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.
Summary
Experimentation can be intimidating to non-data science folk. But Erin wants to get everyone excited about A/B testing. In this talk, Erin shares the Conversion Design process. It centers A/B testing as a way to gather high-quality evidence to make highly informed decisions to improve your digital product. She also introduces the Good Experimental Design toolkit. These easy-to-follow templates usher teams through the logic needed to design trustworthy experiments that you can learn from.
Key Insights
-
•
Conversion design combines design, science, and business to intentionally create measurable improvements, not just arbitrary changes.
-
•
Ronald Fisher’s 1919 work exposed how poor experimental design led to decades of unreliable scientific data.
-
•
Conversion originates from the Latin word meaning to transform or change, which is broader than just sales or profit.
-
•
Traditional linear product development processes miss the complex, iterative nature of real-world systems.
-
•
Systems thinking offers a more accurate way to understand and manage design experiments within interconnected environments.
-
•
A rigorous conversion design process includes seven phases: understand, hypothesize, prioritize, create, test, analyze, and decide.
-
•
Experiments grounded in well-documented hypotheses based on research have higher success rates than guesses.
-
•
Randomized 50/50 AB testing is the gold standard for isolating the true effect of design changes by evenly distributing confounds.
-
•
Experimentation buckets—product foundations, content/motivation, accessibility/usability, and bug fixes—help prioritize work effectively.
-
•
Ethical considerations and guardrail metrics are crucial to ensure changes benefit all stakeholders sustainably and without manipulation.
Notable Quotes
"Conversion means change, not just sales or profit."
"Design is the rendering of intent — bringing ideas into form that solve the problem."
"Decades worth of agricultural experimental data was garbage because of poor experimental design."
"Most product teams stay on the bottom rungs of evidence, relying on opinions or observational data instead of randomized trials."
"You can never purely test an idea, only the implementation of the idea."
"Randomization is magic — it evenly distributes confounds so the observed effect is caused by your change."
"Not all changes create value; some do nothing or even make things worse."
"You have to think critically about how a change impacts all stakeholders, not just the main business metric."
"Ethics evolve faster than laws; just because something is legal doesn’t make it ethical."
"If an experiment I design made the front page news tomorrow, how would I feel about it?"
Or choose a question:
More Videos
"You’re constantly training people and optimizing protocols even after successful implementation."
Sofia QuinteroBeyond Tools: The Messy Business of Implementing Research Repositories
March 10, 2022
"As the product journey progressed, research cadence got faster with prototypes and quicker feedback loops."
Dr Chloe SharpUsing Evidence and Collaboration for Setting and Defending Priorities
November 29, 2023
"The three in the box model is business, design, and technical representation, plus legal and compliance in regulated environments."
Frank DuranPartnership Playbook: Lessons Learned in Effective Partnership
January 8, 2024
"If you're changing workflows in experiments, you need to keep track of impacted teams like call centers to avoid resistance."
Steve Sanderson Alissa Briggs Jeff Gothelf Bill ScottDiscussion
May 14, 2015
"Building generative AI tools is not scary, it’s just very different from the last 20 years of building products."
Peter Van DijckBuilding the Rosenbot
June 4, 2024
"Screen readers let us jump from one heading to another, replicating the experience of skimming a page visually."
Sam ProulxSUS: A System Unusable for Twenty Percent of the Population
December 9, 2021
"Sometimes you have all the best ideas, but it’s not the right time, and that’s OK."
Jacqui Frey Dan WillisPanel Discussion: Integrating DesignOps
November 7, 2018
"It’s not enough to just write down principles—they must be operationalized and made actionable for different teams."
Sarah WilliamsVerizon_A Framework for CX Transformation
January 8, 2024
"The LLM gave many different answers for the same passage, each time reporting it was one hundred percent confident."
Trisha Causley[Demo] Complexity in disguise: Crafting experiences for generative AI features
June 5, 2024
