Here’s a tough search problem (using Google at any rate): I’d like to find a good sample presentation in PPT (not a PDF) that was created by or for an advertising agency to showcase their wares… something that looks good, is in the PPT format. I’ve tried lots of different search combinations with positive and negative keywords in Google, but still haven’t come up with very much compelling. This shouldn’t be that difficult.

One thing like this I’ve done in the past is ask my network on LinkedIn, there are usually a couple of good examples to be had from this type of exercise, but ultimately it takes more time and you don’t get the kind of variety and volume you’d expect from a web-wide search. It’s a bit hit and miss. Why isn’t there a better social search capability for this type of stuff yet?

I was fortunate to stumble upon an example of what I’m lovingly calling “the opposite of keyword arbitrage”, or perhaps it is more appropriately described as “de-contextualizing” a search. Check out the two screenshots — a search for a cheap PC led me to the very appropriately titled Pinpoint search results (see below - click on image to expand). Wow, zero results is probably not delivering very much in the way of ROI here on your Google keyword buy guys. I hate to say it, again, but AOL needs to spend some more time figuring out shopping search, again. Even if they’re getting these clicks for free as part of the Google deal, it’s still a pretty shocking waste.

I just posted a comment on TechCrunch about traffic-counting, so now I’m waiting for my flood of traffic — yeah right! I love nothing more than having to give the “why these numbers and those numbers and their numbers don’t match” speech… well, actually I’ve found it’s much more interesting to actually figure out how to drive the consumer traffic instead of theorizing about it.

But the point of this post was another very new thought: that local search still sucks. Latest case in point was I was trying to find the details for a specific driving range near my house. What should have taken 20 seconds took 5 minutes for me to sort out, as all manner of adsense-laden sites with the same basic list information (but no interesting content, or relevant details about these places, though it does exist— somewhere) came forth from the Google-y depths. I’m hoping that more guys like Matt Berk and Openlist (recently bought by Marchex — good on ya Matt!) can help make sense of this local information mess, and that the engines can actually start doing something here. It’s not a new problem and it just seems to be getting worse. *end rant*

As usual, some good thoughts from Gary about Yahoo! and Google’s approach to bidding on trademarked terms.

Good comments from Gary about search and privacy, re: discussions at ad:tech about it. Will black-box Google make users’ collected personal information available to them to view/edit/mess with? I doubt it.

I was looking for a “list of top search engine terms”, and typed this into Google. Low and behold, I got zero organic results, but I did indeed get quite a few sponsored links for this term. I guess the buyers of searches are sometimes ahead of what people are actually searching for - but then I guess that’s partially what I was trying to avoid in coming up with this list in the first place.

One should now no longer underestimate the importance of choosing one’s words (keywords) carefully, nor even of choosing one’s tags carefully. With the ever-interconnecting Web weaving its magic, there are multiplying consequences, intended and unintended, that will present themselves chaotically. In simple terms - keywords and tags matter, a lot. As a concrete example, one of the top organic listings on Google for LinkedIn now is its Technorati tag which in turn references its flickr tagged-photos and del.icio.us tag prominently with some bizarre but interesting results! And of course not to mention that this post in pinging Technorati might affect that mix itself…

I spoke with a career consultant recently who has several clients who are using LinkedIn to assist them in their job search. She worked closely with them on their profiles, especially on their summary to help position their experience appropriately. The summaries she showed me were very well written and readable and she said that it was sometimes difficult to also get into the summary enough of the appropriate keywords without butchering it. For years now, job seekers have been told to make sure their resumes have enough keywords to be found by programs that scan resumes looking for certain attributes. With hundreds of applicants for some jobs online, it has become increasingly likely that the first pass over your resume might not be a person after all. It gives a whole new meaning to “looks good on paper”, doesn’t it? But now there’s more information out there about us, and a lot of it is under our control, at least at the outset when we post it, upload it, tag it or choose our keywords. Over time it becomes chaotic and less able to be controlled.

How long before we start seeing “personal SEO” or “tagsonomy” consultants offering to help us pop up in the first position (or more appropriately, the right position) in certain types of searches? The tagsonomy consultant would help us tag and have pinged our blog posts correctly to attract the right attention, get into the right feeds and get in front of the right client, dealmaker or employer…

Just as ecommerce and content meta networks have been helping us make better decisions for some time (e.g. comparison shopping sites aggregating, refining, redistributing traffic) we are witnessing the rise of the personal metaweb: a context layer that surrounds, optimizes, characterizes and ultimately points to people and the things they like/want or need. (BTW: Nova Spivack already coined the term MetaWeb in late 2003, and is worth repeating here. I don’t think it’s just about RSS, and it’s only really starting to become a little bit mainstream now.

