The Sweepstake
Challenge
Garmin runs badge challenges inside its Connect app. In the Ahotu Marathon Challenge, a runner joins, records 26.2 miles of running during September, and earns a badge and 15% off race registration at ahotu.com.
When they finish, Garmin gives them a link. That is where Garmin’s flow ends and this case study begins. Everything after that link was ours. Claim, choose, verify, create an account, confirm.
Six screens, and two rules they had to hold at once. Never make the guaranteed reward feel like something you could lose. Never slow down someone who arrived with momentum.
The Garmin side of the journey. The runner joins the challenge in Garmin Connect, earns the badge, and taps Redeem Your Reward, which leads to Ahotu’s claim page. That is where this case starts.
Two rewards, one certain and one not. The design had to keep them from being confused.
Users arriving from Garmin have no account and no familiarity with the site. They arrive with a 15% discount already earned, and an optional sweepstake for one of seven race packages on top of it. Two rewards: one certain, one not. The risk was that people would think entering the sweepstake meant gambling the discount.
Everything after Garmin’s button was ours to design.
Other brands run challenges inside Garmin Connect too: Disney’s 100 Years of Wonder, Starbucks’ March Steps, and ours, the Ahotu Marathon Challenge.
Every campaign is different, but the shape is always the same. A brand builds something inside Garmin. The user completes it. Garmin hands them over to the brand, and that's where this project starts.
Ahotu's business is event registrations, and a Garmin finisher is about as qualified as a visitor gets: someone who has just proved they run, arriving at the moment they've finished. The campaign was a way to turn that moment into an account.
The discount gives a reason to come back and book something. The account makes that person reachable afterwards, and a growing member base is what Ahotu points to when negotiating the next partnership. Arriving from inside Garmin helps too. The handover carries a credibility that a cold ad doesn't.
Success would have shown up in three numbers: accounts created, entries submitted, and discount codes used at checkout. All of it depended on the six screens being frictionless. One link, one arrival, no second chance.
Garmin’s own design files.
Before joining. Finding the challenge inside Garmin Connect, reading the terms, joining.
In progress. Distance tracked automatically, badge locked until the target is met.
Completed. Redeem Your Reward, and then a blank frame. Everything after that button had to be designed from nothing.
Six screens had to do two jobs at once: win over someone who had never heard of Ahotu, and enforce rules they would never see.
Had to convert strangers.
- Cold traffic from a Garmin link, no Ahotu context
- The discount had to feel safe, so entering the sweepstake could never look like putting it at risk.
- The reward had to be easy to find again when the user came back
Had to survive the rules.
- Four ways a link can fail: invalid, already claimed, expired, system error
- Claim check on the account, separate from the code check
- Geographic exclusions enforced at entry and again at draw
How might we make a sweepstake feel trustworthy without ever making the guaranteed reward feel conditional?
The flow took several rounds before it held together.
The first pass was boxes and labels. Package One, Package Two, Package Three, deliberately unresolved. At that stage the job was to find out whether the sequence held, and where a user could fall out of it.
First pass, before the flow was resolved. Packages numbered rather than named, so the sequence could be tested before the content.
A version for Garmin.
Later, once the flow had settled into something we were confident in, a version had to be made for Garmin. They needed to see what happened on their participants' side after the user leaves their app. Not to approve the visuals, but to understand what those users would walk into. Every screen was annotated with its purpose, in sequence, so someone who had never seen the product could follow it start to finish.
Built for Garmin, not for the team. Each screen carries a note explaining what it does and why it sits where it does.
Three principles ran through every screen.
Protect the guaranteed reward. Make the sweepstake choice easy to understand. Make the reward easy to find again once it’s claimed.
The four sections that follow take the flow in the order a user meets it.
The discount comes first, so nobody thinks they have to gamble it.
The three steps below the button set expectations for what follows. The sweepstake appears only inside step one. It is labelled optional, with the discount restated next to it. A user who reads nothing else still sees that the 15% is already theirs.
Four error states.
The same URL resolves four other ways, each with its own copy and its own way forward.
Caught in review, round two.
An invalid country error stayed on screen after the user had picked a valid country. A functional bug rather than a cosmetic one, fixed before launch.
