Log in or create a free Rosenverse account to watch this video.
Log in Create free account100s of community videos are available to free members. Conference talks are generally available to Gold members.
UX Lessons from running more than 1,200 A/B Tests
Summary
Knowing how to solve the right problem is only one small part of the product-success equation. And nailing the execution is harder than most people think. As Erin says in this talk, “There are far more ways to fail than there are to succeed” when it comes to experimentation. To help you learn, Erin shares some of the silly mistakes she made while A/B testing her designs. That way you can avoid those pitfalls as you work to make your digital product better—not just different.
Key Insights
-
•
Nine out of ten AB tests at Booking.com fail, but the few that succeed compound into massive growth.
-
•
Execution often fails on good ideas rather than the ideas themselves.
-
•
Technical details like page load time can negate positive design changes if not carefully managed.
-
•
Edge case bugs in different languages or currencies can silently kill conversions.
-
•
Minor typography choices, such as switching from Times New Roman to Arial, have a large impact on conversion.
-
•
Using system fonts improves legibility and page load speed, positively affecting conversions.
-
•
Tracking should only include users exposed to the tested change to avoid noisy data.
-
•
Qualitative user research drives many experiment ideas by revealing real user challenges.
-
•
Guardrail metrics like customer service calls and loyalty prevent short-term wins from harming long-term value.
-
•
Products should be retested over time as technology and user behaviors evolve, since what fails today may succeed tomorrow.
Notable Quotes
"I’m actually a really big failure — nine out of ten tests fail with no positive measurable impact."
"Compound effect means good upon good upon good eventually builds to incredibly fast growth."
"If a problem keeps coming up, it’s usually the execution of the idea that’s failing, not the concept."
"Increasing page load time by three seconds can nullify all your design improvements."
"Picking the wrong size image can impact customer experience as much as what you see on screen."
"There are as many versions of the website as stars in the sky because of language, currency, and device variations."
"No tracking is perfect because numbers tell you what likely happened, but not why it happened."
"System fonts load faster, improve legibility, and outperform fancy brand fonts in conversion."
"Guardrail metrics help catch unintended consequences like increased customer service calls despite higher conversions."
"You can’t fail 90 times out of 10 without laughing at yourself to keep going."
Or choose a question:
More Videos
"The AI assistant starts with reasoning and not results, engaging users in intelligent conversation to understand needs."
Kritika YadavOptimizing AI Conversations: A Case Study on Personalized Shopping Assistance Frameworks
June 10, 2025
"Evals are everywhere, right? Everybody's talking about evals. It is like one of the key things in developing useful AI products."
Peter Van DijckHands-on AI #2: Understanding evals: LLM as a Judge
October 15, 2025
"Dual track agile means you have the dev sprint track and a separate design/product track that leads the dev track by several cycles."
Toby HaugDiscussion
June 9, 2017
"We use storytelling to turn accessibility from just compliance into a business differentiator that executives care about."
Saara Kamppari-Miller Nicole Bergstrom Shashi JainKey Metrics: Comparing Three Letter Acronym Metrics That Include the Word “Key”
November 13, 2024
"Do we have the feedback loops to feel our way through paths of emergence rather than pre-carved routes?"
Dave MaloufTheme 3: Introduction and Provocation
January 8, 2024
"Most designers join AI projects after the data has been discovered and the model built—that’s far too late for meaningful innovation."
Jodi ForlizziDesign and AI innovation
June 5, 2024
"We decided to limit panel access initially to the research team but are exploring expanding to other teams."
Wyatt HaymanGlobal Research Panels
August 8, 2020
"I don’t care about what’s actionable for a company. I care about what’s actionable for my client’s career and influence."
Robin Beers Sonja Bobrowska Mujtaba Hameed Josh MoralesHow to create actionable insight in the face of politics and silos [Advancing Research Community Workshop Series]
October 12, 2023
"Working on a different word than empower: People already have power, we have to respect it."
Bria Alexander Ariel Kennan Charlotte Lee Sarah Brooks Emily Lessard Gordon Ross Joanne DongReflect and Chart Forward
December 10, 2021
