Sprout Spend
This case study is about the sticker that travels from the POS to the cup to your hands. It's a small sticker, but it carries a tremendous amount of operational weight. When the label is hard to read, drinks get made wrong. When drinks get made wrong, they come back. When they come back, everything slows down. Customers wait longer. Baristas lose their rhythm. Product gets thrown out. The company loses money. And somewhere in the middle of all that chaos, someone is just trying to get their coffee and get on with their day.
Context
There's a rhythm to working a bar at Starbucks. Once you get into the flow of working there, you can make drinks almost without thinking. Your hands move faster than your brain, and orders get finished so fast they seem to fly off the counter.
Then you hit a block.
You get a ticket that is longer than a novel, and your whole system shuts down for a second while you try to figure out where you should even start. Lines and lines of modifications. Extra shot, oat milk, no foam, light ice, two pumps vanilla, one pump hazelnut, 3 pumps mocha, add whipped cream, add cookie crumbles, and add caramel drizzle. There are 10 more orders just waiting in the que that you haven't started yet. Customers are lined up at the handoff station, waiting for their drinks. Your coworker just had to remake a drink because they inevitably missed a topping, causing a bottleneck on the whole flow. And you're staring at a label still trying to figure out where to start, hoping you don't mess it up, because it would take so long to remake.
This is the moment I want to design for.
Industry analysis in 2025 estimated that at a ~94% order accuracy rate, Starbucks was producing roughly 38 million remakes per year, at an estimated product cost of $68 million annually.
Research
Wrong orders aren't just a customer satisfaction or worker issue, they're a measurable operational cost. Industry data on order accuracy and remake rates points to a problem larger than most customers realize.
Data from QSR Magazine’s annual drive thru study shows that overall order accuracy at major fast food chains has fallen from 89.4% to 84.4% in recent years.
Industry analysis in 2025 estimated that at a ~94% order accuracy rate, Starbucks was producing roughly 38 million remakes per year, at an estimated product cost of $68 million annually.
Whether the $68 million estimate is perfectly precise or not, the order of magnitude is credible. And product cost is only part of the picture. Each remake also costs:
• Barista labor time (typically 2–4 minutes per remake)
• Queue disruption: the drink that should have been made next is delayed
• Customer experience damage: the wait, the interaction, the doubt about future orders
The cost of a better label is a one-time design investment. The cost of the current label is paid daily, in every store, with every wrong drink.
Chart of the normal work flow of an order from start to finish.
Chart of the work flow when order gets remade from start to finish.
Example of different types of coffee shop and fast food order tickets
Current Label Analysis & User Interviews
Working within the real constraints of thermal printing (no color, roughly 2" wide), I researched how baristas actually read tickets: scanning, not word-by-word. I mapped the order information needs to flow: Drink size, type of cup (based on iced or hot), drink base, syrups, shots, milk, toppings. The current label presents information in the order a customer spoke it. That's not the same thing. (see example to left)
Due to constraints, I am only able to work with typographic hierarchy, iconography, and spatial grouping as design levers, all achievable with a label template update and no hardware changes required. I tested Bionic Reading and ruled it out; the research doesn't support it as an accuracy aid in trained, high-speed reading contexts. An effort/impact matrix helped focus the work on changes that are both high-value and immediately implementable.
Through conversations with current Starbucks workers, a pattern emerged quickly. The mistakes aren't random. They cluster around specific, predictable failure points:
• Milk type: oat, almond, coconut, soy, nonfat, 2%, whole. Imagine someone with lactose intolerance. A miss like that is enough to ruin a customers day.
• Syrup pump counts: "2 pumps vanilla" and "3 pumps vanilla" look almost identical at speed.
• Toppings and drizzles: these appear at the end of the label and are the first thing the eye skips when rushing.
• Temperature: "iced" being the same text size as the drink base is easy to miss in a hurry.
These aren't failures of attention or care. They're failures of design. The label does not support the speed at which it needs to be read.
The current ticket design with numbers representing the order in which the drink is made
Customer Perspective
Barista Perspective
"I usually miss toppings the most since they are usually at the bottom and it's so hard to remember"
- Current Starbucks barista (user interview)
"When it's slammed, I'm not reading the ticket word by word. I'm scanning it. If something doesn't jump out at me visually, I'm going to miss it."
- Current Starbucks barista (user interview)
Time to Design
The first round of redesigns focused on establishing a clear typographic hierarchy on the existing label format. The goal was to answer, at a glance: what is this drink, what milk is in it, and what else is important?
The first drafts were deliberately minimal, just reorganization and font weight changes, no icons. This established a baseline for testing. Could purely typographic changes meaningfully improve read speed and accuracy?
Next, I created a custom icon set so I had specific metaphors to test on labels. These icons were created to be in brand with Starbuck’s typography and round logo form.
I created another round of variations before testing my designs in front of users. The A/B testing ultimately would help me figure out how many icons benefit baristas, and how many is too many. Iconography language aimed to improve speed can easily cause too much visual noise when done incorrectly, which takes away the benefits.
