A recent report from Kantar found that more than 90% of marketers don’t know what they don’t know, creating problems with consistency and marketers’ ability to act. How can they blend art and science to fill in the gaps?
When John Keenan worked in database marketing at Leo Burnett in the early 1990s, data collection took great effort. Keenan and his colleagues would send coupons to consumers through snail mail, waiting for weeks to get a vague sense of how a campaign was performing. Response from mailers lacked precision; Keenan and his team wouldn’t know where or when the coupons were redeemed. Their hard work yielded a vague understanding of why the coupons were used and they could only approximate whether the campaign had paid off.
Now, Keenan works as executive vice president of marketing analytics at creative agency Periscope and can track consumer data down to a precise moment. Granular data—where a customer bought a product, at what time and even what the weather was that day—is available and easily accessible within a closed loop to any marketer with the right technology. In closed-loop reporting, an incredible amount of data is available to both marketing and sales teams. Rather than the vague approximations Keenan had to surmise from mailers, he and many other marketers now receive consumer data from apps, websites, third-party data vendors and CRM platforms—and new methods of data collection keep coming. With more consumers online and more ways to collect information than ever, data has exploded: A 2017 report from IBM found that 90% of data in the world today was created in the previous two years, a number that has surely risen since.
Although data collection has become easier, marketers still struggle to make sense of their data. Kantar’s 2019 “Getting Media Right” report finds that fewer than 10% of marketers say that they have all the data they need—the rest can only collect a fraction, or they notice gaps in the data they have.
“To me, it’s really kind of astounding, especially [with] North America being such a mature market,” says Aaron Peterson, senior director of marketing of Kantar’s insights division and author of the report. “We hear from [clients] that they have so much data at their fingertips, but it’s not the right set of data that they need to actually make a decision. They just have all these data points and they’re still kind of struggling with what to do with all that data that’s in front of them.”
The Kantar report found that marketers have a hard time translating their data into actionable insights. Kantar surveyed advertisers, agencies and media companies on their confidence in being able to integrate multiple data sources to discover insights—47% of advertisers said they were “not very/not at all confident,” compared to 26% of agencies and 31% of media companies.
And with inconsistent data and insights comes inconsistent measurement. In the report, 76% agreed that it was difficult to assess how well the brand was performing across multiple channels. While the majority of marketers responded that they track reach and frequency, Kantar found that other methods of measurement vary greatly across the industry, disrupting the performance of data. Kantar’s report says that measuring brand effectiveness can harmonize cross-channel currency, but it’s only being used by half of all marketers. And marketers are inconsistent in their cross-channel measurement of sales and ROI—while more than half of advertisers and agencies track these measurements, only 35% of media companies do so.
Inconsistent measurement has likely led to a foggy understanding of ROI. A third of advertisers and agencies and 22% of media companies say that they measure ROI once a year or less, while about a third of all groups say that they measure ROI continuously. In the age of instantly available data, why do as many marketers measure yearly as those who measure moment by moment?
The promise of marketing data has always been about understanding business, taking better actions and getting better results. It’s great to know how people shop, but marketers are in the business of selling, not measuring. And thus far, the data revolution hasn’t allowed for action nor consistent results. Kantar reports that more than half of the advertisers and media companies they surveyed believed they were not successful at acting on data in real time.
The heart of the problem lies in the fact that such a small percentage of marketers have all the data they require. How can the more than 90% of marketers who are struggling with data find what they need?
What’s Your Business Objective?
The same mistake keeps popping up, Keenan says: A company will dive into data that already exists without first figuring out its business objectives.
Recently, a potential client of Keenan’s expressed interest in digging into some data for something they wanted to measure. Keenan sought to understand the purpose of this exercise and what need it filled for client or internal development, but the potential client didn’t have an answer yet. This is how many data inquiries start, Keenan says, which is backwards.
