Most QSR brands can tell you exactly how many orders went through each channel last week. Far fewer can tell you why a guest bought what they bought, or what nearly stopped them. We sat down with Luke from Sprint Valley, who has run more than 100 digital sales experiments across kiosk, delivery and app for global QSR brands, to talk about the gap between what brands measure and what's actually shaping the order.
You've said brands measure transactions and channel volume but miss what's really driving the order. What are they missing?
The thing most teams optimise for is conversion. Did the guest complete the order, yes or no. It's a sensible thing to track and most brands are good at it now.
But conversion is binary. It tells you they transacted. It tells you nothing about whether they bought the right things, or whether they bought less than they wanted to.
That second question is what I'd call commercial optimisation, and almost nobody is doing it deliberately. The brands I meet monitor average ticket, but very few treat the shaping of it as a discipline. They leave it to chance and it’s the gap between conversion and commercial performance where two to five percent of revenue sits, untapped.
Here's a different way to frame it. Growing average ticket by five percent does the same thing to your P&L as growing same-store transactions by five percent. The difference is you don't have to keep spending to keep the benefit. Yet it gets a fraction of the attention.
You're well known for saying you can't just ask guests what works. Why not?
Ten years ago I increased the price of a client's offer from £18.99 to £19. Sales went up 23 percent.
What bothered me wasn't that it worked. It was that nothing I could have asked a customer would have led me there. No focus group, no survey, no interview would have produced "round the price up by a penny." The result was real, it was commercial, and it was completely unsupported by anything a guest could have told me in advance.
That disconnect is the whole reason Sprint Valley exists. People are brilliant at telling you what they think theydo and terrible at predicting what they'll actually do. Stated preference and revealed preference are different things, and the money lives in the gap between them.
So we don't rely on what guests say. We watch what they do, form a hypothesis, run a controlled experiment, and let the behaviour settle the argument. It takes a bit more patience than a survey. It's the only thing we've seen reliably move the needle.
Where does that show up most? What's the friction guests never tell you about?
If a guest abandons an order because the app crashed, you'll hear about it. It shows up in NPS, it shows up in support tickets, someone fixes it. That friction is loud, so it gets managed.
The expensive friction is a lack of decision confidence. A guest who finds the decision slightly too hard doesn't complain. They just pick the safe, cheaper option, skip the upsell, and leave. Everything looks fine on your dashboard. Conversion held. NPS held. But the basket was smaller than it should have been, and you'll never know, because nobody was unhappy enough to say anything.
In our research we've seen well-intentioned kiosk design reduce average check by as much as seven percent. Not through a bug. Through choices about what to show, in what order, and how. There is no neutral way to present a choice. Every decision you make, which products appear, the sequence, the imagery, how price is shown, shapes what ends up in the basket. If you're not designing that deliberately, you're still influencing it. You're just doing it by accident.
A lot of QSR ordering is habitual. How much of guest behaviour is actually a choice?
There's a well-evidenced pattern called status quo bias. Whatever the options, people lean heavily towards the default and the thing they've ordered before. It gets stronger the more choice you put in front of them, because more choice means more risk of regret, so they retreat to what's safe.
For a brand with decades of presencein someone's life, that's a real challenge. If I've ordered the same meal for twenty years, how do you get me to even consider the new premium item? You can't out-argue a habit. You have to change what the guest meets first, and how confident the new option feels.
McDonald's India is a good public example. They reduced the decision effort in the journey and added 1.4 million transactions across 300 stores in twelve months. Not by adding choice. By making the right choices easier to make.
And here's the bit people miss: easier decisions aren't just commercially better, they're a better guest experience. Higher decision confidence predicts more satisfaction and less regret. When you guide someone well, they get the treat they actually wanted, faster. Guest, franchisee and brand all win. The idea that piling on more options is the customer-friendly thing to do is, in my experience, usually backwards.
What changes when AI enters the order journey? Voice, agents, all of it.
A lot, and not in the way most of the conversation assumes.
The reflex is to treat AI as a faster front end. A voice assistant that takes the order, an agent that reorders for you. Useful, but that's the conversion lens again. The deeper shift is who is actually making the choice.
When a guest browses a menu, the choice architecture is doing the guiding. When an agent orders on the guest's behalf, that agent becomes the thing you have to influence, and it doesn't respond to appetising imagery or a clever upsell sequence the way a person does. The behavioural levers that have shaped baskets for years don't all transfer.
So the question I'd put to any QSR leader heading into this is simple. If a machine is increasingly ordering on your guest's behalf, do you understand what drives its choices the way you've spent years understanding what drives a human's? Most brands don't, yet. The ones who'll do well in 2026 are the ones treating that as an experiment to run now, not a feature to bolt on later.
The same is true for voice ordering - part of the reason kiosk outperforms counter is the consistency of upselling. How this translates to voice needs careful thought - I think the most powerful AI ordering interface would blend voice and visual - a responsive, carefully curated set of choices that feels guided and simply presented.
If a brand wants to close this gap, where do they start?
Start where you have the data and the control to actually act, which is your direct channels. On a third-party platform you're renting someone else's rails and fighting the fridge for the drink and the freezer for the dessert. On your own kiosk, app and web you can see the journey, change it, and measure the result on every transaction.
Then do three unglamorous things. Audit how many decisions you're actually asking a guest to make in each channel. Watch real people navigate it, because what they tell you and what they do will differ. And test your way to the answer rather than debating it.
About Luke Battye, founder of Sprint Valley
Luke is the founder of Sprint Valley, a behavioural science and experimentation consultancy working with global QSR brands on digital sales growth.
Advisor to McDonald's, Just Eat, KFC (Global), Gordon Ramsay Group, Mitchells & Butlers Group, and Molson Coors.
Over 100 controlled experiments on how customers make decisions in QSR digital ordering channels. 10+ years working with global brands across Europe, Latin America, and Asia.
www.sprintvalley.com
