Some interface designers believe that the seven plus or minus two analysis has been taken to extremes, or that it is "wrong." Introduced as Myth #23, "Choices should always be limited to 7+/-2," Zoltán Gócza wrote:
Limiting the number of menu tabs or the number of items in a dropdown list to the George Miller’s magic number 7 is a false constraint. Miller’s original theory argues that people can keep no more than 7 (plus or minus 2) items in their short-term memory. On a webpage, however, the information is visually present, people don’t have to memorize anything and therefore can easily manage broader choices.
It's unclear whether the Magical Number principle affected software design to a great extent in the first place, or even that developers and site architects have any patience for empirically based advice on interface design. Let's assume that it has had some impact. Miller, an experimentalist, would have readily agreed that humans engage in multiple strategies for visual information processing. For a behavior as seemingly straightforward as reading, it is by no means clear how all the techniques involved in comprehension, scanning, skimming, diagram interpretation and more interact. Add to this web browser habits, conventions about web site design, and the huge presence of search.
Yet each of these strategies can be impacted by memory constraints. Gócza extols the virtues of Amazon's site as somehow following a different design tactic that ignores memory constraints. Let's look more closely at what Amazon's designers have done.
A frequent user of the site and author of 100 product reviews, for years I found that it was impossible to find the landing pages for recommendations, subscribe-and-save, the media library or even one's profile. If more than a week or two passed between tasks, I could not remember how to find the particular page for that task. That is a longer-term memory failure, but it may be exacerbated by the number of unpredictable clicks required to reach a destination.
For some tasks, including those mentioned, it's gotten better, but some pages, even though they appear in multiple menus, were not retrieved with search results, which filtered only products. Today my Amazon "Account" menu, which is one of three home page menu tab groups, contains 12 options. None of these selections under "Account" is gift cards, even though one of these was purchased a week ago. That appears in a different menu group. Why? Over the past decade, Amazon has moved the gift card selection from menu to menu, and for a time it did not turn up in search results, since it wasn't treated as merchandise.
|Tab Group 2|
|Tab Group 2|
|Tab Group 1|
This breakdown shows that there are only four menu groups on the top of the home page (there's more in the footer), and that these do not surpass the Magical Number by much -- none using more that 13 menu items, and most much less.
What about usability? I have not looked for research on this, but suspect that Amazon visitors who occasionally return to the site to perform specific tasks are sometimes not cued at all by the menu system. For instance, while Amazon has considerable information about how the site is used by specific customers, the menu system does not leverage it. "Mark's Amazon.com" tailors recommendations (and not very well at that, but that's a story for knowlengr.com) but little else. It's plausible that Amazon's designers wanted me to see that there was an "Amazon Mom" link in the 65-item "Your Account" page, but after seeing it once was enough. Beyond that it, and many of the other menu items are just clutter that interfered with the task at hand.
Why does this matter? Because it introduces a memory burden for navigating the site. One is forced to use menu cues, search and scanning to navigate the site because it's not possible to remember much about the site's organization.
Consider also that every ecommerce site uses "My Account," so in one sense that site cue is delivered by metaknowledge about how such sites are to be navigated. There's no need for a memory constraint there.
Still, even that page is grouped into 5 groups, which works out to around 13 menu options per group on average. Allowing for some subgrouping of similar items, this isn't vastly different from the Magic Number's upper limit.
So what about memory constraints?
Except for the Departments listing, the home page menus tend to have fewer than a dozen items, and there are (chunked?) groupings within those lists, such as Cloud Player and Cloud Drive, which are together (are users clear on the difference anyway?). In that sense, the home page honors the constraints, albeit with three separate and somewhat overlapping menu systems.
All this is opinion and anecdote to some extent. Amazon.com obeys some Magical Number rules of thumb, at times and ignores it in other facets of site design.
There are many cognitive factors at play. For instance, while presenting evidence for menu breadth vs. depth, a Human Factors newsletter post cited by Gócza also concluded that:
Several factors are thought to influence users' success in learning and traversing information hierarchies. These do include the breadth/depth of hierarchy, but additional critical factors are: the transparency of the category and sub-category labels, qualities of information scent, relative size of categories, and the shape of the hierarchy.
Especially problematic is Amazon's help system, which, like many sites, seems to concede the game to search. That too is a discussion for another day and another blog.