Airbnb ran a lot of experiment on their property pages between 2019 and onwards. Some of these I managed to capture here: goodui.org/leaks/?company%5B%5D=3436 These were both big and smaller experiments (probably mostly smaller from what I can see).

Thanks for the encouragement. 👍👍

Thanks

Hey Ivan. I only analyzed the US (.com) site on one or two products which showed the direct addtocart interaction. Afterwards I noticed an alternative floating cart shelf interaction on the Canadian site. So chances are that different countries or products might have some variability. That's inevitable with such a huge company. Interestingly, the floating cart interaction still shares some of the potentially positive properties (progression, clear next CTA, and feedback or confirmation). As far as implementation on the mvideo site you have to be aware that ab test outcomes and implementation decisions are not always 1:1. Or that it's still a very fresh ab test and there might be a lag. Unfortunately Andrey just moved on to a new company and I'm not sure what the implementation decision on that ab test will be. Thanks for being sceptical and trying to verify or peer review these findings. 😃

I placed the screenshots into Adobe Illustrator. With a Intous tablet, I also like the way the brush works for quick annotating. :)

Thanks for the comment. I'd be curious to learn more about what you mean by "quicker experiments for signal". Are there some examples you can describe to paint the picture? :)

Historically, we know that the A version was used as the control in the past. Hence B was the contesting variant - which of course did not pass the success criteria and was rejected (a degree of failure). :)

@@JakubLinowski the contesting variant failed, but the title implies that the experiment/test itself failed. The experiment was a success 😉

Hi Victor. The cards serve the purpose of expressing how confident we are in each little change. Going left to right, we have: 1. A total count = sums all points from Evidence and our Subjective guess. 2. Evidence points come from counting the frequency of winning-losing a/b tests (historical data from goodui.org/fastforward) 3. Subjective confidence = we express the likelihood of something being a good/bad idea ranging from -3 (definitely bad) to +3 (definitely good), with 0 being neutral. 4. Occasionally we also reference the potential impact of a change (also pulled from historical data using medians from numerous tests) So that's how we're building confidence in what to change. Does this help?

Adobe Illustrator + Techsmith Camtasia + My Sketching Template :) = linowski.ca/sketching.php

@@JakubLinowski on what hardware? iPad or some other device?

Desktop + Intuos Tablet

Adobe Illustrator :)

CC? How did you do these lines with points at the start?

Interesting question re: testing for the timing of the coupon code choice. Thanks for sharing - would make a good experiment.

Adobe Illustrator + ISN linowski.ca/sketching.php

I see that they want to be different. Having access to previous a/b tests however I had to make a bet in favour of greater specificity over aspirational copy. Would be great to run some experiments on whether one or the other approach is truly better. I think your idea to test copy that moves away from "giving you work" is also a solid one!

Yes, that would be a good test to see.

That's a bit different focus (with obvious overlap) for GoodUI, but this may help (also about Dropbox): www.useronboard.com/how-dropbox-onboards-new-users/

Thanks for the suggestion. Can you elaborate on "context of use" and which situations you are thinking of? On goodui.org/fastforward/ you can filter patterns by page type or metric type. Anything else?

One more thing. You tried too much to sell the product. By putting so many arguments, people can think that you are forcing them, that you are too insistent. It's a bit like shop sellers, who insist on selling when you just want to watch for yourself.

Good observation. Thanks.

The effect range is between -1.2% and 26%. I've tagged this as a suggestive / possible test result.

Brutal Humor If you open the url in more browsers (or clear cookies) you might notice they are doing a follow up test.