Why We Don't Know What Marketing Analytics Means

Nothing keeps you up at night like a number.

My career in marketing started in the middle of an industry-wide transformation. Managing “the brand” didn’t cut it anymore. CEOs expected quantification, metrics, numerical proof marketing dollars were moving the needle. Clicks, conversions, cash coming in the door was what mattered.

I didn’t realize any of this at the time, of course. Inexperienced, I assumed this was always how marketing worked. But I also quickly learned the anxiety that came with accountability for a metric while being short on answers for why it languished or thrived.

Around this time marketers started adding “data-driven” to their LinkedIn profiles. Nearly every webinar, conference session, and meeting used the term “marketing analytics” so regularly I assumed it carried the intuitiveness of phrases like “free beer.”

But as I’ve advanced in my career, age and number of gray hairs, I’ve become more convinced that we don’t actually know what marketing analytics means. Sure, we can envision charts and graphs, reports and dashboards — but as to the actual purpose of marketing analytics we can’t easily define.

I’ve held this suspicion for a while now. So I started to investigate. And, helpfully, there’s actually data to explore.

A Quick Summary of the Problem

It’s a modern meme to laugh at clips of Katie Couric and Bryant Gumbel wondering aloud what the Internet is. Funny in hindsight, sure. But the speed in which aspects of life, big and small, moved online is difficult to find a historic precedent for.

One estimate finds the amount of data created, captured and consumed is expected to grow from 6.5 zettabytes in 2012 to 120 zettabytes in 2023 — around an 18X increase. For comparison the number of books printed in Western Europe is estimated to have grown tenfold over a century after the advent of the printing press.?

Much of this output is driven by the exponential digitization of consumer behavior. Internet traffic is expected to grow from 40.8 exabytes in 2012 to 331.6 exabytes in 2022. And how we purchase services and products, as well as consume information, is increasingly digitized.

Ecommerce sales in the US in the third quarter of 2012 were just below $59 billion. In the third quarter of 2021 that figure had risen to nearly $215 billion. Meanwhile, the number of households with at least one video streaming subscription is 82%, and 53% are subscribed to three more streaming services. When Napster ushered in a wild west era of online music, it peaked at 80 million users. In 2020, the music streaming industry added 100 million subscribers. Even food is leaving a vast digital footprint. Food delivery services grew by 7X between 2018 and 2021, and the percentage of consumers purchasing groceries online grew by 41% during the pandemic.

And the pandemic is adding fuel to the fire. McKinsey found that 75% of consumers tried a new shopping behavior since COVID-19 wreaked havoc on daily life, with 36% of consumers trying a new brand and 20% using a new digital shopping behavior.

In other words, nearly every aspect of life can — and often is — conducted digitally, tracked, captured, and stored. Saying your business is in the middle of a digital transformation initiative at this point is a little like a monk saying they have a plan to shift from hand-copied manuscripts to printed books in 1800.?

The impacts of this shift are felt by every business function, of course. But for marketing, the implications are sweeping.

For one, even from its etymological origins, marketing has always been about “transacting business in a market.” The market is now digital, so much so that the term “digital marketing” is oxymoronic. Marketers have rushed to adopt technology to capture, analyze, and act on the data these transactions and behaviors generate.

Consider the epic rise of marketing technology (referred to as martech from here on out). In 2012, Scott Brinker cataloged a total 350 vendors in the space, but by 2020 that number had reached 8,000 vendors. This spans a swath of categories, analytics being just one of those.

Still, marketing analytics is cited as a regular, even predominant, area of investment. Since 2008 Duke University has conducted a survey of marketing leaders, capturing data around economic confidence, technology use, spending, and more. When you look as far back as 2012 (the earliest the survey asked about marketing analytics spend), the percentage of marketing budgets dedicated to marketing analytics has been relatively stable at around 6%.

What has fluctuated more is the expected spend. In 2012 the expected percentage of marketing budget allocated to analytics for the following three years was around 9%. That surged to nearly 22% in 2017 before sliding back down to 9.45% in 2020. But the expected spend always outpaced the actual spend each year. Marketers seem to put greater emphasis on the expected value of analytics than the delivered value justifies.

