Loading ...« Kelsey ILM Coverage: Local PPC Targeting | Home | Kelsey ILM Coverage: We have money to spend; help us spend it! »
Kelsey ILM Coverage: The Cambrian Age of Search
By Matthew Berk | November 29, 2007
Today finds me in Los Angeles, at Kelsey’s ILM’07 event, listening to a session titled “Convergence of Local Media and Directories”. The convergence, it turns out, is all about how consumers search for local services. The panel’s doing a great job hashing through what it might mean to have effective local search, and there’s disagreement aplenty. Is local search about the long tail or the short? Do searchers refine, or is the first search the only search? Are consumers willing to keyword search, or do they prefer to navigate? There’s even a striking moment, when Malcolm Lewis, from Local.com, claims that given the prevalence of basic name/category search behavior in local, “getting it right” just isn’t that hard. Emad Fanouse, of YellowBot, who has been in the search trenches for many years, including a hands-on stint at Citysearch, vehemently disagrees.
Amid the back and forth, one thing’s clear. The standard for comparison is always Google, and most often the basis is the survey of one (”I was recently looking for an x on Google, and I saw y and z”). In one stunning moment, one of the panelists talked about Google’s related searches, and how they let him dig deeper into the result set (questionable). After a few comparisons, Robyn Rose chimed in with a great point: that consumers are forgiving when it comes to Google, and even if the results are dead off, consumers are wont to fault themselves for not guessing the right keywords.
What’s increasingly clear, is the brilliance of the risk profile of the limited Google search metaphor. At first blush, the box-and-list metaphor is the riskiest interface imaginable, because you basically have only a single (type of) input, a single interaction, and a limited set of results to present. (Note that universal search only stresses the box-and-list metaphor.) But over time, under the ever-steadier aegis of the Google brand, the limited search metaphor is in fact the greatest hedge against risk; with little experimentation on metaphor (even Google “experimental” doesn’t significantly complicate the box-and-list style) and consumers willing to fault themselves, everyone else in the search game is forced to run one-off experiments.
Cases in point: all the products/sites represented by the panelists (ZipLocal, YellowBot, Local.com, and Idearc) vary considerably from the box-and-list standard; they offer related searches, refinements, sorting, tag search, and a bevy of other functions that, when compared to the rigid simplicity of a Google, are difficult for consumers to even parse. Elaine Kunda testifies to the risks in operating your own style of search, saying that consumers are only willing to refine their searches once you hit them over the head, promote the feature front and center, and get them to understand the nature of the functionality.
We are all swimming against the tide here, and every one of our search interfaces is at best highly experimental when compared with the Google standard. To our own credit, and our own risk.
Another example. Recently, we threw ourselves into a tizzy over the fact that on Open List because a search for “finance” included results for orthodontists and other unlikely businesses which, under certain refinement and sort conditions, were listed front and center. At Google, by contrast, even those those same questionably relevant results are present in the result set, but the lack of sorting and filtering features, and the primacy of rank as a governing principle (relevance = rank) guarantee that those listings will never be on the first page.
We’re in the Cambrian age of search. The shark already swims the seas, an efficient machine that will stand the test of time. Our one-off antennae, fins, eyes and appendages, great as they may be, need to adapt to survive. My question, as I sit here and listen to folks sweat through understanding whether their search metaphors are the right ones, is simple, and won’t have an answer for quite some time: when will we see the evolution of the dolphin?
Topics: The Kelsey Group, Local Search |
November 29th, 2007 at 9:39 pm
[…] to my employer’s Local Point blog (where I post not often enough), but you simply must read Matthew Berk’s thoughts on the Google model of local search (which is crude, offering a search box and a set of results) […]
November 30th, 2007 at 9:06 pm
Hi Matthew,
To clarify, Emad’s comment was that providing relevant results to the user’s “initial query” in a local search engine is hard. I respectfully disagree. In my experience, most consumers using a local search engine or IYP search on short, common and easily resolved search terms like hotels, plumbers, realtors, etc. In fact, depending on the UI, most consumers don’t even type the search, instead clicking on a “popular search” link. Getting it right in those cases is idiot proof, which is why the links are provided in the first place.
You mentioned that “In one stunning moment, one of the panelists talked about Google’s related searches, and how they let him dig deeper into the result set (questionable).”
That was me too. What I actually said was that Google now allows you to refine some searches by suggesting keywords that you can use to refine/filter your results. Try searching for “shin splints” in Google, which is the example I gave on the panel. You will see that Google provides a special “Refine results” section with clickable links for “symptoms”, “diagnosis” and “treatment”.
It’s very similar to the excellent results filtering interface you provide on Open List.
IMO, it’s only a matter of time before Google provides this kind of refinement/filtering on extended local search modifiers (kosher, kid-friendly, handicap access, etc.) in their local search results.
I’m a big fan of capturing and exploiting extended search modifers in local business profiles. As we discussed at the airport, I believe that soon we will see filtering along another dimension, ie availability. Eg, I imagine being new in town looking for a local dentist because I just chipped a tooth. I need someone that takes my insurance, is well rated, and, most importantly, is available this afternoon!
December 2nd, 2007 at 7:33 pm
[…] Marchex LocalPoint […]
December 3rd, 2007 at 12:24 am
Matthew,
Until gas stations that sell flowers and have a mailbox stop showing up as Florists and Post Offices in the data we’re still basically dragging our knuckles.
Great seeing you at the show.
Andrew
December 3rd, 2007 at 10:09 am
Malcolm,
Thanks for your comments. Great meeting you at the event. Just a few re-directs:
1. Having built a local search engine from scratch, and having evolved it over several years by looking at user behavior, I’m still on Emad’s side here. Local search, even for an initial name/category search, is very difficult to get right, especially when you let folks work with the result set. Even allowing someone to keyword search is a pretty risk enterprise, given the state of the data and the polysemic nature of the vocabulary.
2. The “refine results” feature at Google is really about helping a user refine their _query_, as opposed to the result set. In fact, many of the refinements actually yield much larger result sets. There’s a really important distinction between refining a result set and refining a query. In 2003, when Open List was just getting started, everyone thought that Google would eventually adopt result set refinement, but five years later, they’re still more interested in helping folks generate more queries (it’s a philosophical position, very worth talking about here some other time).
3. Refining by availability, as you suggest, would be a fantastic addition to any local search experience. It’s hard to get right, but you’re dead on when you talk about being able to refine by a dimension that can reflect the immediate need of a searcher. We’re going to think hard about that here….
Thanks again for the clarification!
Best,
Matthew
December 5th, 2007 at 12:42 pm
Hi Malcolm. I, too, respect your opinion and agree that doing “search” is easy…but I believe doing “search” *correctly* is far more complicated. I’ve done algorithmic (web) and local search for years. Local adds an adds more dimensions such as a name or geospatial limitations, despite having far fewer records. If someone searches for “cafe,” are they searching for cafes in general, a location that they remember has “cafe” in the name, etc…then you must also factor distance into the relevance/order of the results (people are more/less willing to travel further for restaurants versus industrial widgets). Many other companies that need local search (i.e. newspapers) don’t even understand or have the team to build these tools so they leverage data and technology (search engines, etc) from other companies such as Planet Discover and others.