Keenan says that marketers should start with a requirements analysis to define the expectations of the search for data and then use the analysis to find the right data. Even 25 years ago with mailers, Keenan says that his teams would always start by finding the objective and asking, “Why?”
And it’s not enough to simply know the questions and start analyzing the data—there’s an art to data science, Keenan says. Marketers must consider the exact business contexts that matter, those that will truly affect their objectives. The art must precede the science. “If we’re thinking about something with our analysts, they shouldn’t go back to their desks and start coding right away,” he says. “[If they do], they haven’t been thinking about what they should be doing first.”
The process of data analysis and insight discovery is often overwhelming for overly eager clients, but Keenan says it can be made easier with patience—stating objectives, asking questions and planning thoroughly. Although diving in to available data right away may seem more appealing, Keenan says that it usually leads to greater challenges—and it’s likely a big reason why many marketers find holes in their data.
“In order [to] not be overwhelmed, you have to know how [the data] plays into what you’re trying to achieve,” he says. “You have to start with context. Then, everything else becomes easier.”
Whittling Down the Metrics
Nichole Urigashvili, senior data scientist of research science at food-ordering service Grubhub, determined the business objectives and asked all the requisite questions, but she was left with about 150 different metrics in her data model. “I talked to my manager the next day and they were like, ‘You know what? We just need to take some out,’” she says.
Nailing down business objectives and the contexts in which they exist are essential steps in gathering data, but narrowing the metrics to be followed may be just as important. Both steps consolidate an overwhelming mountain of data into a climbable hill.
If you’re following too many different metrics, Urigashvili says that you can work logically to cut them down. For example, if two metrics are similar, remove one. Or thin the herd by analyzing the data further and focusing only on the metrics that most affect business objectives.
It’s important to remember that even when you know the direction you’re taking, the amount of consumer data can be cumbersome. Often, data can become too granular. For example, it wouldn’t likely be fruitful for Grubhub to attempt to change the actions of 10 people who live on a single block, but the company could find benefits in analyzing consumer trends at a neighborhood level.
When data gets too granular, analysis becomes too much about reaching a few people rather than large swaths of consumers. Keenan recalls a job where he mined grocery store data and had 5,000 shopping indicators on individual shoppers before he could even attach the data to their digital footprint. He had information as specific as whether shoppers bought paste or gel toothpaste, carbonated or non-carbonated water, Coca-Cola or Pepsi. Analyzing data this granular often casts too small a net. “There’s a lot of data out there, so it’s about deciding what’s relevant and actionable in service of the objective,” Keenan says.
Urigashvili says narrowing your data is sometimes about digging deeper and asking more precise questions. For example, if Grubhub is researching churn rate, they’d first need to ask if there’s something they can do about the lapse in support. Who’s leaving? Is it drivers, restaurants or customers? From there, they gather potential metrics to follow and analyze to determine which metrics have the closest relationship to business results. If there’s little or no relationship, that metric can be put aside.
The metrics that marketers follow must serve the business objective—just because something is measurable doesn’t mean that it matters to the organization.
Translate Data for the Innumerate
During a day-long meeting at 4C Insights, a data and marketing technology platform, CMO Aaron Goldman asked Ashley Tomzik, his team’s new marketing data scientist, to help ground the session in data. Goldman had just hired Tomzik away from 4C’s engineering team a few months prior and was blown away by her work in that short time.
“She pulled up a few dashboards,” Goldman says. “There were all the charts we were used to looking at. But then at the top, she had written actual words—imagine that—telling the story or the key insight of what we should take away from each of those [charts]. It was the first time in as long as I can remember at 4C where we didn’t just immediately dive into dissecting a chart. We actually started with some context and a story, a key insight that was drawn out.”
Tomzik says that she knows not everyone easily understands numbers—while it’s her job as a data scientist to make them palatable, she acknowledges that they can make people confused, sleepy, frustrated or bored. When people looked at the charts she presented, she wanted them to see numbers in the proper business context and understand what they meant.