A technology is well understood when its purpose is applied often within the business. The same Duke survey asked respondents what percentage of projects use or request marketing analytics before making a decision. While there’s been some improvement between 2012 and 2019, it has never achieved even 50%.

But it’s worse than that. The same survey tracked whether marketers formally evaluated the quality of their marketing analytics. If there’s a silver lining in the data it’s that in 2020, for the first time, a slim majority of marketers said “yes” — 53%. (This question was not included in the 2014, 2016, 2018 or 2019 surveys.)

Despite the slight momentum here, 47% of marketers not evaluating their analytics and less than 38% using analytics when making decisions suggests there is confusion in the category. A product or practice that doesn’t inform decision-making, and with no formal evaluation, means that the very term “marketing analytics” is, at best, vague or misunderstood.

How Did We Get Here?

In a recent article, we laid out a basic history of marketing analytics. In that timeline the explosion of marketing data corresponded with the rise of digital advertising and CRM. Eventually, this evolved into automation of digital content, messaging, and advertising based on the behaviors of individual customers. Gradually, demand increased for solutions to better visualize the trends within the data, birthing a generation of dashboards.

If we map that ecosystem to the goals it helps marketers achieve, it would look like the following: The systems we use enable us to track customer behavior, then collect and sort the data (either centrally or disparately), use the data to automate the delivery of messaging, and dashboards to visualize the data and explore it in its raw form.

But are those the right goals? I mean, when we use the term “marketing analytics,” what are we actually trying to do?

Let’s consult some definitions for marketing analytics. Here are three of the top definitions, according to Google Search.

What is Marketing Analytics?

“Marketing analytics is the practice of managing and studying metrics data in order to determine the ROI of marketing efforts like calls-to-action (CTAs), blog posts, channel performance, and thought leadership pieces, and to identify opportunities for improvement.” - Source: Marketo

“Marketing analytics comprises the processes and technologies that enable marketers to evaluate the success of their marketing initiatives. This is accomplished by measuring performance (e.g., blogging versus social media versus channel communications). Marketing analytics uses important business metrics, such as ROI, marketing attribution and overall marketing effectiveness.” - Source: SAS

“Marketing analytics is the practice of measuring, managing and analyzing marketing performance to maximize its effectiveness and optimize return on investment (ROI). Understanding marketing analytics allows marketers to be more efficient at their jobs and minimize wasted web marketing dollars.” - Source: Wordstream

I offer these definitions as a useful, if ambiguous, way to explain why marketers are seeing so little value from their analytics.

Each of these definitions emphasize that the purpose of analytics is to measure the success or effectiveness of marketing activities. This may be at an individual tactical level or across programs and campaigns. But the expected outcome is that you are able to pin a metric value to marketing’s efforts.

On the other side of this coin, these definitions emphasize “efficiency” and minimized waste. The purpose of marketing analytics is to assign a value to what marketing does and — perhaps simultaneously — to reduce the cost of that effort.?

Both are fine goals. And if these are the goals, the visualization abilities of BI or the embedded reporting features of tools in a martech stack may suffice. (The keyword here being “may.”)

But I would argue that these definitions are out of step with the demands on marketing today.?

The purpose of marketing is not to defend marketing. Yet, marketing analytics, as loosely defined today, makes that a primary goal. Reporting, attribution, dashboards — too often these are associated with vague phrases like “proving the value” of marketing.

But marketing’s role is not shrinking into practical indefensibility. It’s growing in influence, responsibility and visibility. Consider the sheer number of responsibilities senior marketers claim. In 2017 and 2021, 50% of marketers cited 12 different functions that marketing was primarily responsible for, ranging from analytics to PR. That’s a big remit.

It’s not just an expansion of duties, the visibility has shifted as well. In 2021, senior marketing leaders were asked how often they are asked by a CEO or CFO to participate in board meetings. Selecting from a 1-to-7 scale, the largest share (nearly 43%) said “all the time.” Only 13.4% selected “never.”

