Cohort analysis is becoming a competitive necessity, especially as tracking and targeting individuals becomes more difficult.
With third-party data going the way of the dodo bird, digital marketers are looking for ways to do without cookies. Call it “data dieting.” But something must replace those bits and bytes of third-party sweetness. If you can’t drop a cookie, track a cohort.
Any group of customers engaging your web site can be counted as a cohort, provided you are tracking what they do. Are they just going as far as the landing page? Are they filling the cart, but not checking out? Did they buy something before, but haven’t shopped lately?
Churn, drop-off, customer lifetime value — all can be tracked as cohorts. But the online vendor must know what measures are most relevant to their business to make the most of cohort analysis.
(Segment [cohort]): Get it?
“Segments” and “cohorts” are terms sometimes used interchangeably, but that would be incorrect.
Google defined cohort this way: A cohort is a group of users who share a common characteristic. “For example, all users with the same Acquisition Date belong to the same cohort. The Cohort Analysis report lets you isolate and analyze cohort behavior.”
In contrast, segmentation means organizing groups of users around common characteristics, like demographics, geography, personality, or value. It can also group customers using more than one characteristic.
“A cohort is a form of a segment. All cohorts are segments, but not all segments are cohorts,” said Eric Sloan, director of strategy at performance marketing agency Thrive Digital. Cohorts can be understood as “time-based segments”, for example, a set of users signing on at a web site at a particular time.
Sometimes the two terms get mixed up because of the analytical tool being used by the vendor or analyst, noted Adam Greco, product evangelist for digital optimization platform Amplitude. A cohort is “a group of like users based on interest,” he said. Segmentation “is like a filter,” Greco continued. A segment is an activity. Cohorts are people. And a “cohort depends on identity resolution”, he said.
Simply identifying a cohort is not enough. The analyst will have to drill down further to identify cause-and-effect. “It’s the only way to make cohort analysis meaningful,” Sloan said. The biggest pitfall is just assuming “the time-based cohort caused what you are looking at,” he said.
Asking the right questions
Which leads to the data. There is an answer in there somewhere, provided you ask the right question to get it.
“We spend time using data to figure out the cohort that is meaningful for the business,” Greco said. Take the example of a cohort defined by behavior — customers going through a multi-step process to complete a transaction.
“You need the right data to build the right cohort,” Greco said. You don’t need to worry about tracking people who add items to the cart. “Just because you can track something does not mean you should.” He added. “Too few companies start with the end in mind.” If you start by listing the relevant cohorts you want to track, and work backwards, then you are more likely to be successful, he pointed out.
For Sloan, the cohort is part and parcel of root-cause analysis. “[When] you see KPIs change, you look at all the different factors that caused the change.” Again, correlation is not causation, he stressed, but you keep drilling down through the cycles and ask intuitive or logical questions, finding the data that answers the question.
“Start with a cohort. See if it is time-based.” Sloan said. Spot the drop-off from period to period. Include new visitors as older ones drop off. Look at the face value of all behaviors and events, going through the initial period, then to 30, 60 and 90-day increments. “A cohort is the first step in eliminating some of the noise,” he said, as the analyst tries to measure the customer experience with the web site.
Greco offered other paths. One approach uses the data to isolate groups of identified users so that groups can be compared. He called this a “persistent cohort.” For example, tracking the number of online shoppers for a seven-day period. New users will naturally enter this cohort while others exit it after the time set. Those who purchase are counted while those who don’t buy are tallied as drop-offs.
Then Greco outlined the “predictive cohort”. One example is looking at the number of shoppers who visit the web site to make another purchase. There may be a group that is 90% likely to buy something next week; another that is 80-90% likely to make a purchase, yet another group that is 70-80% as likely to acquire an item. The marketer can use that data to offer discounts to each cohort, only increasing the discount to entice shoppers in the next group less likely to buy something. “You use the cohort in combination with marketing and promotion to get people to convert,” he explained.
Making the most of data
Cohort analysis is an approach that requires marketers to change their thinking to make the most of their data. Our experts have the same starting point, but pursue their goals along different lines.
To use cohort analysis, “start with the question.” Sloan said. “Tie back to business results that are possible to answer…Understand where and how to drill down…Make sure the KPIs are meaningful.” Make sure the data you are analyzing reflects reality, he added. Data can skew if the same online shopper is accessing the same web site using different devices at different times, he cautioned.
Greco framed cohort analysis as a competitive necessity. In the e-commerce realm, every online shopper is just a “click or a swipe away”, he noted. The burden is on the marketer “to figure out how they are losing people and bring them back.” The faster problems are solved and fixed, the more likely an online web site will be successful.
The post How to make the most of cohort analysis appeared first on MarTech.
MarTechAbout The Author
William Terdoslavich is a freelance writer with a long background covering information technology. Prior to writing for Martech, he also covered digital marketing for DMN. A seasoned generalist, William covered employment in the IT industry for Insights.Dice.com, big data for Information Week, and software-as-a-service for SaaSintheEnterprise.com. He also worked as a features editor for Mobile Computing and Communication, as well as feature section editor for CRN, where he had to deal with 20 to 30 different tech topics over the course of an editorial year. Ironically, it is the human factor that draws William into writing about technology. No matter how much people try to organize and control information, it never quite works out the way they want to.
(22)