The code the user never sees.
When a Garmin user completes the challenge, Garmin issues a code that identifies them as someone who genuinely finished it. Without that code there's no way to tell a legitimate participant from anyone who found the link.
The early wireframes put it on the landing page, on the assumption that a code has to be visible to be validated. Talking it through with Garmin established that it didn't. The link could carry it, and the page could validate it silently. The check still happens. The user just never has to think about it.
The claim page. Discount confirmed first, sweepstake second and marked optional. One button, three steps, no code.
Seven race packages, one choice, and always a way out.
Nothing opens on the selection screen until the user asks for it. A single closed drawer, a headline, and a skip link, with nothing else competing for attention. Skip the sweepstake stays visible at every stage, so opting out never requires backing out.
One screen gets the full background image.
Every other screen is a task: claim, verify, create an account, confirm. This one is the aspirational moment, so it’s the only place that gets the full image.
One package, not several.
Letting people pick three or four would have felt generous, and it would have been easy to build. But more selections read as better odds, and that isn’t how the draw worked. One entry, one package, the same chance whichever you choose.
The package cards had to fit ten languages.
Shipped across the site's ten languages. French event names cramped metadata that sat comfortably in English, so the card layout had to be built around the longest string rather than the one in front of us.
Drawer closed. One decision on screen, and a way out of it. The 15% discount is restated above so the user is reminded what they keep regardless.
Seven packages. Running, trail, cycling and triathlon. Discipline, name, date and location on every card. Enough to choose from, not enough to read.
The detail modal. What's included sits above the fold, because it's what decides the choice. The CTA stays pinned while the content scrolls behind it.
New and existing users end up in the same place, with the reward waiting.
Two kinds of user arrive at this point: users who already have an Ahotu account and users who don’t. Both end up with the same confirmed entry and the same reward attached to their account.
The country check comes first.
Country of residence is checked against the excluded list before any of that. Before signing in with Google, before creating an account. Asking afterwards would mean telling someone they had been disqualified after they had already done the work.
We asked whether Garmin could run the check on their side, before the link was ever issued. That would have been fairest: an ineligible runner would never have reached us at all. Garmin didn’t want to hold that data, which is fair enough, so the check landed with us. Putting it before anything else was the next best thing.
The reward card had to carry the whole page.
For most of these users, the account is brand new. They are looking at their Ahotu profile for the first time, and the only thing on it is the reward they just claimed. So the reward card had to carry the whole page: state clearly what they have, when it expires, and how to use it, without assuming any familiarity with the rest of the site.
Not one profile page, but three.
The profile page fetches the user's reward data, so it always needs a loading state. The question was what that loading state should look like, and the answer turned out to depend on who was arriving. Users could reach the page in three different conditions:
- Both rewards, for someone who entered the sweepstake and still has their discount.
- Discount only, for someone who skipped the sweepstake.
- Sweepstake only, for someone who entered and has since used the discount.
A single placeholder layout with two cards would have worked for one of those and misled the other two: the page promises two cards, then resolves into one, and the user is briefly told they have something they don't. So each state got its own placeholder as well as its own resolved layout, all three specified and handed over rather than left for the developers to work out.
Country first. The selected package is confirmed at the top so the choice isn't lost. Excluded countries are named in plain text above the field rather than discovered by selecting one. Terms sit directly above the CTA, which stays disabled until both are satisfied.
Log in or create. One screen for both kinds of user. The heading says what the account is for: to get your rewards. The step is part of collecting the reward, not a gate standing in front of it.
Loading. Placeholders match the shape of what's coming, so nothing moves when the data lands.
Three states, specified. Each one drawn twice, loading and resolved.
Both rewards. What the user lands on once the account exists. Sweepstake entry with the date results are announced, discount code with its expiry and a copy control.
Four emails, each written for the path the user took.
- Confirmation with both rewards.
- Confirmation with the discount only, for users who skipped the sweepstake.
- A message for users who were not selected, sent to everyone entered who wasn’t drawn.
- A winner email, which goes to a handful of users and was left deliberately simple at this stage.
The not-selected email was the one worth spending time on.
