You are here

TorCHI talk: Stories behind the links

AttachmentSize
KEI-TorCHI-StoryLink.pdf1.68 MB

On June 24th, I was honored to be part of the TorCHI program and talk about information architecture on ibm.com. The talk was entitled There is a story behind every link:

Ever wonder how a large corporate web site navigation system is designed, and how it evolves over time? How certain links find their way to the corporate home page, while other ones do not? How major changes in the corporation's business affect the web site navigation? And how the IA of the corporate web site can give clues about how well the business is doing?

If so, then join Keith Instone as he tells stories about the information architecture of ibm.com. Example stories may include:

  • The evolution of sitewide navigation categories like "solutions" and "services"
  • Worldwide information architecture challenges
  • The cultural shifts required to do something as (theoretically) simple as adding a "sign in" link to the masthead
  • Tracking the evolution of a single web page as it signals fundamental shifts in how the business is operating "behind the scenes"
  • The effect of selling off a key division of the company on the navigation system
  • Balancing the strategic goals while handling day-to-day requests for changes to the navigation
  • Techniques for dealing with executive home page link requests

The stories Keith tells will be in part determined by what you want to hear. Come prepared to select some high level links on ibm.com and see if Keith has any interesting stories to tell about them.

I prepared about 100 slides of possible things we could talk about, but, by design, the session was driven by what the audience wanted to talk about. Here is my (incomplete) list of what we covered:

  • Sign in / register, country location in the masthead today (logic for placement today, and in the future)
  • Who owns what, how to manage the millions of pages
  • Solutions, services, products and corporate strategy over the years
  • Masthead sign in challenges and compromises
  • My IBM: use as anonymous user, gateway to various applications
  • Who does the 3 prongs of user feedback: User research studies, analytics, qualitative user feedback
  • Search challenges: technology/budget, tagging, UI (easiest of the 3)
  • User-generated content (silos likely initially, integrated over time)
  • Accessibility challenges overall, why the link in the footer
  • Role based navigation (home page Learn about tab), task based navigation
  • Standards for page design ("what we offer" module as an example)
  • Content management, sharing, the "4th tab owners" of a solution page
  • Tactical, medium-term, long-term planning for changes (add 1 link now, work with groups for better tactical changes later, save some things for a big, strategic redesign)

As it happens with these things, you really had to be there to get any value out of my artifacts from the talk. For those of you who were there, you can download (below) a PDF of screen grabs from some of the parts of ibm.com that we talked about. And I included some of the text slides that I prepared ahead of time that were relevant to what we discussed. I included the "history of the ibm.com masthead categories" slide that several people have asked for.

There are lots more stories to tell: I will have to save them for some other talk.

Comments

One correction: the masthead links you show for 1995 are actually late 1996 or even 1997.

You may have seen this already but I've compiled an ibm.com homepage history site. It's more from a tech perspective than UX/IA.

In the early (1994-1998) days you got onto the homepage navigation mostly by succeeding in escalations up the IBM management chain. This was extremely infrequent though as the bar was pretty high. One additional hurdle was that I had to give a rather straightforward tech sign–off: you had to prove that you could handle the routine traffic off the site (I don't remember exact statistics, but certainly through 1997 it was a given that *every* link off the ibm.com homepage received at least 20% of the traffic that the homepage received, regardless of whether it was good, valid traffic, or 'bots, if ibm.com receive 100k views, your site would get at least 20k). The funny thing (funny now, in 10 years' hindsight) was how many sites/organizations succeeded in justifying a link off the homepage (or out of the top nav bar, which was even more infrequent) yet did not / could not come up with the resources to handle the minimum traffic coming off the site. One group which I cannot name publicly had a server which fell over dead if it received more than 10k hits (that's hits, not page views) in an hour, and had succeeded through escalations in getting a link off the home page. When I demonstrated that their server could not sustain the minimum traffic they then escalated to have their link removed from the home page (it never actually went live).

Hi Ed - thanks for the comment! I probably visited your historical page a while ago, but did not remember to reference it in detail for my TorCHI talk. My dates could be off by a year or 2, sure. I will make updates as needed based on your historical record.

One note, tho: I was focused on the sitewide navigation - what users see across the many pages of ibm.com, not just on a country home page. So my point was that back in the mid 90's, there did not appear to be a cohesive sitewide navigation system. Your history seems to back up the fact that v9 (1999) was the turning point: "The goal of v9 was to move beyond a common design system to a common information architecture." So I guess v9 is when we switch from "BC" to "AD" and I was only interested in the New Testament. (^:

A few comments on your page. We are indeed on "v16". Our "extra tab" experiment for the home page is over and we are working on making that a permanent addition. If you ever want to fill in the missing years, the Wayback machine is pretty useful (e.g., archived pages for www.ibm.com/). I used that a lot to prep for my talk, but did not actually use much of that in Toronto. Like your home page history, I have one for "IBM Solutions" that has a lot of interesting stories behind it. Maybe for my next talk.

Right, v9 (aka "Bullseye") was the first real site-wide navigation attempt. While there had always been a set of common links in the nav bar and footer, they weren't very consistently maintained (there'd been two previous attempts at site-wide navigation, with the initial launch in 1994, and the February 1995 redesign, but these were very top-level).

What we tried to pull off (and did, mostly) with v9 was to have a very consistent, tightly integrated set of links from various deep parts of the site to other deep parts of the site (for example, we had some sort of "Buy now" link on every product page which linked to a URL off www.ibm.com which in turn redirected to the appropriate real link on shop.ibm.com. This redirection setup made it easier for us to map ShopIBM's byzantine URLs to more straightforward URLs which could be hardcoded on the actual product pages. ShopIBM's URLs changed every time they updated their product database, but instead of shipping the updated URLs to every ibm.com site, they only had to send the update to www.ibm.com, which remapped everything into appropriate redirects).

I guess another way of explaining: prior to v9 if you wanted to link from one ibm.com site to another, say from Product A's page to Product B's page, where A and B were produced by different divisions (and hence on different sites, with different management systems and software infrastructures), anyway, prior to v9 you just linked to Product B's top-level page because to link any deeper was to risk link rot, especially as Lotus Notes and other CMS’s rolled out. With v9 what we tried to do was provide a central place off ibm.com with a standard URL taxonomy so that the guy maintaining the Product A page could just link to ibm.com/products/category/B (and /products/category/B/shop, /products/category/B/support, etc) and rely on the ibm.com URL database to redirect to the "right" site.

Unfortunately, after maybe ten months (I left in September '99, about seven months after v9) the whole thing started to fall apart because individual groups failed to maintain their links at the central repository and the maintenance of the over all infrastructure sort of fell apart.

From '94-99 there was a set of fixed links that every page (in theory) was to maintain, but these were all back to the corporate web site (ie, if you had a link called "Search" it had to go to ibm.com/search. You could have a site-specific search as well but it had to be labelled as such. "Products" was to go to ibm.com/products, "News" to ibm.com/news). One enforcement approach we had at our disposal was the search engine: if while doing the web crawl we detected that a given page did not have the common links, it got flagged and did not get added to the master search index. More often than not the threat of being dropped from the search engine was enough to get people to keep the common links (which were more of a design thing than an I/A thing).

Alex Wright could fill in better than I about how the shift from design to I/A occurred. While v9/1999 was the watershed event, preliminary steps had been taken in 1997 and 1998. What happened with v9 is that we got the weight of g–d (Lou Gerstner) behind our efforts, where with v7 & v8 we had to beg people to implement stuff.

Add new comment