I mentioned in an earlier post (Amazon as a Benchmark) how much I both love Amazon and yet find it frustrating because of poor Findability .

I experienced poor Findability (and therefore this new term “web stress”) last week on Amazon as I tried to look for a desktop computer for my colleague Dan Barbata. Dan had been looking forward excitedly to a new computer he’d ordered from Dell about 3 weeks ago but for some mysterious reason Dell cancelled the order – something to do with a parts shortage. Some instant yet painless gratification seemed appropriate and as I’m an Amazon Prime member I thought I’d see what we could get by the end of the week.

Dan’s key criteria are 8GB RAM and a 64-bit CPU to make use of the RAM so he can have lots and lots of applications open at the same time without bringing his computer to a crawl (a problem with 32-bit computers because they couldn’t use more than 4 GB RAM).

First Search Attempt on Amazon

I go to Amazon, search for “Desktop Computer” and quickly navigate to the Desktop Category.

Amazon has lots of filters (a.k.a. facets or attributes) to refine my search:

• Operating System
• Included RAM
• Condition (New, Refurbished, etc.)
• Shipping Option
• Brand
• CPU Speed
• Hard Disk Size
• Average Customer Review
• Green?
• Price
• Seller

RAM is my first criterion but the filter choices provided are useless – the maximum range is 1 to 249 GB which is going to produce one thousand one hundred and forty-nine (1,149) results! Not exactly useful plus how many people nowadays are going to buy a Desktop with less than 1 GB of RAM?

I think it must be 20 years since I had a computer with less than 40 MB of RAM so either Amazon is selling an antique in the above 1-39 MB range or there’s a product data error – of course it’s a product data error:

Note this is a product that Amazon is selling itself and not one sold by an independent merchant through Amazon marketplace where you might think that data quality would be worse.

Second Search Attempt on Amazon

Ok, so now I’m forced to go back up to the search box and try again, this time searching for “8 GB RAM” while staying in the Desktop category. I filter the results by Windows 7 Home Premium and Hewlett-Packard (let’s give them a try after Dell) and Amazon returns 5 results, all of which are sold by Amazon MarketPlace sellers. Notice the reason these show up in the search – MarketPlace sellers put a phenomenal amount of product information into the title - apparently Amazon does not search Technical Details in the full product description which in turn explains why no computers sold by Amazon show up.

Third Search Attempt on Amazon

I don’t give up but only because I’m already thinking of writing this post. I cancel my 8GB search, filter by Hewlett-Packard and Windows 7 Home Premium and click on the products to get to the product description page, scroll down to Technical Details (what a pain) to find the RAM. After 4 or 5 tries, I eventually find a Hewlett-Packard computer sold by Amazon that could be delivered by week’s end. But I’m not going to suggest buying from Amazon to Dan because there’s no way we can compare the features of all the computers they stock that have the basic criteria (8 GB RAM, 64-bit CPU) – we simply can’t find them.

Search & Navigation and Product Data - Why doesn't Amazon do a better job?

This incident is not a one off - I frequently experience problems with Amazon's Search & Navigation and Product Data and it mystifies me why this should be the case with the largest and most successful Internet retailer. In this case they clearly have RAM as a separate attribute in a structured field – albeit with some data errors – but the ranges are messed up. In other cases I’ve seen key criteria “just sitting there” in semi-structured form (i.e. contained in text in a bullet point under Technical Details) that are not even addressed in product filters (a.k.a. faceted navigation). What’s particularly intriguing (frustrating from a consumer perspective) is why terms in the search box are not being run against structured values. For example if RAM is a separate attribute in a structured field, this presumably means that the value “8 GB” is associated with the HP Pavilion P6320F computer that I eventually found. In which case why did this computer not show up when I searched for “8 GB RAM?” And why is Amazon missing the key attribute of “64-bit?” Admittedly it’s only been a year or two that such computers have been available but they’re pretty common now and 64-bit is a mandatory requirement now for anyone who’s in the market for a high performance computer with lots of memory. Even if “64-bit” is not part of any structured data provided by the manufacturer, it’s a trivial text mining job to pull it out.