Yes, I believe online display advertising is underhyped and search is overhyped. We’ll see how this shakes out, but here’s some interesting research from Dynamic Logic about how display ads encourage users to search:

The consumers who were served display ads conducted 61 percent more searches on keywords related to Harrisdirect–that is, those keywords purchased by Harrisdirect for search marketing purposes. Also, the group that was served the display ads clicked through on Harrisdirect-related terms at a rate 249 percent higher than consumers who were not exposed to the ads, and clicked on links leading to the Harrisdirect site at a rate 139 percent higher than those not served display ads.

Not extremely surprising results, but hopefully this will encourage some more work in the field to help understand further how search and display advertising (along with contextual and behavioral components) can work together to create more value for consumers.

So I stopped by Ad:Tech yesterday in downtown San Francisco. It was not only great to run into a lot of people I know (hard to walk 10 metres without doing so) but awesome to see the buzz and the sheer number of people there. Things are hopping in the online ad world once again. I didn’t make it to any of the sessions but did manage to catch up with some of them via the Ad:Tech blog. One set of data that seemed particularly interesting was some of the “heat maps” of where people look on the page when searching online. Here’s a link to some of the information presented including an example map — top three organic results all get ’seen’ by the user. Top sponsored link on Google gets 50%, then 40%, 30%, 20% from there. Nice. Think we’ll see a lot more science on this type of stuff to come… hard to believe we’re still at such an early stage in search/contextual/behavioral marketing. Very exciting.

Here’s someone who works at a comparison shopping company, funny that his name looks quite familiar to me. Some interesting posts, check out Greg Yardley’s blog site.

Damn interesting - my immediate reaction to Amazon A9’s new Yellow pages search. For 20 large cities, they’ve gone around and taken digital photos of the streets, of storefronts. And when you do a search for a particular business, it shows you the front of the store, and you can “walk” up and down the block browsing through the other storefront photographs. Apart from the interesting side light of the people who were randomly caught in front of the camera while this was going on (in some of the shots you can fairly clearly see these people) and any fall out from that, this is a cool new toy. And it is actually quite useful, since how many times have you been looking for an address and it is poorly marked or the store sign is difficult to see, and such a service would prepare you in advance.

I must say, though that is a bit freaky to see the front of your old apartment (a few years ago, in San Francisco) featured in these photos (for fun, work backwards to figure out where I used to live if you can!), but then again this is a step in the inevitable direction of cheap, ubiquitous cameras and 247 peer-to-peer surveillance.

Clay Shirky posts in reponse to Louis Rosenfeld about folksonomies - as I understand it, a metadata taxonomy that is dynamically created and maintained by the users of the network/service (as opposed to a “controlled vocabulary” which is defined by the builder/creator of the network). Shirky rightly points out that the costs of defining and maintaining such a vocabulary can make it prohibitive, and that it is easily undermined by users who jam exceptions into “other” categories. In fact, if the system is designed smartly, metadata (some user-created, some automatically created) will build on itself and the system will start to self-organize.

It often depends though, on how the system is used (e.g. search vs. browse) as search can be a way for small mistakes to be ignored. If users, for example, categorize objects in a certain way that is “wrong” (with respect to how the majority of these objects are found in the system) then their objects will not be easily found. They will tend to correct this and/or searchers will have to take this into account when searching for these objects. A simple example of this might be defining your educational institution in a business networking service like LinkedIn. The service doesn’t have a definitive list of all the educational institutions in the world, and so users can freely enter their institution and often there are several versions thereof. Smart humans search a few different ways - and over time these systems can easily “suggest” alternate searches based on observed behavior - thus allowing an automatic way to help the first-time user navigate properly.

Mentioned in Mark Naples’ Friday column this past week are the comScore numbers that we heard about earlier in the week at the Majestic Research conference (it was a really great, focused event, BTW, kudos to Seth and the Majestic team).

Here is the release on Yahoo! News. It is interesting, but not really counter-intuitive: Using a combination of online behavioral observation and consumer survey data, comScore estimated that 92 percent of all buying activity following a CE/C search occurred offline. where CE/C is consumer electronics/computers.

In addition comScore reports that “85 percent of conversions occurr[ed] in a latent (or non-search) session” with “nearly 40 percent of all purchases… “5 to 12 weeks” after the search session. This makes sense given the longer buying cycle and high degree of consideration for these categories.

Returning to the earlier point, the ratio of offline- to online-purchases for this category is not that surprising. In March 2000, while at Jupiter I built a model of offline-online spending where we projected the % of commerce that would be started online (research, referral etc.) but completed offline - the ratio of offline (influenced) retail in the US to actual online sales for 2004 that we projected was about 78%. Given the 90-day timeframe of the study I’d say this gels quite well with what we at Jupiter predicted several years ago.