What does this have to do with how we define marketing analytics? The point is that the focus of analytics needs to move from justifying marketing’s existence to empowering marketing as a driver of growth. Board meetings are not an inquisition. Marketing is not offered a seat at the table to provide a dissertation on its raison d'être. The expectation is to help illuminate the why — why is customer behavior shifting, why are trends fluctuating, why is this market opportunity potentially more lucrative, etc.

Remember that chart I just shared on the number of primary responsibilities marketing teams hold? Well, if you look at the changes between 2018 and 2021, it provides further evidence that marketing analytics (and marketers) need to make this shift.

The largest shift in reported responsibilities was “digital marketing” — an 11.5% increase between 2018 and 2021. In other words, as commerce rapidly digitizes, marketing is expected to hold and grow the bag.

As the spotlight intensifies, marketers need analytics to reveal more than what is happening, they need to be able to elucidate why it's happening so they can understand how to respond. Marketing is not getting this out of their data and analytics today. They struggle to even explain the why internally, much less act on it.? A survey of CMOs bears this out.

The biggest problem CMOs and VPs of marketing face when presenting to the executive team is simplifying all the operational data they’re fed into meaningful insights. In other words, the why. Not just conversions are down or sales are up, but why…and how is marketing going to drive the change rather than be driven by it.

All of this data indicates that, as a function, marketing doesn’t really know what it means by marketing analytics. We need a new definition that meets the wider-ranging goals of the modern marketing team.

A New Definition for Marketing Analytics

In order for a definition of marketing analytics to be useful, it needs to describe the minimum requirements for what the software should do. There are three basic requirements to meet the needs of modern marketers:

1. Marketing analytics should provide an analysis for changes in specific KPI performance.
2. These changes should not be constrained by arbitrary time periods.
3. The software should be owned by the marketing function.

Let’s do a quick review of what each of these mean.

Automate Actual Analysis

You’d be forgiven for assuming that a solution described as “marketing analytics” would provide marketers actual analysis. In reality, most provide visualization of changes that occur in KPI performance, but not explanations for why the change occurred.

This is largely due to the fact that the data analysis process is mostly manual. Changes in performance can be flagged, but if I want to know why the change occurred I need to form a hypothesis, look at raw attributes within the data, cross-reference these attributes in search of patterns that substantiate or disprove the hypothesis.

Let’s say I’m an eCommerce brand noticing that web conversions have dropped. I can see the trend line and when the trend began. But if I want to know why using traditional analytics platforms that requires combing and cross-referencing the attributes in the raw data. For instance, maybe I want to look at time of day, referring source, landing page, and time on site, to see if a decrease or increase in any of these — or some combination of these — are correlated with the overall drop in conversion.

This process is manual, sophisticated, and assumes I have stumbled on the right theory about why the change in KPI performance occurred. And this would be a process that would need to be repeated for each change in each KPI. When you consider the number of KPIs that marketers are tracking, and the frequency in which they track them, getting to why a KPI is over or underperforming is an impossible task to scale.

But advances in AI, particularly unsupervised learning, mean that marketing analytics platforms can provide automated analysis of the data. This is automated analytics. The changes in the KPI performance is still tracked within the software, but instead of deep-diving into the raw data to draw conclusions, the user is provided analysis in the form of multi-attribute patterns that describe why the change occurred, and how much each pattern contributed to that change.

Using the same example above, in an automated version of marketing analytics, I would see when my web conversions dipped. But from there, I would be provided the key drivers for this change in the form of patterns like “Customers who landed on the Puppy Food Sale page that entered from Google search and spent 1:25 - 2:16 minutes on site converted at 31% less than the overall baseline conversion rate.” The key differences here: the user is immediately exploring the analysis the AI generated from the raw data, the specificity of the insight, and that the analysis is not arbitrarily contained to the attributes I thought to explore.

Today, marketing leaders report getting very little value from AI for marketing analytics. But this is largely due to AI primarily being used at the end of the analytical process in order to provide predictive models about the future. But automated analytics actually make it possible to use AI to provide explanations for why changes in KPI performance occur. And in marketing explanations for why a change occurred is more valuable than predictions about what may happen.

