A self-install flow that matched the box in the customer’s hands
AT&T was shipping redesigned hardware packaging, and the in-app install flow no longer matched what customers were actually unboxing. That mismatch was generating very expensive support calls at national scale.
Senior designer, mobile apps
MyAT&T, Smart Home Manager
Millions of customers
Legal, care, packaging, retail
Month-over-month reduction in self-install-related support calls.
Faster average page load, from reworked visual assets and Lottie files.
Aligned on a single flow — legal, care, packaging, retail, engineering.
Every failed self-install is a phone call somebody pays for.
Self-install exists to keep a technician’s truck out of the customer’s driveway. When it fails, the customer calls care, and the cost of that call plus the eventual dispatch dwarfs whatever the flow saved. Internet providers had earned their reputation here, fairly.
AT&T was mid-refresh on hardware packaging, which set the timeline for me: if the app told customers to look for a label that no longer existed, calls would spike the week of launch. Hard external date, no negotiating.
It isn’t really a flow. It’s a physical task with a screen next to it.
The existing sequence followed internal process order. Customers experience the task in the order the box hands it to them — unwrap, identify, connect, wait, verify. So I re-sequenced the entire flow to the hardware and made the app behave like a second pair of hands instead of a form to complete.
That one reframe settled most of the content argument, because it handed every team an objective test: does this screen match what the customer is physically holding right now? Much easier conversation than debating preferences.
One action per screen.
A customer kneeling behind a TV cannot parse a checklist. Each step asks for exactly one physical action and confirms it before moving on.
Visuals matched to the real hardware.
Illustrations and Lottie animations rebuilt against the new packaging, so “the white cable” was, in fact, the white cable.
A recovery path at every failure point.
Instead of dead-ending into “call support,” each step leads with the most likely fix — which is where the call reduction actually came from.
Lower stress for a high-stakes flow
Walk through the updated prototype on the right to see how I took a disconnected and confusing flow and made it into an easier, guided experience.
A few prototype notes;
Reset the prototype to the initial page at any time using the controls under the prototype frame.
Error states are peppered throughout this flow. Try tapping on help links or dismissing messages to see where they populate.
Of course, the animation and imagery in the real flow are proprietary and owned by AT&T. I’ve swapped in simpler CSS versions here to demonstrate what a user sees while cooperating with their policies.
Twelve teams, one flow, a fixed date.
Legal, care, packaging, retail, motion graphics, and engineering all had legitimate claims on this flow. Every review round threatened to add another screen. I used the physical-task principle as my scope defense: content that didn’t help someone finish the step in front of them went to a tooltip or a help article, not into the flow itself.
Care’s call-reason data became my prioritization input, which made the reviews about evidence instead of taste. It also made me a lot of friends in the care org, which never hurts when you need a favor two quarters later!
The scope defense was never “no.” It was “not directly in the flow, but here’s exactly where it should go instead.”
28% fewer calls, and a faster app on the way there.
Service-center data showed a 28% month-over-month drop in calls from customers who hit an impasse mid-install. Alongside the flow work, I did an extensive rework on visual assets and Lottie files across MyAT&T and Smart Home Manager (as many needed to be updated or replaced), cutting average page load by 1.4 seconds. That time matters a lot when your customer is on the connection they’re still trying to set up.
The lesson I took forward: in a flow with this many stakeholders, the most valuable design work is establishing the one principle everybody can test their request against.