It goes to the largest group by a wide margin, and it's the only message in the set that has to deliver disappointing news. It leads with the effort rather than the result, and it puts the discount back in front of the user at exactly the moment they might otherwise disengage.
Both rewards. Same hierarchy as the profile page: sweepstake entry with its announcement date, discount code with its expiry. A user who sees one surface recognises the other.
Discount only. For users who skipped the sweepstake. Nothing in it hints at what they opted out of.
Not selected. The result is stated once, briefly, and the rest of the email is about what the user still has.
The badge.
The final flow looks obvious. Getting there wasn't.
Several ideas were drawn or built before being dropped. Each looked right on paper. Building them was what made the problem visible.
The best-looking card tested worst.
- A Show more control that expanded in place.
- A card that was simply clickable, with everything behind the tap.
- A card carrying the event photo as a full background, borrowing the aspirational pull the user arrived with.
The background treatment looked best and tested worst. The image competed with the metadata a user needed to compare seven options. What shipped keeps the photo small and the text legible, and moves the aspiration into the modal, where there’s only one event to look at.
Multiple selection was built and then cut.
The states were drawn, the component worked, and it still shipped as one selection, because more entries would have looked like better odds the draw never gave.
First prototype. Everything in this version calls it a raffle. It shipped as a sweepstake, a change that touched every screen, every email, and the terms.
Multiple selection, before it was cut.
Notes on the work. A later pass, annotated in place: which terms apply where, what happens on skip, where a loading state was still missing.
Five people tested it cold, and all five understood both rewards.
Five people tested a clickable prototype of the full flow, not the live site. None of them had seen any part of the project: colleagues had been through it often enough that their reactions weren't worth much. Each was given a set of tasks to work through and reported in a Google Form as they went, rather than being observed.
All five correctly identified that they had received both the discount and the sweepstake entry. That was the one thing the flow had to get right, and it held. The useful findings were elsewhere.
The idiom stayed, even though one tester couldn’t parse it.
A tester who was still learning English couldn't parse two phrases on the claim page: bucket list, and either way. Both are idiomatic, and neither is taught early.
What made this worth taking seriously is where those words sit. Either way is the phrase doing the heaviest lifting anywhere in the flow. It's the sentence that tells a user the discount is safe no matter what they choose.
We discussed replacing both, and kept them. The audience is global but the product ships in English, and the plainer alternatives read as instructional in a way that didn't match the tone of the rest of the page. A plainer alternative would have been “Your 15% discount is yours, whether or not you enter the sweepstake.” It's less playful, and it carries the guarantee without relying on an idiom.
The next step would be to test both with more readers at different levels of English, rather than settling it on one tester's reaction.
The link was right, the content wasn’t.
A tester selected an event, wanted to know what the prize actually included, followed the link, and landed on the event page, which says nothing about the package.
The instinct was to question the link. The link was fine: someone considering a race genuinely does want to read about the race. What was missing was the package detail on the modal itself.
So the modal was expanded. What the package includes now sits near the top, instead of being something a user has to go looking for.
Two bugs showed up.
A discount code stayed in its loading state far longer than it should, which the developers looked into. And a support link pointed nowhere, because it hadn’t been connected in the pre-production build. Finding both before launch is exactly what the test was for.
Two findings were real but outside the flow.
The terms link pointed at Ahotu's general terms because the sweepstake terms hadn't been built yet, already scheduled and left alone. And a newsletter popup during login turned out to be running on the tester's own machine, not on the page. Worth chasing both down rather than assuming. A finding that isn't yours is still only knowable after you've checked.
One tester, three real problems. Response to: was anything confusing or missing when browsing the packages?
Bug reports. A discount code that never resolved out of its loading state, and a support link that pointed nowhere.
The flow is live.
A Garmin participant with no Ahotu account becomes a member, holding a reward they can find again, in six screens.
Delivered: the end-to-end reward flow, account and profile states across ten languages, and four email variants.
Validated: all five testers understood that the discount and the sweepstake entry were separate rewards, and that the discount was theirs either way.
Not yet established: whether the flow improved account creation, completion or redemption in production. That needs post-launch data I don't have yet.
Who you test with decides what you learn.