Bryan Eisenberg Explains Amazon's Relentless Customer-Focused Optimization [INFODOODLE]
We have all done it.
We have a purchase in mind. We drive to the store, any store, and focus all of our attention on completing this mission successfully.
We wander aimlessly around crowded aisles and end caps stuffed with closeouts and uncategorized “junk”. Giving up the scavenger hunt, we now look for a customer service attendant to help us out.
We wait. We continue to wait until we catch the eye of a reluctant employee. We ask for help and low and behold, they are out of stock.
What do we do now? We order it on Amazon.
You still might be asking yourself, if only 10% of purchases are made online, how is it that Amazon is kicking Walmart’s butt?
- Highest Customer Service Ratings
Jeff Bezos says
“We are not in the business of selling books, but of helping our customers buy books.”“We are not in the business of selling books, but of helping our customers buy books.”
- 2/3 of Amazon visitors are return visitors.
- Amazon can tell an email was unsuccessful as soon as it goes out.
- Amazon changes their pricing 2 million times per day – no one can compete with that.
Brian Massey attended Conversion Conference San Francisco and captured this infodoodle of Bryan’s Keynote Presentation: How Amazon Uses Relentless Customer-Focused Optimization to Crush Competitors.
Share this Image On Your Site
Please include attribution to Conversion Sciences with this graphic.
“Amazon can tell an email was unsuccessful as soon as it goes out.”
Might be a language barrier issue or something like that but I can’t wrap my head around this sentence… “unsuccessful” how ? And how could they know ? I must be missing something, sounds more like magic or mind-control than some technical superiority or better infrastructure…
The implication here is that Amazon can tell from early returns if an email will be successful. They learn and innovate very quickly. Unsuccessful will most likely mean a low sales performance.
And, that they don’t send the email out to their entire audience all at once. First, targeted at that segment that who is predicted to respond at higher rates and then choosing a sub-sample of that (likely: randomly) to see which of multiple versions of the email does well with that segment, and then letting the winner go to the progressively larger proportion of the segment. It’s almost certainly a combination of multi-arm bandit (epsilon greedy) and Bayesian AB stats
Thanks for providing more detail. What tools are available to the rest of us that offer multi-arm bandit and Bayesian AB testing?
I ask you to back up the claim, that Amazon spend $6 billion on testing. That’s what your infographics says. That’s not pocket change. 6 billion is a lot of money. Where did you get that statistic? That’s a fair question.
Joe, Bryan mentioned that the $6 billion claim includes acquisitions of companies that provide their technology. It’s really an investment in innovation, testing and test capability.