We Need Better Marketing
Banks today are looking to drive significant improvement in their marketing efforts by using real-time analytics. And among the latest efforts, there’s a lot of talk – and promise – about Next Best Offer. No one really agrees on what Next Best Offer means.
It wouldn’t be the first time we all use terms we don’t really know what they mean (uh…big data, maybe?)
But I want to clarify what Next Best Offer should mean. Or at least tell you what it shouldn’t mean.
When I say Next Best Offer, I mean the offer that adds the most value to the customer – as measured by an increase in their overall expected customer lifetime value.
Expected customer lifetime value is made up of two parts:
- The profitability if they accept the offer, multiplied by
- The probability they accept the offer.
Most Next Best Offer engines only offer you the second part: the probability. They have models that tell you which kinds of customers should get which product. But they don’t speak to the individual profitability of the customer.
Three Phony Next Best Offer Solutions
I want to talk about three solutions that often get called Next Best Offer that aren’t. If you have one of these, they’re not bad; but they’re not Next Best Offer either. They can be improved.
The Demographic Segment Solution
There are many vendors who have taken marketing data – both demographic and psychographic – and created customer segments in retail banking. They can take a individual customer and map them to these segments. Once they’ve mapped a customer, they have chosen – based on industry-wide data – which product that type of customer should be offered.
Essentially it maps customer segments to products.
This is sub-par. It doesn’t leverage the specific behavior of our own customers and it doesn’t take product profitability into consideration.
What if one of the customers in a segment was being mapped to a product they couldn’t possibly be approved for – or afford?
The Static Rules Solution
The next type of solution that bills itself as Next Best Offer is a decision system where the user creates a rule set. That rule set takes data about each individual and executes a decision tree to find which product to offer. The problem is, the rules were just “made up” by someone (or some committee) based on what they think made sense.
For example, customers with a credit card account, but no auto loan will be offered a car loan. These kinds of static rules may get close to the “right” answer, but they’ll never have the complexity necessary to truly be “best.”
In addition, when things change – as they will – there is no feedback loop in the system that ensures the static rules get updated. Nor, since the rules were just created based off what sounded good at the time, is there any structured way to know when “it’s time” to change them.
These rules, even if they start off right, get stale quickly and no one will even know. That is, until you start missing your numbers.
The Response Modeling Solution
The latest type of solution actually builds statistical models that predict an individual’s response rate to a particular offer. These models do a great job of predicting how an individual customer will react when presented with an offer.
It doesn’t answer the question: “which offer should I present?” But the system basically just suggests you offer the one with the highest predicted response rate.
Again, sub par. Recall that to be a truly Next Best Offer, we need two components: probability and profitability. This solution only offers one: probability.
Will the “Real” Next Best Offer Please Stand Up?
So what makes up a true Next Best Offer solution? Three components are required. If you’re missing any of these three, you don’t have a true Next Best Offer solution.
- Individual Response Modeling
- Individual Profitability Modeling
- Constraint-Based Optimization
Let’s address each individually.
Individual Response Modeling
Response models are built by taking a database of historical offers and building a statistical model that predicts the likelihood that someone will accept an offer based on their individual customer characteristics.
But beyond customer demographics and psychographics – the things that just don’t change about the customer, really good response models are now using transactions and interactions as independent variables in those models. In other words, you might get a cross-sell offer for a new type of credit card after your debit card is compromised, as one example tells.
This requires you to have an offer repository that stores the history of each offer made to each customer – and their response to that offer: accept or reject.
It’s best when that offer repository is cross-channel and that each offer is tagged with the channel through which the offer was made. But often, we don’t have that handy. But if you are interested in executing Next Best Offer and don’t know where to start, this is a very good place.
Create an offer repository.
Individual Profitability Modeling
Here we’re talking about customer lifetime value models. Rather than a statistical prediction model, we’re referring to a formula that tries to capture the lifetime profitability of a customer in a particular product.
Many believe that by leveraging big data techniques and tools, banks will finally have the information they need to understand profitability at the customer level.
This seems like a tall order. Start small. Develop a formula that predicts the profit of a credit card product, or an auto loan. Use those first. Don’t try to get one big formula that takes the entire customer relationship into consideration. There’ll be time for that later. Just having the profitability of the individual products in the analysis will drive most of the value of the solution. So start there.
Create a profitability formula.
This is a technical term for “goal seeking” where you have some rules that need to be followed.
The typical scenario is this: you are running a marketing campaign. You’re looking to maximize profitability, of course, but the campaign needs to sustain itself, so you need to have a minimum number of accepted offers. In addition, one of the products is a credit card product. Risk Management has told us that under no circumstances can the weight-average FICO score of new customers be less than 700. Marketing also insists that at least 60% of all new customers come from the new “premier” segment of customers.
Those goals aren’t all aligned. Maybe higher profits come from lower FICO customers and so by forcing an average FICO of 700 actually takes away some of the gain. The same may be true of the customer segment constraint.
But constraint-based optimization handles all of that. Its goal is to find the maximum profit while obeying all the other rules you give it – assuming of course, you haven’t given it so many rules that no answer is possible. It’s the same technology that’s used in price optimization, a technique which is picking up steam across the globe.
Without constraint-based optimization, you can predict which customer will take which offer, and you may even be able to predict how profitable they will be in that new product, but you’ll never answer the question: “which offer should I present?”
For this, if you don’t have the expertise in-house (and most banks don’t), then you’ll need to hire that out as a consulting engagement or a software license engagement. Whether internal or external, you’ll need some assistance. So start there.
Your To-Do List
So, I’ve pretty much given you your to-do list if you’re interested in rolling out Next Best Offer.
- Create an offer repository (if one doesn’t exist today)
- Create a profitability formula (if one doesn’t exist today)
- Get help!
And of course, if you need help, we’re your resource. Just ask by scheduling a free consultation.
Generally Next Best Offer means cross-selling to existing customers. We’ll be talking a lot about that over the next month in preparation for our webinar “Gaining Share of Wallet: Best Practices for Effective and Profitable Cross-Sell”
We’ll also have a report out soon on the topic. Want to be notified? Then sign up for the webinar and not only will you be given the login information for the event, but you’ll be added to the Bria Strategy Group notification group that receives the latest content and offers.