Non-Arbitrary Time Analysis

Because of the limitations outlined above, marketers are forced to use arbitrary time periods in which to measure changes to KPI performance. We measure in hours, days, weeks, months and quarters. Within that are near endless permutations. But cognitively you can’t track them all, so you pick a time in which to measure change.

This is a natural inclination given the constraints of exploring raw data. But this is not how change occurs.

Changes to KPI performance doesn’t only occur when we choose to look at it, and it’s not constrained to neat columns. For instance, conversions could increase or decrease over a short period of time (i.e. a day) or over a relatively long period of time (i.e. a month). And these changes are the result of spikes or dips occurring within the underlying data, which might take place at different points in time and over varied durations.

The only way to track and respond to change is to see it in the non-arbitrary periods in which it naturally occurs. And this can’t apply only to the top line metric, such as conversions, but also to the attributes within the data, such as time on site or referring source, that impact the topline change.

Consider the example we used above. The eCommerce brand sees a dip in conversions. Ideally, this would be seen over the time in which it took place. The significant drop occurred over one day, but a longer, more subtle decline took place over several weeks. When we examine each on their own, an automated analysis provided by AI would reveal that the short drop could be the product of sudden declines in time on site and traffic to a specific landing page.

The longer drop could be the result of multiple declines or rises in other attributes occurring within the time period. For instance, time on site declined but traffic increased driven by a newly ranking landing page that is sub-optimally converting, contributing to an overall decline in conversions.

Shifts of both large and small magnitude are what marketers need to be able to track and quickly diagnose. But it can only be done when the analysis of the data is unfettered by arbitrarily defined time periods, which is provided by AI automated analysis.

Marketing Must Own Marketing Analytics

This seems obvious. But it isn’t.

A recent data from the Harvard Business Review found that 71% of business teams relied on data analysts and data teams for insights. Given the pace of change in consumer behavior, data sources, trends and fads, along with the rising responsibilities and visibility of the marketing organization, dependency on other teams for access to the insights buried in data is untenable.

If the role of marketing is expanding, then the function can’t wait on insights to inform decisions. Certainly, a data and analytics function can provide value, but marketing needs to be able to get to analysis quickly. And that analysis needs to be in a decipherable format. If it’s not understood by the primary user, it’s not useful.

Marketing’s expanding responsibilities require that decision-making needs to be informed by a thorough analysis of the available data — and often that data will emerge from the very channels and software marketing uses. For larger strategic projects, partnership with a centralized data and analytics department can be fruitful. But daily decision-making requires a faster approach to analysis that marketing needs to own. Marketing owns the number, the execution, and the decision — it needs to own the analytics as well.

A Definition of Marketing Analytics

Taking these three concepts together, here’s a proposed definition for marketing analytics:

Marketing analytics refers to software run by marketing teams that analyzes big data to identify changes in KPI performance over time, and surfaces the contributing factors to that change, in order to improve the decisions marketers make.

Putting the Definition Into Practice

If you’ve followed me this far (congrats!), then you’ll rightfully be asking, “Great, we have a definition. What do we do now?”

The short answer is to evaluate whether what we have in marketing lives up to the definition. This requires answers to some pretty basic questions:

• Is the software run by marketing or another team?
• Is that software providing us an analysis of the underlying data (i.e. why a KPI performed the way it did) or just data visualization?
• Are we regularly consulting the analysis before making decisions?
• Is the change in performance tracked over preset time periods or based on the time periods in which the change naturally occurred?
• Does the software identify how much individual factors contributed to the change in performance?
• How many clicks does all of the above take?

This Kind of Marketing Analytics is Possible

This isn’t a pipe dream. Given the advances of AI, all of what I’ve described here is possible.

The challenge is changing our perception that AI should automate decision making, rather than automating the analysis to improve our decision making. And that any technology applying analytics, AI or big data is better suited for a team outside of marketing.

If marketing teams are to meet the digital market, and to fulfill the duties of its expanded remit, they can no longer accept these limitations. When we say we’re investing in marketing analytics, we need to know exactly what we mean.?

And insist it lives up to what we say.