Praveen Manchi
Journal№ 03
  • UX Design
  • Case Study
  • Best Practices
By Praveen Manchi · Aug 15, 2024 · 13 min

UX Mistakes That Cost Millions: Lessons from Real-World Failures

Three Citibank employees signed off on the same $900 million transfer in 2020 because a loan-servicing interface made an unusual payoff look routine - a case that took a federal appeals court two years to fully unwind. This piece checks five of UX’s most-cited failure stories against primary sources and corrects the numbers and endings that don’t hold up in the original tellings.

UX Mistakes That Cost Millions: Lessons from Real-World Failures

UX MistakesCostly Failures

  1. Confusing Confirmation

    Citibank's Flexcube screen let three reviewers approve a nearly $900 million payoff when they meant to wire $7.8 million in interest.

  2. One Fewer Step

    Replacing a forced "Register" button with "Continue" lifted conversions 45%, an estimated $300 million over the following year.

  3. Friction for Danger

    Hawaii's live missile alert sat beside a near-identical drill option with no confirmation step; a correction took 38 minutes.

  4. Staged Rollouts

    Snapchat's 2018 redesign shipped to everyone at once; daily active users fell from 191 million to 188 million.

[ FAILURE MODES ]

Each failure hit an untested moment

Clippy, Citibank, the $300 million button, Hawaii and Snapchat were built by capable teams. They failed at specific moments - an unusual transaction, an interrupted workflow, a rushed alert, an unwanted redesign - that user testing exists to catch.

  1. Unclear confirm
  2. Forced step
  3. No friction
  4. Big-bang launch
On this page

Three Citibank employees independently reviewed the same $900 million transaction in August 2020 and all three approved it - the loan-servicing software's confirmation screen was confusing enough that trained professionals couldn't tell they were about to pay off a loan eight years early. The interface functioned exactly as coded. It just made an unusual, catastrophic action look routine. That single case set off a two-year legal fight, and it's one of several documented UX failures at large companies with a paper trail precise enough to show exactly what the interface did wrong.

1. Microsoft's Clippy: Research Misapplied

The mistake: Clippy - officially "Clippit" - shipped with Office 97, built on real human-computer-interaction research from Stanford: Clifford Nass and Byron Reeves' "Computers Are Social Actors" work had shown people respond socially to interfaces that display personality. Microsoft's implementation missed the nuance. Clippy interrupted mid-task based on a probability model that frequently guessed wrong - popping up to ask "It looks like you're writing a letter" while a user was doing anything else - and gave no way to learn from being dismissed.

The cost: Clippy became a fixture of "worst tech" retrospectives, including a spot on Time's list of the 50 worst inventions. Microsoft disabled it by default starting April 2001, kept it as a buried optional feature through Office XP, and removed it entirely in Office 2007 - a decade of quiet reputational drag before it was fully gone.

2. Citibank's $900 Million Wire Transfer

The mistake: On August 11, 2020, Citibank was servicing a Revlon loan and intended to wire $7.8 million in interest to lenders. Using the Flexcube loan-servicing platform, employees instead triggered a full principal payoff - sending nearly $900 million, years ahead of the loan's 2023 maturity. Citibank's own "six-eyes" process required three separate employees to review the transaction; the interface convinced all three that the amount they were entering matched their intent.

The cost: Ten lender groups refused to return a combined $504 million, arguing the payment discharged the debt. A district court sided with them in February 2021, but the Second Circuit Court of Appeals unanimously reversed that ruling in September 2022, restoring Citibank's claim to the funds. The parties settled two months later, with roughly three-quarters of the disputed amount returned to the bank - meaning the interface error still cost Citibank tens of millions of dollars in unrecovered funds, on top of two years of litigation, even after the reversal.

3. The $300 Million "Continue" Button

The mistake: In 2009, usability researcher Jared Spool's firm, User Interface Engineering, studied a major e-commerce retailer whose checkout forced first-time shoppers to register an account before buying anything. Session recordings showed a specific failure mode: returning customers who'd forgotten which email or password they'd used were abandoning checkout rather than resetting a password.

The fix: UIE replaced the "Register" button with "Continue" and added a line clarifying that an account wasn't required. Nothing else on the page changed.

The result: Conversions rose 45%, adding $15 million in the first month and an estimated $300 million over the following year - from renaming one button and removing a forced step. Spool never named the retailer publicly, but the case remains one of the most cited data points in UX research for why a single unnecessary decision point can outweigh an entire redesign.

4. Hawaii's False Missile Alert

The mistake: On January 13, 2018, a Hawaii Emergency Management Agency employee sent a live "BALLISTIC MISSILE THREAT INBOUND" alert to every phone in the state during what was meant to be a drill. The alert system listed nine similarly worded options in one interface, including "PACOM (CDW) - STATE ONLY" (the live alert) directly alongside "DRILL - PACOM (CDW) - STATE ONLY" - differing by one word, with no visual separation and no confirmation step before sending.

The cost: It took 38 minutes for a correction to reach the same phones. Residents abandoned cars on highways and sheltered children in bathtubs believing an attack was underway. Design researcher Don Norman, writing about the incident, identified three specific fixes the interface lacked: a separate, clearly differentiated test mode; a default state of "test" rather than "live"; and a second-person confirmation step before any real alert could send. None of the three existed.

5. Snapchat's 2018 Redesign

The mistake: Snapchat's February 2018 redesign separated friends' Stories from publisher content and reorganized navigation without a staged rollout. The change was disorienting enough that a petition to reverse it collected 1.2 million signatures within days, and Kylie Jenner's tweet calling the app "so sad now" coincided with a single-day $1.3 billion drop in Snap's market value that same month.

The cost: The backlash compounded over two quarters. Snap's May 2018 earnings report - the first full quarter under the new design - showed user growth had stalled so badly that the stock fell roughly 20% in a single trading day. By its August 2018 report, Snapchat posted its first-ever quarterly decline in daily active users, dropping from 191 million to 188 million. CEO Evan Spiegel publicly attributed both slowdowns to the redesign.

Key Lessons for UX Designers

  • Test before a full rollout. Snapchat's redesign shipped to its entire user base at once; a staged rollout would have surfaced the backlash at a fraction of the cost.
  • Removing a step beats redesigning a flow. The $300 million button didn't need new visual design - it needed one fewer decision.
  • Confirmation screens must show what will actually happen. Citibank's six-eyes review failed because the interface, not the process, was the point of failure - more reviewers looking at the same confusing screen doesn't fix the screen.
  • Irreversible actions need friction, not convenience. Hawaii's missile alert needed a confirmation step precisely because it was catastrophic and immediate - the interface treated it like any other dropdown selection.
  • A personality-driven feature still has to read the room. Clippy was grounded in real research about people responding to interface personality; it failed because it ignored context and timing, the two variables that research also said mattered.

The pattern across all five cases isn't a lack of design effort - Citibank, Microsoft, and Snap all had functioning products built by capable teams. It's that none of these interfaces were tested against the specific moment they failed at: an unusual transaction, an interrupted workflow, a rushed alert, a redesign nobody asked for. That's the gap user testing exists to close.

References & Further Reading