Friday, January 09, 2004
The Semi-Organic Navigation
A few months ago I was looking at a set of reports by one of my colleages that was generated for one of our larger clients. Amongst them was a report about some statistics relating to that client's on-site search engine. There was the usual stuff about what the top search terms were and why we thought new ones were showing up and other, previously popular ones were dropping off. Usefull information in itself.
What I hadn't really paid that much attention to in the past, but struck me as extremely important this time, was when he did a report on what the most popular search terms were by page on the site.
This report provided valuable information about what people were looking for on what pages in the site. And, as the site was sectioned off according to audience segment (business vs. personal), we had an instant view of what different visitor types were looking for on the site, where on the site they were looking for it, and how they talked about what they wanted.
This report lead to the creation of a couple of pages for content that was in high demand, and the addition of new search terms in the meta tags on other pages where it was shown that people were describing the content of those pages in different words.
I hadn't though much on an expansion of this theme in the intervening months, except to be bullish on on-site search engines as a valuable way of optimizing a site. And then I was doing an analytics plan for the site of a different client - one to whom I was going to recommend that they set up an internal search engine even though they have a pretty small number of pages on their site - when I got to thinking:
What if you were to eliminate all other kinds of navigation except the search engine?
The rest of this entry will examine that scenario and see how it plays out.
Search-only as Navigation
So, what if you just had a Search box, and no other kinds of navigation, what would that mean?
Well, first, it would mean that you'd have to tell your visitors what to do. You'd probably want to make the Search box very visible, and have a very short line of text that indicated that they should just type in what they are looking for. The text should also indicate that they should enter their query just like they would in a normal search engine. This latter point serves a couple of functions: a) People are generally more concise when they are using a search engine than they might otherwise be; and, b)It eliminates the bothersome need to try and parse natural language queries (though this would be an obvious extension of the idea in the future).
So the text might look something like this: "To navigate this site, enter what you are looking for in the Search box like you would in a search engine."
Ok, now if we're going to treat the Search box as navigation, then instead of returning a list of results, we need to return an actual page (no sense in making the number of steps to get to a page any greater than necessary, and I've rarely seen search results that most people would describe as quick and easy to read). It makes sense that we'd want to return the top match (we'll talk about how we determine that more in a bit), but given the organic nature of both language and you're visitor's relationship to your content, we can't be sure that that's the document they wanted, so we'll want to have some way of giving them the next top 5, say, results.
I should take a quick moment to point out that this problem (not getting the content you were looking for) is a problem in standard menu-based navigation as well, but since we're introducing an entirely different navigation process, and people approach search engines with less of an expectation that they'll get what we want, I think it's valuable to do a little handholding.
Now, if they typed in some query, and they didn't get the page they wanted, odds are good that they will modify the query and try again (especially if their options are limited to either that or one of the alternate results returned). My thinking is that we need to capture this behavior as, in later analysis, it will give us much more information about the specific kind of thing people were looking for, lessening any ambiguities that might have cropped up by just looking at the first search term they entered (indeed, this ambiguitiy might be why they didn't get the results they were looking for the first time). So I'm proposing a checkbox right next to the Search box that says something like "Still looking".
This will need a bit of explanation itself, but now that we know they've used the Search box at least once (and hopefully get the idea), we can replace the help text we had that explained how to navigate, with text that explains the new check box. Something like: "Check the 'Still Looking' check box if you didn't find what you are looking for and are trying again."
Leaving the land of the user interface for just a moment, let's look at what we can do with and learn from this on the back end.
A couple of obvious things are the things that I mentioned at the beginning of this entry. We can learn what things people are looking for on the site.
- When we know what the most popular sorts of things people are looking for, we can add new pages to the site that cover the most popular ones that aren't already covered.
- For popular search terms that match content already on the site, but not referred to the way visitors my be referring to them, new keywords can be added to those pages so that they are found more readily.
- Best of all, we know, for any given page, what people are looking for next. We get an understanding of how the major segments of visitors want to get their information. And these segments are self-selected, meaning they most accurately represent the differing groups of visitors to your site (determining what people in the same group have in common, and how they differ from other groups is no trivial problem, I'm guessing, as it may have as much to do with how they gather information as it does why they are gathering it, but that's fodder for another entry).
