Using Advanced Segments to Guess Amazon aStore Performance

March 9, 2010

UPDATE: As of 14th March 2010, due to copyright infringement, I’ve changed the original uk-kindle domain name in this article to a new domain: cheap-find.com/kindle-uk/

This is the third post about my ongoing ‘affiliate’ experiment. I’ve set up a very simple Amazon aStore to allow people to buy the Kindle in the UK (see what I did there? SEO in action, baby). I then advertise the URL using Google AdWords, hoping that the commission made through sales outweighs the cost of the AdWords. It’s a tried-and-tested method; nothing new (except that it’s new for me).

I’ve previously complained that – because the final purchase transaction occurs on the Amazon server – it’s impossible to track conversion/goals using Google Analytics: which visitors actually go on to make a purchase? And, more importantly, which keywords produce visitors that convert most often?

I’ve come up with a hack/workaround, that I think works. With aStore, the Amazon mini-shop is embedded at your URL using a frameset. The Google Analytics (GA) code is placed within this frameset (not within a frame, as Amazon controls the frame), so as the user navigates the shop, it doesn’t record the individual URLs. In fact, I’m not convinced that GA even registers individual page requests. But, it does seem to record time on site.

So, to ‘best guess’ users who have made a purchase, we can use the GA ‘Advanced Segments’ feature to create a segment that represents users who have reached our goal. Instead of our goal being an accurate ‘checkout success’ page, we’re going to use the time they spent on the site, as that’s one of the only things we have access to.

I created a segment that matches all visitors who spend more than 3 minutes on the site. In theory, these are users who do more than ‘bounce’, and are more likely to be those who are going through the checkout process.

Based on the number of users in this segment per-day, and the number of sales I make each day (it’s actually about 5 per week, on average), I’d say this ‘best guess’ metric is about 80-90% accurate. There generally seems to be 1 or 2 visitors who match this segment on the days that I make 1 or 2 sales, and 0 visitors on the days without sales.

I can then use this metric to report on which of my AdWord keywords create ‘converting’ customers.

As you can see, I’m probably wasting my money buying clicks for generic phrases like ‘kindle’ and ‘amazon kindle uk’, and should concentrate on the more specific (and therefore higher intent) keywords. Note that all keywords are limited to UK visitors only (there’s no way I’d bid on a generic, global ‘kindle’ keyword!).

Anyway, there you have it: Advanced Segments to the rescue.

PS Feel free to copy my keywords and site; this project isn’t about making money, but about undergoing the process itself. I hope the information I post is more valuable than the odd few quid I currently make through the scheme (otherwise I’d be blurring out all my keywords and stats above).

One Response to “ Using Advanced Segments to Guess Amazon aStore Performance ”

  1. thefluffanutta on October 12, 2010 at 3:09 pm

    There is a trick you can do with Google Analytics and Amazon aStores to track your visitors. You need to use the “No Javascript” service that has been setup, and put an image tracker in as the logo of the aStore. Works with any image based stats trackers…

Leave a Reply