Perhaps I’m being overly tough on Amazon. When we see Findabilty problems on small to mid-size retailer’s sites, we tell them about them and only write about what we’ve seen in a disguised form. But I think Amazon is fair game because they really ought to be top of the heap. The really big driver of this post though is that Amazon is my primary online store and they’re giving me lots of unnecessary web stress.

Customers get stressed if you make them work harder than they should have to on your website and your performance metrics such as bounce rate and conversion rate will suffer.

 CA (formerly Computer Associates) recently commissioned a study Web Stress - A Wake Up Call for European Business in which 13 volunteers wore skull caps to track their brain waves as they tried to search for and purchase a laptop PC and travel insurance. The study was conducted in Scotland by British usability experts Foviance. See some highlights here:

Brain wave analysis indicated that shoppers had to concentrate up to 50% more when using poorly performing websites, leading to greater agitation and stress.

It’s a given that when customers get frustrated they give up or go elsewhere – CA’s own (2009) estimates are that poor performance leads 40% of people to go to a competitor’s site and another 37% to give up entirely. What’s fantastic and exciting about the notion of tracking web stress is that this type of research gets us on the road to being able to measure usability; i.e. as well as experimenting with usability changes and their impact on conversion rates we’ll be able to:

1. Put an actual number on a website’s usability using an index or score.
2. Measure the impact on this Usability Index of making specific changes (e.g. improving Findability , changing design, streamlining checkout process).
3. Most excitingly, to create industry wide metrics that indicate the relationship between the usability index and performance metrics such as conversion rate. These metrics can then provide rule-of-thumb estimates as to how much an ecommerce site’s conversion rate is being held down by poor usability and what kind of improvement could be achieved if the site could move into the usability sweet spot.
   • My own guess is that the relationship between usability and website performance is something like an S-curve. Wouldn’t it be great if a retailer could find out they were at position A and by moving to position B they’d double their conversion rate?

   

   • This isn’t the whole story to conversion rate of course (other factors come into play) but adding this kind of knowledge to industry wisdom would help us all in a big way.

Amazon has emerged as the 800–lb gorilla of E-Commerce and some pundits believe that smaller retailers will get squashed in the expected slugfest between Amazon and Walmart. Based on my own experience with Amazon, however, I’m not at all sure.

I LOVE Amazon and buy everything I can there. Often, as is increasingly the case with many shoppers, I’ll go straight to Amazon and search for what I want before – or without -trying anywhere else. I’m an Amazon Prime member and 2–day free shipping is fabulous – especially when shipping from another retailer costs an arm and a leg. On average I buy something from Amazon once per week – last year I spent a little over $3000.

On the other hand, Amazon constantly drives me nuts because their Findability in some areas – particularly electronics and computers is annoyingly poor. It’s OK if you know exactly what you want but if you need to look at and understand alternatives, choosing the right product on Amazon can be a real pain because the attributes you need to narrow your selection are just not there.

Perhaps Amazon’s failure to provide attributes of significant depth in many categories is due to product data problems – like Comparative Search Engines, a lot of data is provided by independent merchants. Whatever the reason, Amazon loses a fair chunk of my business – e.g. today I bought a set of toner cartridges for a color printer ($200) on another website (one I have never used before) because I wasn’t able to understand – or trust – the choices presented to me – and I wasn’t sure I had isolated all the relevant choices.

I’ll be posting more examples of Amazon’s pluses and minuses but, for now, the key takeaway is that while Amazon is definitely the 800–lb gorilla there’s plenty of room for savvy merchants to hold their own.

Our shift to Drupal is a bit painful right now. Posting to the blog is easy enough (famous last words!) but getting pages loaded is tiresome. We hired someone (thanks Pawan!) to get our HomePage working but had expected that we could take care of basic pages ourselves. Looks like we’re going to need some help for a while, especially as IRWD is in less than two weeks. Keeping our fingers crossed!

This is partly a test posting to see whether blogging is working properly on our new Drupal site – we have great hopes for Drupal, particularly when Drupal 7 is released as it’s reputedly nearly 100% focused on making Drupal easy for non-techies. This post is also to announce that we’ll be exhibiting at the Internet Retailing Web Design & Usability (IRWD) conference in Orlando February 15-17th. More on IRWD to come.