Google Instant Insights

Posted on Thursday, September 9th, 2010 by Print This Post Print This Post

Categories - Featured, News, SEM, SEO

Google unveiled a new search feature, Google Instant, on Wednesday, which promises to provide users with an easier, faster and predictive way to search. Some pundits are calling this the most significant change to the Google search page since the advent of ads.  Google Instant anticipates what you are searching for, similarly to Google’s suggestions, and provides search results as you type.  For a great overview of the ins and outs of Google Instant, we highly recommend Matt McGee’s Complete User's Guide. Any new feature from Google is sure to get everyone's attention and generate a lot of buzz.  But what impact will this actually have for both advertisers and users?  We asked our Search Agents to weigh in on what Google Instant really means for SEO, SEM, and our search behavior: SEO – Potentially a boon for big brands and long-tail queries. Daniel Tsui: I would think that those searchers looking for something very specific but who normally would just enter 1 word pre-Instant would now at least tend to click on/hit enter for what may match more closely their intent. Thus instead of a 1-term search it would easily become a 3+ term search.  And it will make it “more convenient” and induce them to add terms on the fly and  to be much more specific in their search than they normally would be  have been otherwise. 3 initial thoughts on impact: 1.       Keyword demand numbers will drop for short tail, general terms as a % of total searches (whose volume will multiply many fold) 2.       Long tail optimization becomes even more important as volume will increase 3.       Descriptions of topics that grab a searcher’s attention but whose topic was not the original intended one will get more traffic vs. the old system Richard Wong: I think that if anything this Google Instant feature will give SEO and SEM editors an easy way to do more keyword research. You can type in “mortgage” for example and see all of the auto-keyword fills Google provides for you as you type in each letter of the query then add all of the relevant ones that show up to the optimization list. Linking specialists can then focus on those keywords as well. One thing that hasn't been mentioned is that many people have installed the Google toolbar on their browsers and thus rarely have a need to use search on Google's actual site. For those who do most of their searches this way, nothing really changes with the addition of Google Instant. Caragh McKenna: Firstly, I presume we’ll see a hugely more accurate keyword research tool being released based on what Google can now glean from search behaviour at a top level...agreeing that long tail search numbers are going to drop significantly. Secondly, the speed of Instant which (I assume was caffeine’s goal all along?) surely encourages more regular site updates, with more engaging content...we are certainly seeing in the UK a significant increase in the news, images and video results when using Instant.  Blended results have been limited to date. This gives us an opportunity to really encourage deeper optimisation of these verticals! Esha Nandi: What struck me was that the number of impressions would go up remarkably for more frequently searched/popular websites than less popular ones. For example, in Google UK, “bbc” is the most popular keyword suggestion starting with the letter “b”. So, the BBC organic listing would be shown even when a person intends to search or type “british airways” or “Barclays” or even “brand value” (The listing would be shown as soon as I type in “b” but before I type in the second letter of my search term). But Barclays organic listing would be shown only when a person searches for “Barclays” or “band” or “banner” or “base guitar” (i.e, terms starting with “ba....” and before the searcher types in the third letter). So the increase in impressions would be significantly high for Barclays but not as high as it would be for BBC. SEM – Could result in higher CPCs on high-volume keywords (or not). Frank Lee: There will be a need to expand your keyword buys to include more “partial terms”.  For example, if you are buying the term “honda,” you should consider buying “hon” and “hond” since your “Honda” ad could be broad-matched to the search results with Google Instant.  Furthermore, if you don’t have a robust list of broad or phrase match terms in your account, you may not show up on the SERP results at all for these partial queries.  It potentially expands your keyword list in a significant way.  I wonder if their keyword estimator tool will now give impression counts in “partial terms.” Google is changing the way it counts an ad impression.  Since the results can change in real-time, Google will now define an ad impression if a user pauses on the search results for 3 seconds or more (based on the assumption that a cognitive pause was three seconds long and they were now digesting the contents of the search results).  Although they claim this shouldn’t have a material impact to clicks, only time will tell if we see a difference. Ami Grant: 1.              Google Instant will change searcher behavior - increasing engagement, while potentially making searchers more savvy. 2.              CTR may increase as a result.  A user may see Google’s auto fill queries as more relevant than their own and therefore, be more likely to click on an ad. 3.              Impressions will change on head/tail terms depending on the vertical. In this real estate example, users may be steered away from searching on more generic, non-geo keywords and encouraged to select the more geo specific, relevant ad. 4.        