“The numbers aren’t going to be obvious at first,” she says. “You need to understand the whole picture to be able to understand why something might be happening and how things are relating to each other to draw insights from your data.”
There’s a demand for data scientists like Tomzik. LinkedIn co-founder Allen Blue told attendees of a Wharton School of the University of Pennsylvania Knowledge@Wharton town hall that demand for data scientists has seen 15- to 20-times growth over the past three years. But Tomzik says that data scientists in marketing must be able to translate the numbers into words or stories for the rest of the company—if the data scientist understands the numbers but the executive doesn’t, the marketing team won’t likely benefit.
Kantar’s Peterson says that a problem for many marketing departments is that their numbers are crunched by those outside of marketing. “For a lot of companies, there’s somebody within their IT department that’s actually collecting and organizing all of the data and then pushing it back to the marketing team,” he says. “A lot of businesses … are still creating a bit of a silo situation. Marketers need to be able to talk more directly with those folks that are managing the data to really get them to understand what is going to be useful for them versus just pushing the data out.”
It’s important to have someone on the team who understands the business contexts. Goldman says that Tomzik’s ability to translate the data is much better than crunching the numbers themselves or having someone else do it. With Tomzik’s help, Goldman says that the marketing team has narrowed the number of metrics it follows, become more confident in its ability to track data that matters and gotten quicker to act—this year, 4C started working toward unique quarterly goals rather than annual benchmarks.
Follow the Numbers
As data analysts, Keenan says that it’s important to know what people are looking for to understand how data can best support their goals. An organization can be successful with differing views of the same problem, but marketing data analysts must understand what these contrasting views mean.
Urigashvili says that data could be more consistent if everyone better understood what’s being measured. For example, if Grubhub was analyzing the cancellation rate for orders, some in the company may view it purely as a user issue, but that disregards the potential that a cancellation may have come from the business or issues with the delivery driver. Getting more specific on the matter of cancellations can get everyone on the same page, a good step toward solving the problem.
As Tomzik says, it takes good communication to clearly explain these analytic differences to others in the company. But data analysts must serve the business by delivering actionable insights.
Finding Insights and Taking Action
If data is a rainbow, insights are the mythical pot of gold at the end—which most marketers fail to find. Keenan says that the process of finding actionable insights should originate with your objective and the context of the data you search for, as insights are the interpretation of the data you’ve found. These insights will be used to determine what steps the business takes next.
For example, if Keenan’s team was performing a campaign analysis, he says that they’d need to understand the objective of the campaign and how they reached out to consumers to find insights. The team’s analysts would be armed with the contexts of the campaign, such as the kind of offer and style of communication. Then, instead of reporting back the response rate within certain subjections, the analysts would know the context of the interaction and interpret it themselves. Finding insights is concluding why consumers responded the way they did and what that means going forward. Was the response rate lower or higher than usual? Why? How should this affect the next campaign?
Once a business starts gathering insights, acting on them and seeing success, Grubhub’s Urigashvili says that they’ll have an easier time defining and solving future problems, gaining even better observations in the process.
How to Start
If Urigashvili was starting a data team from scratch, she says that she’d first look to understand the business at a high level. For example, how could data potentially be used to help the business?
By knowing how data can be leveraged to reach business goals, Urigashvili says that marketers should be able to score some quick wins, which can prove that being data-driven works. These quick wins will be essential for the most important part of becoming driven by data: getting the entire organization, especially leadership, on board. Hiring smart, creative people who ask the right questions and know the business objectives is great, but change requires a senior leadership team that’s willing to listen and see that data can have a positive effect on the company.
“Companies that aren’t data-driven, especially if their leadership isn’t data-driven, make it hard for a data team to start pushing and advocating for analytics,” she says. “It’s really going to be this organizational culture change. We can use data; this is how we can use it. If we have this type of problem, don’t just go with your gut, let’s utilize data to be able to help with that.”