We have a measurement problem in loyalty. We measure purchases, trips, nights, visits, opens, redemptions and share of wallet, and then we tend to bundle all of that behavior together and call it loyalty. Which is convenient, because all of those things are relatively easy to count. But they don’t necessarily tell us whether a customer is actually loyal.
That distinction came up repeatedly in my recent Customerland conversation with Jaclyn Wands of Phaedon, and I think there’s something important in it. The real test of loyalty probably isn’t whether a customer continues to choose you when choosing you is easy. It’s what happens when it isn’t.
Jaclyn used the phrase “loyalty of choice,” which gets pretty close to the heart of it. If I choose your airline when somebody else has the more convenient flight, or your hotel when another property has better availability, something beyond convenience is at work. And if I come back to you after you’ve disappointed me — a bad stay, a service failure, a screwed-up order — that tells you something considerably more interesting than another successful transaction does.
It suggests that whatever relationship you’ve built has enough stored value to withstand some pressure.
And that may be a much more useful way to think about loyalty.
Maybe we’ve been measuring the wrong thing
A lot of what we describe as loyalty is probably some combination of habit, familiarity and inertia. I already have the app. My account is set up. I know how the product works. The location is convenient. There might be some points sitting around that I don’t want to abandon.
None of that is trivial. Making yourself easy to do business with is an enormously valuable thing. But repeat behavior isn’t necessarily evidence of preference, and preference is where loyalty starts getting interesting.
One of the ideas from my conversation with Jaclyn that has stuck with me is measuring what she calls “return after point of failure.”
Does the customer come back after you’ve screwed something up?
Think about what that captures. A successful transaction tells you that your system worked. A return after a failed transaction tells you something about the customer relationship. There’s trust in there. Forgiveness. Stored goodwill. Some belief on the customer’s part that the experience they just had isn’t representative enough to warrant going somewhere else.
We spend an awful lot of time and money trying to measure emotional loyalty through surveys and proxies. It may be that some of the best evidence of it is already sitting in the behavioral data.
And yes, you can buy loyalty
This is where the conversation took another interesting turn, because there’s a tendency in the loyalty business to romanticize all of this. We like to believe that “real” loyalty has to be emotional.
I’m not sure that’s true.
Amazon is probably the easiest counterexample. Amazon built an extraordinary amount of loyalty through convenience, selection, price and removing friction from buying things. I don’t need to have an emotional relationship with Amazon for that loyalty to have enormous economic value.
But that also raises a more important question: what is actually producing your customer’s loyalty?
If your advantage is convenience, optimize convenience. If it’s recognition, optimize recognition. If it’s expertise, assortment, service, access or price, understand that and protect it.
Instead, much of the industry has converged around a fairly generic prescription for creating loyalty: more personalization, more communications, more offers, more journeys, more engagement.
And now AI gives us the ability to do all of that on a scale we couldn’t have imagined a few years ago.
I’m not convinced that’s entirely good news.
Relevance beats personalization
We’ve spent years treating personalization and relevance as though they’re basically the same thing. They’re not.
Personalization starts with: What do we know about this customer?
Relevance starts with: What would actually be useful to this customer right now?
That’s a meaningful difference, particularly when AI gives brands the ability to act on increasingly granular amounts of customer data.
An airline knowing that I frequently fly between two cities is personalization data. Alerting me when fares drop on a route I actually fly might be relevant. Recognizing that I’m standing in an airport dealing with a cancellation and giving me useful information before I have to go looking for it is definitely relevant.
Same customer. Same data. Very different experiences.
Jaclyn made a point during our conversation that I think is worth hanging onto: don’t start with the brand’s KPI and then work backward through the customer data to figure out how to move it. Start with the customer problem.
That distinction is going to matter more as AI gets better, because AI dramatically lowers the cost of producing customer interactions. And we know what companies tend to do when something gets cheaper to produce: they produce more of it.
More messages. More recommendations. More interventions. More “personalized experiences.”
Which makes me wonder whether the looming customer experience problem isn’t insufficient personalization at all.
It may be industrialized irrelevance.
AI won’t fix the plumbing
There’s another part of the AI conversation that gets considerably less attention because it isn’t nearly as exciting.
AI isn’t going to magically repair bad customer infrastructure. In a lot of cases, it’s simply going to amplify whatever is already there.
If your customer data is fragmented, AI gets fragmented data. If your CDP is incomplete, AI gets an incomplete picture of the customer. If your governance is weak, you’ve just created a much faster and more powerful way to expose that weakness.
Jaclyn describes AI as an amplifier, and I think that’s a useful mental model. Before asking what AI can automate, maybe the first question should be: what exactly are we asking it to amplify?
That brings us back to some fairly unsexy work: AI readiness, martech architecture, data quality, governance and understanding what customer outcome you’re actually trying to produce. There are plenty of things AI can and should take over right now — repetitive operational work, clicking and sorting, pattern recognition and other things humans don’t need to spend their days doing. But there are also good reasons to keep humans at meaningful checkpoints.
AI doesn’t need to be infallible to be enormously useful. We just need to stop designing as though infallibility is right around the corner.
The bigger change may be somewhere else entirely
Toward the end of the conversation, Jaclyn and I got into what I think may be the more consequential question for loyalty leaders.
What happens when the interface between the customer and the brand changes?
For the past couple of decades, we’ve built customer relationships around destinations controlled largely by brands: websites, apps, emails, search results and loyalty portals. But AI assistants increasingly have the potential to insert themselves between the customer and all of those things.
Instead of searching airline sites, checking hotel rates, opening loyalty apps and figuring out which points or benefits apply, I can increasingly imagine simply asking an agent to sort it out for me.
Find the best flight. Book the hotel. Figure out which rewards I have available. Tell me which combination gives me the best value.
That’s a very different customer journey.
And it means brands — and loyalty programs in particular — are going to have to become searchable, interactable and data-ready for machines as well as humans.
There’s a technology challenge in that, obviously. But I think there’s a much more interesting loyalty question buried underneath it.
A surprising amount of what we currently call loyalty is supported by friction. Switching takes effort. Comparing alternatives takes effort. Understanding the value of different rewards takes effort. Figuring out whether another option is actually better takes effort.
AI agents have the potential to make a lot of that effort disappear.
And if they do, we may finally get a much clearer view of which customers were actually loyal and which ones were simply staying because leaving was inconvenient.
That’s what made my conversation with Jaclyn Wands of Phaedon particularly interesting to me. We talked about AI, personalization, customer experience, measurement and the future of loyalty, but they’re increasingly becoming different parts of the same conversation.
Because underneath all of the technology, data, programs and increasingly sophisticated machinery we’ve built around loyalty, there’s still a remarkably simple question:
When your customer has a real choice, why do they choose you?
We might want to get considerably better at answering it.