The generic vs. specific search term issue is a quite complex one, and comScore’s data here is interesting, but I would really want to actually understand the methodology and data in much greater detail before commenting. This is especially the case considering the 90-day timeframe, and that this is a survey combined with analysis of search engine results. As someone who has frequently conducted research studies and just released “teasers” to the press, I understand and sympathize with the practice, but for a study that is sponsored by someone (Overture-Yahoo!) who clearly stands to gain from the results I hope and ask that they make this study and all the data that feeds into it available for public scrutiny.

Finally, to my earlier post about navigation-searches - I don’t think this conclusion in the comScore release makes much sense: comScore also found that generic search terms are likely to have influenced even those consumers who converted to purchase after conducting a retailer trademark search (e.g. “Best Buy” or “Gateway”). Fully 84 percent of these buyers searched using a generic term earlier in the buying cycle, reinforcing the importance of reaching consumers early in the search process when they are defining their consideration set. This seems very client-serving, and I would say if you’re doing a “retailer trademark” search later in the cycle it is quite likely you are looking for a physical store to make the product purchase and that those earlier searches (probably for product and not retailer research, quite some time ago) have very little latent influence left.

This is a minefield and I applaud comScore’s efforts to study the important search-commerce link in greater depth and I hope (but we will see if) they and others who follow will be transparent with the data and results so we can truly find answers to help us all better spend our marketing dollars online (and off).

More hard data showing how consumers are using search engines as navigational tools at least as much as they are doing true searches, with this Hitwise data quoted by MediaPost.

86.7 percent of the top 500 unique search terms for the week ending Dec. 4 were related to corporate brands such as eBay and Wal-Mart. Nearly 11 percent of the top search terms related to generic products

It is quite amazing - Google’s search engine (and others) has become the defacto browser address bar, with people typing in things like ‘ebay.com’ or even ‘www.ebay.com’ in the browser instead of just typing it in the address bar. I will try to dig up some other interesting snippets and post later.

Nice. Microsoft launches their new search site and already it’s “temporarily unavailable”. Way to go guys.

A few weeks ago, I downloaded the newest version of the Amazon a9 toolbar now that Amazon is pushing it and saying it is fully ready. Naturally it had the almost-obligatory-for-a-toolbar popunder-blocker. Unfortunately I noticed that when I clicked on a lot of advertisements on the web, it was blocking these! I got in touch with a couple of my Amazon contacts and they in turn informed the a9 team who acknowledged the problem:
we’ve known that this can be a problem for some clicks (especially those that aren’t launching the window directly, but are initiating _javascript that in turn initiates a new window. we’ll add it to our list of things to look at near-term.

I yesterday noticed that the popup blocking functionality had disappeared from the a9 toolbar… at least from mine! The issue of how to actually block popunders without inhibiting other functionality (e.g. a lot of sites use javascript to popup dialog box-type windows) is already a big one. Now with XP SP2 we have “duelling blockers”. Even still, I went to Weather Underground yesterday and saw a SpecificPop popUP defeat the XP blocker and paste itself OVER the site’s content, a bit like a roadblock but it was a typical pop-under creative! Things are going to be ugly in this space for a while to come.

I am currently running a microtising campaign on Google, where I’m buying Gary Stein as a keyword to drive traffic to this site. Microtising - using a large scale advertising medium to reach a very specific audience, much smaller than a niche, I’m talking one or two people. Anyway I had lunch with Gary today and he was somewhat amused. He should be happy though - I don’t get much search engine love anymore - while “gary stein” had 37 Overture searches in July 2004. So I went ahead and paid for lunch

I did a bit of research using hitwise data to compare the big search engines with shopping-specific search/shopping sites (DealTime, NexTag, PriceGrabber, Yahoo! Shopping etc.) so I figured I’d post the topline results here. If you’re interested in the full results and the paper I wrote, let me know via email (rob at analystblog.com). Here’s the summary:

Upstream traffic (as defined by hitwise - site people were at right before) for the top 10 online retailers that was from Google or Yahoo! Search increased 27 percent from June 2003 to February 2004. During this same time period, Google and Yahoo! Search increased their marketshare of overall US visits by 15 percent, thus they gained net share of traffic leading to these top retailers.

However, over the same time period comparison shopping sites? share of upstream traffic to the same top 10 retailers grew by 76 percent – taken as a group, comparison shopping sites grew their marketshare of aggregate visits by an amount similar to Yahoo! Search and Google: 14 percent from June 2003 to February 2004.

It thus seems that shopping search sites are humming (fair disclaimer here: I *do* work for one of them! Now you know why…). Send me your comments.

This entire blog is much more of an experiment (mostly a creative outlet) than anything else, so I thought I might as well as some Google AdSense to the bottom right. Initially it has ads for non-profits etc. unitl they actually scrape the page to figure out what the content thereof is. I’ll resist the temptation to select content I think will sell (geez!). Now I just have to get real people to come to this site…