Crunching the Loyalty Numbers
How to Calculate Loyalty Program ROI
In today’s increasingly competitive market, brand loyalty is emerging as a hot topic for forward-looking brands to differentiate from their competitors. With more and more choices available to consumers and customer acquisition costs rising, businesses are constantly seeking ways to retain customers and foster true long-term relationships.
Loyalty programs promise to be the silver bullet. But as these programs evolve, so do the challenges associated with assessing their true value.
If you’re a brand looking to build a loyalty program, it’s crucial to understand how to measure your program’s effectiveness and return on investment.
How to measure loyalty program effectiveness
First, brands need to set clear objectives for what they want to achieve with their loyalty programs. In most cases, this is measured by incremental revenue – which improves through increased customer retention (lower churn), higher average order value, or a greater frequency of purchases.
We should pause here to highlight that the key is incremental revenue, meaning revenue that is in addition to any revenue that would have otherwise occurred, even if the program did not exist (we will dive deeper on this topic in the next section).
On the flip side is the cost of the loyalty program – this includes variable costs, such as the cost of rewards (the incremental cost associated with giving out more value to customers through rewards), as well as fixed costs (such as software or administrative costs, which are flat and do not increase proportionally to the number of rewards given out).
Together, we compare the incremental revenue to the cost of the loyalty program to arrive at ROI. If ROI is positive, it indicates that for every dollar the brand spends on the loyalty program, the company is earning more than a dollar in return. If it's negative, the company is losing money on the program.
Calculating Incremental Cost (“Cost of Rewards”)
Most of the values on the cost side are straightforward to calculate. Let’s break down the rewards into 4 types and discuss the cost associated with each:
For tangible items, this is the cost to the business of purchasing or producing an item (“COGS”). The most common example of tangibles are free products.
For example, if you’re a cafe and one of your rewards is a free coffee, the cost of the reward would be the cost of purchasing the beans, transporting them to your shop, and turning them into beverages. Effectively, this is what comes out of your gross margin.
For services (like access to an exclusive event), the cost would be the cost of delivering the service – a fixed cost that is divided across the number of customers who earn the reward and, theoretically, do not cost the brand an incremental amount to issue more of.
For example, let’s say a brand is hosting a concert for its top 100 customers, and the entire event costs them $5,000. The cost of rewards would be $5,000 divided by 100 rewards issued, or $50 per reward. If they wanted to open this event up to 200 people rather than 100, assuming the cost of the event stayed the same, the cost of rewards would drop to $25 per reward ($5,000 / 200 rewards issued).
For discounts, the cost of rewards is the amount of revenue forgone by offering the product or service at a reduced price.
For example, let’s say a coffee shop sells lattes for $5, and the customer receives a 20% off discount as a reward. The cost of this reward would be $1 – the customer pays $4 for what otherwise would have been $5 in revenue.
It is worth nothing that this method is true if a customer would have already visited. However, if the trip is an incremental visit, meaning the visit would not have occurred without the loyalty program, then the cost of the discount is no longer $1, but ($1 x COGS margin). Similar to the cost of rewards in “tangibles,” a discount on an incremental purchase would be the same, but only a fraction of that tangible (in our example, 20% of a free latte).
The last category of rewards is intangibles, which typically we assume have zero marginal cost to brands. These could be special perks like “skip the line” or “early access,” as well as cosmetics, such as digital badges on their account or access to special features that don’t cost the brand to issue more of.
For example, maybe the brand issues access to a special group chat as a reward for its most loyal members, or the right to vote on the new coffee flavor of the month. These are rewards which customers find value in, but which cost the brand no incremental dollar value to issue.
Beyond the incremental costs discussed above, there are other fixed costs, such as platform & technology costs for third-party software platforms, or the cost to train staff. These costs typically only occur upfront and do not scale proportionally to the program, so as the program produces more incremental revenue, ROI improves.
Calculating Incremental Revenue
On the incremental revenue side of the equation, the picture is less clear. Let’s start with a discussion of how most brands calculate loyalty program ROI today, what’s wrong with that approach, and then dive deeper into better ways to do it.
How incremental revenue is typically calculated
Historically, loyalty program incremental revenue has been calculated by dividing customers into two groups – loyalty program members and non-loyalty program members. The non-loyalty program members are essentially the control group, and revenue is then measured for both groups over the same time period.
The incremental revenue per customer is then found by taking the average revenue per loyalty member minus the average revenue per non-loyalty member.
That incremental revenue per customer is then multiplied by the total number of loyalty program members to arrive at the total incremental revenue of the program.
This approach is flawed.
Why? Because the types of customers who join a loyalty program are already more likely to be more valuable than the types of customers who would not join the loyalty program (even if the program did not exist).
In statistics, this phenomenon is referred to as a “self-selection bias.” No matter the industry, customers who opt into a loyalty program are likely already brand enthusiasts, frequent shoppers, or those who have a higher propensity to engage with the brand. All of these factors would indicate that they inherently spend more, visit more often, and are generally more engaged than those who don’t join, regardless of any incentives provided by the program.
The result? This bias skews the results in favor of the loyalty program, making it appear more effective than it truly is.
How do we address the self-selection bias?
There are a few ways to account for and address the self-selection bias.
1) Difference-in-Difference (DiD Analysis):
The first (and most obvious) is to change the way we select a control group to compare the program against.
Rather than splitting our groups into loyalty members and non-members, we split the groups into loyalty members after they joined the loyalty program and those same customers before they joined the program.
The idea is to observe trends before and after the introduction of the loyalty program into the brand’s business so that the difference in trends can be more isolated to the existence of the program itself, rather than behavioral differences among different types of customers.
2) Randomized Controlled Trials (RCTs):
Another robust method would be to randomly assign customers to a treatment group (those who get the loyalty program) and a control group (those who don’t get the program).
For example, let’s say a coffee chain has 10 locations. We might take 2 randomly selected locations to initially launch the program and compare performance improvement at that location to the locations who do not have the program.
This random assignment ideally ensures that the inherent characteristics of the customers are equally distributed between the two groups, minimizing bias.
While partially reducing bias, this approach is not perfect, as it does not account for external factors that may be impacting the performance of certain locations (regardless of the existence of the loyalty program). For example, maybe one location is in the proximity of a major community event, which artificially inflates sales, even if it didn’t have a loyalty program.
3) Propensity Score Matching:
A third way to balance the control group and test group (non-loyalty members and loyalty members) is to divide them based on their propensity to join the program. This approach requires more robust analytics capabilities, but it uses observable characteristics to estimate each loyalty customer’s probability of joining the program (before they joined) and then matches them with non-members who have similar propensities. The two groups are then compared over the same period of time.
While this method reduces the self-selection bias, it requires enough data from observable characteristics about customers before the program is launched to accurately assess how likely someone is to join the loyalty program, which for most small brands with limited data can be difficult to predict.
In summary, while a straightforward approach to measuring incremental revenue can provide a starting point, it’s essential to account for the self-selection bias, which recognizes that people who join the loyalty program were already stronger customers to begin with.
In addition to the quantitative approach we described above, it may also be useful to ask customers directly why they joined the loyalty program and how they personally say it impacts their behavior. Answers to these questions can help brands understand underlying motivations behind customers who join.
Using these methods to calculate costs of loyalty and incremental revenue, you can better understand your loyalty program’s overall ROI.
As part of utilizing Hang’s platform for your brand loyalty program, we provide extensive analytics dashboards to constantly monitor and optimize your loyalty programs over time. If you’re interested in learning more, check out our website here.