Let's look deeper at this last element for a minute. If we know the most popular places that people want to go from any page, it seems to me to make sense to go ahead and show them these places so that they don't have to click to get there. Let's call this the "Top 5 Next Pages". This isn't altogether different from something you might do with straight path analysis and a regular menuing structure, with the important difference that every path is an option (not just the ones where the existing menu leads), and that Search doesn't become a black hole where you have to do separate analysis, but is rather the focus of the analysis with all of the information in one location. If this list was built dynamically with, say, the information collected over the last 3 months (or some timeframe that makes sense for your site), then this navigation component changes organically as your customers change (due to changing business climates, or simply where you are targetting your ads), or as their needs change (as their own business landscape morphs).
Ok, back to the user experience for a minute (I know, we were really there in that last paragraph, just bear with me). One thing that seems missing from this equation is a method for understanding how well the page the visitor landed on suited their needs. Now I'll admit to not being totally convinced that this step is necessary, but let's walk through it and see where we get.
Suppose that, since we aren't providing traditional navigation, we want to add the ability for visitors to rate the page they arrive at on some scale from "Exactly what I wanted" to "Not at all what I wanted". Does this provide us with any useful information from the point of view of Search-only navigation? Certainly it would tell us if a page where people abandoned significantly was a page they found useful, and that they left satisfied, or not useful, and left out of frustration. We could get this same impression, however, by just studying the abandonment rates on any given page and just assuming that high abandonment means frustration (or bring our knowledge of context of the page to help qualify this opinion). So it is of some import, but not terribly helpful.
We could say that such a rating would help us determine more generally how well a page matched a search term, but I feel like we could infer this by assuming that the last page they visited in a search process is probably the one that most likely matched their expectations. Now if they went from one search process to looking for something entirely different, I guess that page ratings would tell us if their search ended well, or they moved on to other things out of frustration.
We could also say that the page ratings could help us sort search results, all other things being equal. This seems like it has more potential, and there's probably something to be had by combining these last two pages. I'm always a big fan of soliciting user feedback where ever possible, so I wouldn't dismiss the idea out of hand.
Something else we now know is not just what people who are on a particular page are likely going to want to know about, but what people who have taken a particular path to get to a page may want to know that's different from what people who came to the page via a different route might want to know. An example might be that certain people want to know your pricing model up front. Others might want to know about your technical infrastructure. If they both then go to the product features list, the first group of people may want to go to the technical infrastructure page next, whereas the second group may want to go to the pricing page. A process that took into account just the page the visitor was last on, and the page that they arrived at next, and then generated the list of likely pages to go to from there would probably add significantly towards customizing the user experience in a way that wouldn't be too taxing to develop.
Further insight would come from the "Still Looking" checkbox. Similar to looking at the last page and the current page to generate a list of next pages, looking at the last term a user entered, and the current term (with the "Still Looking" box checked) is likely to add much needed context to what the user is looking for without having to go back to the first term and follow it through (if they checked the "Still Looking" box repeatedly). This would inform the search results to more effectively serve up the desired page.
Combining these last two ideas, knowing where a person is, and the search term they entered would inform the page that the visitor should be taken to. For example, if they typed "contact info" into the Search box on a support page, it probably means something different than if they typed it at an investor page.
But sometimes visitors don't know what they need. And this is just one of many valid reasons to provide links in your body content. When it naturally flows that visitors might want to visit another page from the current one, perhaps because one provides context for another, there's no reason to stifle this natural flow, especially when it gets visitors heading towards a conversion.
So, now we have a different breed of navigation set up. A Search box as primary navigation. Some means of indicating other results for the same search, and some means of indicating the most probable next page, the latter two largely self maintaining. This means that as you add new content to your site, it will fall into its natural place organically. You also have a site that is automatically optimized for external search engines with the content that visitors want referenced it the language they want to use to find it.
Some quick reality checks:
- This type of navigation would be a BIG risk for anyone who actually relied on their web site for a living. That said, it would be an interesting way to get "innovation credit" from your audience if you're in a market where that would count.
- I've glossed over what would be a lot of work setting this kind of thing up to be self-maintaining
- There would be a LOT of discovery involved in figuring out what is worth paying attention to to make this work well.
But ultimately, I think the payoff in the ease of navigation (after a while, admitedly) and the amount of stuff you could learn about your visitors would be tremendous.
Thanks for tagging along!