Segments, segments everywhere
“Segmentation does all these great things, but it doesn’t make coffee,” began Jennifer Gary, marketing manager at Entergy in the “From Data Chaos to Touchpoints, Transactions, Tests, Targets & Tailored Messaging” session on the final day of Utility Analytics Week in Atlanta. “You still have to do your own homework, apply your own thinking.”
Gary’s team at Entergy created a residential customer segmentation program to help them design new products, offers and messages.
“Like many companies, Entergy didn’t have a formal, 360-degree lovely data view of our customers, we had to go find it. That took 5 months,” she said.
Gary called segmentation a journey and with Entergy’s segmentation journey, they began where all analytics programs do: They gathered the data.
Then, they created a marketing database and built a segmentation model. After that, the algorithm was applied to the customer base and each customer got a score. Then “the fun started,” she added. The fun was developing segment strategies to optimize every segment.
That led to developing a learning plan to, well, continue to gather that info and learn from that data beyond the initial segmentation.
“Now you know all of this stuff, but what do you not know? What do you still need to find out?” Gary asked.
Entergy’s segmentation was latent class using a multi-dimensional data set. Their segmentation was less demographic than behavioral with variables like use, call history, media preferences, billing, outages and more.
What they found out: The 80/20 rule applies here at Entergy. 76% of their customers were opportunity to engage. They could apply more services, use more programs, get more involved. 24% of their people fell into a “customer service challenge” bucket.
The need-to-engage segment fell out into these groups: secure traditionalists, up and coming families, thrifty renters, old-fashioned service seekers. And they learned they needed to solve issues for these groups: the over-extended, the financially desperate, and the stressed and struggling.
Gary emphasized that the success of the segmentation, for them, involved a test plan---a discussion about what you learned today and what you want to learn six months from now. One example: a learning objective around what strategies are most effective in driving enrollment which leads to hypothesis on how to answer that question and then a plan to answer it.
Gary turned the floor over to Kristy Lovett, senior manager of customer satisfaction and customer experience, and Mike Mattingly, manager of data analytics, both with Ameren Missouri.
“We all know that the marketing mandate overall has evolved,” Lovett said. “There are different expectations---and varied expectations---these days.”
Ameren Missouri had a vision to streamline and individualize customer communication. But the problem was none of them knew who was talking to whom about what. The customer might be getting hit by energy efficiency program marketing, general marketing and other customer outreach without the utility understanding those marketing efforts within different streams of their own silo.
“We wanted to talk to the right customer with the right offer in the way they want to be talked to,” she said. “And we wanted to incorporate more analytics to improve message resonance via those targeted communications.”
In other words, like Gary and Entergy, Ameren wanted that 360-degree view.
Mattingly helped Lovett in the first steps to reach that vision, which started with a goal of 20% paperless enrollment. He’s the data guy with the data team.
“And then I threw her a curveball and said, ‘I want to talk about eCustomer instead,’” he said, talking about how they could promote both ideas since most eCustomer segments are also paperless---over 76%. So, in pushing an eCustomer approach, they could essentially get those paperless numbers, plus more.
With this push, they noticed that current eCustomer enrollment occurs within 90 days of becoming a customer. Without knowing that, they may have incentivized behavior that customers would have already taken within the first three months.
Mattingly focused on models vs. strategy and understanding that “segments drive strategy” and “models drive efficiency.” For Ameren Missouri, this thinking led them to want a response model but they didn’t have the data. So, they started with a clone model in absence of a known marketing history.
“While the clone model isn’t optimal, it allows you to build,” Mattingly added. They targeted with that cloned model and then built the dataset needed for an eventual response model.
And, keeping that knowledge on eCustomer data and the focus on paperless, they excluded billing accounts opened 90 days (along with, of course, those already enrolled in eCustomer).
The process in simple jargon: data extraction followed by data redoes, outlier and missing value treatment, training and validation split, a clone model logistic regression, and then a response model. The modeling was successful, with a segment identified and a strategy built.
Both Ameren Missouri and Entergy are in the process of finishing up and building out larger programs combining marketing and customer segmentation.
Being in that process also means both are in Gary’s: “what do you know, what do you want to find out” stage and growing. The customer segmentation story will continue, with Entergy, with Ameren Missouri and likely withing your own utility.
Get more insights from Utility Analytics Week:
- Follow all the excitement of Utility Analytics Week on Twitter
- Dealing with dollars, DSOs and the divide
- Data, customers and climate are changing the industry
- Utility Analytics Week has tasty, sweet Halloween opening session
- 7 tidbits of in-the-trenches advice for your analytics program
No discussions yet. Start a discussion below.