine if you made marketing campaign decisions based on the predicted health of your customer base. Imagine if you knew a bad quarter was likely 30 days before the quarter started. Some marketers do this today with their media planning. Essentially, they take the output of a predictive model as the rule set for a prescriptive model. But this type of analytics has more potential and can be used to better manage customer lifetime value or even product development cycles. But for most of us, ingesting probabilistic outcomes (e.g. there is a 62 per cent chance this product’s sales will cover development costs) and weighing multiple scenarios to enhance planning is an area of tremendous opportunity.
(Gartner clients who want to see one-way prescriptive analytics applies to media can checkout Master Marketing Mix Modeling to Maximize Your Media. For a more general approach to the analytics behind foresight, Gartner clients can access Combine Predictive and Prescriptive Techniques to Solve Business Problems.)
If you have used prescriptive analytics to inform marketing decisions, I’d appreciate hearing from you. Whether your first foray of prescriptive analytics failed or you’re an optimisation veteran generating eye-popping returns I hope you share your story. It’s easiest to message me here.
* Analysis based on data found here. Note this study was selected because of it longitudinal nature, but includes many smaller companies then much of the research I use.
- This article is reprinted from the Gartner Blog Network with permission.