Simpson’s paradox occurs when we observe a certain trend in the aggregate data but not in the underlying segments that comprise the data. In the A/B testing domain, Simpson’s Paradox can occur when the overall mean conversion rate and/ or average order value of the experiences tested point to a result different from the mean conversion rates and/ or average order value of the underlying segments.
Let me illustrate this with an example from Georgi Georgiev’s blog post, instructor at CXL. Suppose you run an A/B test between Page A and Page B and see the following results:
I have been exploring how experimenters* work in and with product teams. Towards that end, my last post sought to demystify Scrum, especially its three pillars of transparency, inspection, and adaptation and covered how experimenters can align their processes with Scrum events and artifacts. This post will cover ways product owners can use experimentation to refine product backlog.
A product owner essentially creates a roadmap for a product with the aim of delivering value to customers. This roadmap known as the product backlog is a prioritized list of items that when developed is expected to deliver that value to customers…
This post covers some Scrum vocabulary and processes that product teams use and how they align with the process experimenters’ use.
Let’s start with a basic question — why do experimenters need to understand Scrum?
Experimenters need to understand Scrum because a large number of organizations are using Agile & Scrum practices and you are likely at one such organization. Typically, such organizations are focused on adding value to their customers through their product. Running experiments provides insights all throughout the product lifecycle from discovery to exploration to validation. …
The Weekly Buzz has gone through a few iterations already in true experimentation spirit — not really backed by data yet, but we’ll get there. For now, in case you are here and want to catch up on some of the important posts in the last few weeks, here they are on LinkedIn.
The LinkedIn posts have takeaways from the articles/ posts below for easy reference.
👉 Which experimentation metrics should you be measuring? Katie Kelly from Speero (formerly CXL) provides a primer.
🤔 Your turn -> Which metrics do you track for your experimentation program?
In case you are wondering what this is all about, read my previous post introducing The Weekly Buzz and what to expect. And now without further ado, here’s the latest installment of The Weekly Buzz. Scroll down to the video for key takeaways.
While I took a break from writing in the last few weeks, I didn’t take a break from creating experimentation focused content. In December, I began collaborating with Experiment Nation. Spearheaded by Rommil Santiago, Experiment Nation connects experimenters from around the world and is a platform for sharing their experiences and learnings.
It is a way for busy professionals to stay in touch with the latest in the experimentation space. Every week, I pick 5 important stories covering experimentation news, A/B testing stats, UX research etc. and create a 2-minute video which has bite-sized takeaways from these stories for quick…
I got my Conversion Optimization certification from CXL Institute about two weeks back. And pursuing this course has been quite a journey for me with a lot happening alongside — dealing with the layoff, supporting my daughter’s online schooling, starting this blog and starting off as a freelance consultant. So I am really reveling in the sense of achievement at least for now! It’s also that time of the year. :)
An emotion is a complex set of physiological changes in response to a perceived threat or opportunity. They’re automatic and mostly unconscious, which is why we’re never fully aware of all the changes we’re experiencing.
Emotions drive behavior or put another way motivation is an emotion that facilitates action. If you have read my previous post on Using Neuromarketing to Drive Conversions, you came across a fascinating study done by researchers at UT Austin where participants were shown photos of two chickens. …
You launch a new A/B test after weeks of planning. Your boss asks you for an update one day into the test and you peek at the results. The test variant is up heavily vs. control! So you tell everyone that the test variant is already outperforming the control. You think your intuition is right — this idea is a winner! You always knew it!
A week later, you are preparing to give an update in the same meeting. You peek at the results and the lift has vanished! What happened? …