User Round 1 and 2 A/B Testing & Design Iterations
A/B testing was conducted with 9 current and former Starbucks baristas. Participants were shown pairs of labels — one with icons, one without. One with boxes, one without. They were asked to not think too hard about it, and on instinct, choose which was quicker for them to read. I told them to picture what it was like being in a rush.
The results showed that box was preferred over no box. Also, the icons to differentiate between sections were not preferred over the plain text. Also, the boxes separating steps was preferred.
One current Starbucks worker stated that the icons would be most helpful when there is a longer list of the same item, which means this concept is worth continuing exploration on. Ex. 4 vanilla, 3 mocha, 1 caramel etc. The other suggestion was to include flags for important changes, ex. a non cows milk for someone who’s lactose intolerant. Also some more bold text instead of just the larger size.
I then changed the designs to be longer tickets, and did another round of AB testing, this time with 7 participants.
The results remained the same. Workers liked separated content, and did not like the icons on the ticket, other than for hot/cold and flagged modifications.
Tickets ready for Round 1 of A/B Testing
Round 1 of A/B Testing Results
“As nice as it is to have the visuals and boxes, to me, on bar it’s too much visual language to process when I get an order. I just need to bare information to make the drink.”
— Current Starbucks Employee
Round 2 of A/B Testing Results
Final Designs
The final design applies typographic hierarchy, and tastefully integrates the icon system for the hot/iced drinks, and a flag for special intolerances like no cheese for a food item, or a plant based milk modification. The order of ingredients lines up with what goes in the cup when, and like is grouped with like to keep the barista’s brain organized by section. The name zone is proportionally larger than in the current design as well.
All of this works within thermal printing constraints. The design is achievable with only a label template update.
Training Material
I decided to add a training material that goes over how to read orders, and what things to look out for. This helps bring attention to the most important things that new baristas might not have on their mind when first getting used to the bar. Learning bar is one of the most overwhelming parts of training at Starbucks, so I want to make this as easy as possible.
Final Takeaways
How Would Starbucks Implement This?
Implementation means updating the label print template in Starbucks' POS system, a software change rather than a hardware one. Icons would be embedded as printable assets, and a single visual training page would introduce baristas to the new system.
Starbucks has already shown willingness to invest in operational tech when ROI is clear, as Green Dot Assist demonstrates. A label redesign is far simpler and targets the same problem the company is already spending to solve: order accuracy and remake reduction. The most realistic path is piloting in select stores, measuring remake rates and barista confidence, then rolling out. That is how Starbucks typically tests operational changes.
How Does This Directly Impact Bottlenecks?
The bottleneck at the bar is not the espresso machine. It is the moment a barista pauses to re-read a ticket. Every pause breaks flow, and multiplied across a rush-hour queue, those pauses add up to real throughput loss. A label that communicates priority information at a glance keeps the bar moving, which correlates directly with speed, accuracy, and barista wellbeing.
Is This More Accessible to All Workers?
Not all baristas read at the same speed or process visual information equally fast. New hires are especially disadvantaged by the current design, which relies on memorized abbreviations. A label with icons and clear hierarchy is more legible for everyone, but most valuable for workers still building fluency. The principle comes from accessible design: optimize for the most constrained user and you improve the experience for all. A label a first-week barista reads correctly is one a five-year veteran reads even faster.
Improvement for Customer Experience
Fewer wrong drinks means fewer waits, fewer disappointments, and fewer awkward returns. A clearer label also lets customers self-verify, confirming their milk type and modifiers before walking away. That small act catches errors before they become complaints.
Starbucks is working to rebuild customer trust after years of declining satisfaction scores. The label is a small touchpoint, but small things across thousands of stores are not small.
Credits & Sources
Research Sources
QSR Magazine Annual Drive-Thru Study. (2019). Order accuracy data across major QSR chains. qsrmagazine.com
Intouch Insight. (2025). Drive-thru accuracy benchmarks by chain. As reported in Fortune.
Niccol, B. (2025, October). Remarks on Green Dot Assist and operational accuracy. Salesforce Dreamforce. As reported by Fortune.
International Journal of Research and Innovation in Social Science (IJRISS). Font Matters: Investigating the Typographical Components of Legibility. rsisinternational.org
Snell, J. (2024). No, Bionic Reading does not work. Acta Psychologica. ScienceDirect. doi.org/10.1016/j.actpsy.2024.104322
Možina, K., Kovačević, D., & Blaznik, B. (2025). Usability of Bionic Reading on Different Mediums: Eye-Tracking Study. SAGE Journals. journals.sagepub.com/doi/10.1177/21582440251376158
Singer Trakhman, L.M. (2022). Can Bionic Reading make you a speed reader? Not so fast. The Conversation. theconversation.com
Doyon, T. (2022). Does Bionic Reading actually work? Readwise Blog. blog.readwise.io
Starbucks Newsroom. (2025, June 11). Meet Green Dot Assist. about.starbucks.com
DigitalDefynd. (2025). 10 Ways Starbucks Is Using AI. [Note: industry analysis site — used for context on scale estimates; specific figures unverified by Starbucks]
Acknowledgments
Special thanks to Matt Glynn for guidance and mentorship throughout this project.
Thank you to the current Starbucks baristas who participated in user interviews and A/B testing. Your feedback allowed this to be made for you in mind.