In a world with the internet and social media at our fingertips, direct mail marketing often seems like a relic of the past. While digital has become the highest priority channel for many lender’s marketing strategies, traditional direct mail campaigns have consistently proven effective in reaching potential customers. One reason for their continued success: advancements in technology have paved the way for data-driven strategies, which often include grouping potential customers into different segments based on various traits and building models to predict their future performance. Each segment will have their own offer and subsequently exhibit different response behavior; understanding the optimal way to distribute and prioritize mailings is a critical component in maximizing campaign effectiveness.
In this two-part blog series, we will discuss credit card direct mail segmentation strategies and how they are used in part one, along with different mailing optimization solutions and their advantages and disadvantages in part two. We will answer the following two big questions:
- How do lenders determine how many direct mail pieces is optimal to send to customers?
- How do lenders determine to which group to send out more mail than other groups?
Direct Mail Overview
Direct Mail (DM) Marketing is a widely used strategy across many industries, especially in financial services. In 2023 Q2, over 2 billion credit card acquisition mailers were sent to consumers.
The general process flow of a credit card direct mail campaign is as follows:
(1) Identify which product to promote. This could be a new product offering or an existing product with increased focus on acquisition volume.
(2) Identify the target customer segment(s). Different products will attract different customer types with varying behaviors; ensuring the prospect’s needs match your product’s offerings is crucial.
(3) Create an offer strategy. Each segment and subsegment will respond differently to various language and tone used, offers presented, and application channels; altering these attributes maximizes response rates or return expectations, depending on the goal of the campaign.
(4) Identify optimal mailing allocation. Once all segments and mailing strategies have been determined, a lender must decide how to allocate their marketing budget to the different segments.
(5) Implement the campaign strategy and collect responses. This data crafts future campaigns, specifically in the construction of future response models and ways to distinguish correlation and risk.
A “successful” marketing campaign can be defined many ways; high response, high return, etc. But no matter what your organization’s definition of success is, best-in-class campaigns leverage strong analytical capabilities to empower data driven approaches in marketing campaigns. 2OS has consistently seen the best lenders build high value-generating campaigns using their analytics to identify top marketing segments, build differentiated offer strategies, and optimize allocation of their varied creatives.
Segmentation and Offer Strategies
As noted in step #2 above, a key component of any marketing strategy is identifying the customer base. This includes clustering potential customers into segments or groups based on different attributes. Building these segments can be challenging; best-in-class lenders use a combination of varied data sources to divide customers into segments based on their risk (associated with probability of default, “PD”), response (how likely the customer is to respond to direct mail), and behavior (whether they are a transactor, revolver, or mixed).
All of this, among other data, helps differentiate customer expected performance and optimize campaign allocation. These attributes will often come from a variety of different data sources including general demographic data (e.g., age, income, geographic location, etc.) and credit history data (e.g., credit scores, spending patterns, etc.). Lenders can also use different online marketing affiliates (such as CreditKarma or LendingTree) to find credit-seeking prospect leads and data about their desired products.
By combining similar customer profiles, lenders create custom tailored mailing strategies to ensure they are sending only relevant offers to each group, increasing the likelihood of engagement. Examples of different segment groups could include students, travelers, rewards enthusiasts, credit re-builders, heavy spenders, or retirees. Once the segment groups are decided, each is assigned their own offer strategy. For example, a lender may focus on a low-APR and/or low-fee with minimal rewards products for credit re-building prospects, while offering a high-APR and/or high-fee with high rewards products for rewards enthusiasts. If lenders partner with specific merchants, their data can be used to create even more granular behavioral groups, focusing on different hobbies or interests, and offering rewards or discounts from specific targeted merchants.
Mailing strategies incorporate more than just the product-prospect fit, too; each segment can have different language used throughout the letter, different content structure to promote the key benefits, and targeted channels where the prospect can respond and apply. As an example, DM offerings to students are likely to include clear explanations of how the payment process works and benefits of paying on time and building good credit, with a link to the online application via QR code. Offer letters to frequent travelers are likely to be short and to the point, highlighting the high base rewards, spending bonuses, and perks included (e.g., lounge access, partner discounts, etc.) with the APR and fees being mentioned offhandedly at the bottom. Creating the optimal mailing strategy for each segment maximizes both response and future customer performance.
Once the segments are identified and corresponding mailing strategies are crafted, the same risk, response, and behavioral attributes can be used with some performance assumptions to generate segment-level response rate expectations and financial value projections. We see most lenders use a “Net Present Value” or “NPV” metric to calculate the prospective customer’s lifetime expected value. Beyond NPV, other value metrics such as Return on Assets (ROA) and Risk-Adjusted Margin are also commonly used to provide extra insight into the value of a new booking. These response rate and value projections are used to help identify how much mail volume and marketing budget should be allocated to each segment.
Identifying the optimal budget allocation for a campaign can be difficult; lenders must balance the tradeoffs between response, marketing costs, and value estimates to maximize a specific target metric. Common optimization metrics we see used include total number of booked accounts, number of active accounts, and value estimates like ROA and NPV. Sophisticated optimization techniques can even optimize for multiple different target metrics at a time, with primary metrics of focus and tertiary metrics that must meet a specific threshold.
To achieve a successful campaign, additional broader factors and constraints beyond campaign return maximization must be accounted for. These factors include:
- Incorporating partnership agreements (i.e., revenue/loss sharing),
- Respecting maximum loss thresholds,
- Adhering to segment-level spend allocation limits, and
- Assessing impacts on the whole portfolio relative to institution-wide goals.
Given the amount of data to balance and the myriad dimensions and constraints to optimize against, we would not say DM optimization is a “solved” problem. We have seen different optimization approaches yield widely differing results for multiple lenders depending on their needs.
One of the more common solutions we have seen across multiple institutions is to build a rank-ordering algorithm, which allocates mail successively to the highest ranked optimization metric, which is done ideally per customer and per segment. If the lender were optimizing on ROA, the algorithm would identify the highest ROA segment and allocate the maximum pieces of mail until either a budget or custom constraint is met. It would then move to the next highest ROA segment and repeat the process. While this is an effective solution that can be used in multiple scenarios, our experiences have shown these algorithms have difficulty scaling up. When applied to many segments and constraints, rank-ordering algorithms often leave significant swaths of the population excluded from campaigns, simplifying their execution but leaving lenders in the dark about future optimization possibilities.
At 2OS, we found the use of Linear Programming to be an efficient and effective alternative to rank-order decisioning. Linear programming works by mathematically expressing all constraints at once, rather than individually handling one at a time. This reduces the overall range of outcomes to a much smaller “feasible region” in the campaign itself, which allows for the maximization of multiple metrics at once and requires far less computational power, resulting in much quicker run times for a variety of scenarios. These models are also more flexible and are (generally) less error prone given their decreased reliance on manual changes.
Stay tuned for part two of this blog series, where our team will do a deep dive on the mathematical underpinnings of Linear Programming; we will discuss some of the technical advantages it brings to organizations across the world and some of the core methodological principles behind the algorithm.