Advertisers may need to buy additional keywords to optimize performance and ensure there are no missed impressions. In the example above, if a user is “steered away” from a non-geo query, the advertiser will miss out on impressions if not also bidding on geo queries. Alex Campbell: I’d predict two significant effects for search queries: 1) Google Suggest becomes the de facto gatekeeper for short queries, making targeting keywords featuring here more vital than ever; and 2) tail terms will see an increase in usage, but queries that are physically long (as in many words) will be used less as users see the results they want before writing out every word. I think despite Google’s new approach to measuring impressions many AdWords accounts will see an inevitable dip in CTR and potentially QS, but I’d envisage Google working to normalize this behind the scenes as the rollout gains pace. Matt Grebow: Although even Google doesn’t dare predict the effect of Google Instant on paid-search advertisers, it could, conceivably, increase the average cost per click of keywords. While Google’s previous drop-down search suggestions could be easily disregarded, moving these to the search box itself may lead to a greater proclivity for allowing Google to complete its suggested searches. This could consolidate queries—increasing searches for high-volume (Google-predicted) keywords, while reducing searches for so-called tail terms. This could shift budgets to more expensive keywords (for advertisers who can afford it) and force advertisers who can’t to reduce their spending. Nathan Price: In terms of my own personal (non-work-related) search behavior, this week was the first time I can remember in months that I’ve actually gone to google.com to type in a search (to see how this new feature worked) rather than using my browser search box. I’m sure there are plenty of people not like me whose default is to go to google.com to do their searches, but knowing what that proportion is would be useful as a modifier in gauging the potential scope of the impact here. AUTHOR’S NOTE: According to Gigaom.com, ad network Chitika is reporting that 18% of Google’s traffic yesterday came from its home page.    We would welcome any additional data points on the % of Google searches originating from browser toolbars vs. Google.com Bradd Libby: What aspects of SEM bid management are going to change the most due to Google Instant? I'm going to go against the current and say: Not much. There's the idea that Instant might drive more searches and ad impressions, but search and impression volume have already been going up robustly year after year. Google started the US roll-out of Instant yesterday. Are there large-spending accounts (like TSA's clients) that saw, even from this partial day's data, noticeably higher impression volume on Wednesday than the previous day or week? Because I don't see any. Some are worried about the possibility of higher CPCs and lower Quality Scores, but as I explained in Maximize Your CPC and The Highest Quality Score is not Always the Best Quality Score, these effects would be a good thing. The fundamentals of bidding, like Google chief economist Dr. Hal Varian's description of using the Bid Simulator to determine your profit-maximizing bid, are solid. Bidding and data analysis will still use the same techniques and formulas, we just might have different specific numbers to look at. The biggest possible impacts I see will be on user experience first and perhaps keyword selection, not bid management. SEARCH BEHAVIOR - Faster may not always be better. Grant Simmons: Tried a typical search looking for “cooling fans” (yes, I’m geeky like that).  A few observations/key takeaways:
  1. It got where I wanted to go, not as “quicker” as Google inferred in their demos and ‘time-saving clock.’ The slow down came when my eyes went up and down from suggestion to search results... It was a little distracting... What’s interesting is that paid search, especially when there’s images at the top of the right search results, drew my eyes immediately... Good call Google on better CTR on those ads.
  2. There’s going to be many more partial search queries – I found myself pressing enter quicker and actually ending up with a different search result than displayed e.g. “cooling f” <enter> doesn’t deliver “cooling fan” results.. Even though the auto-complete is there. It’s no different from the former Google Suggest in that regards, it’s just that the instant results shown are from the “suggest” not from what you get by pushing enter... Major disconnect here.
  3. Why have “I’m feeling lucky” in the drop down suggestions? It’s distracting and I’ve never understood the value there, nor the ‘fun’ aspect it infers. I wonder how many people click on that?
  4. Distraction, distraction, distraction. Having all the movement  whilst I’m search is a little off-putting. Power users who type fast are treated to a frenetic slideshow.
  5. It’s not perfect (yet) - I tried typing in my old company “Simmonet Marketing” - It took until “Simmonet Market” for Google to present the correct results, but add another “in” to the search query and BANG Google thinks you’re onto something else (Simonet) - so is this part of their thought process? If you don’t select the result it thinks you should after two tries, the suggestion moves on? Interesting logic, something to test.
  6. FINAL THOUGHT – Wow! Google has an awful lot of computer power.
Aryn Kennedy: Since it seems to work most often when the user is signed in to their Google account, it will primarily affect searches when the user already knows what they’re looking for and has performed a similar search previously. For example, if I start typing “home”, it gives me options like “home depot” “home goods” and “homeworld.” I assume that’s because I visit Home Depot and Home Goods with some frequency. However, I was actually searching for “homework,” and it didn’t give me that as an option until I typed the “k.” Thomas Ciszek: There is little research on the effectiveness of this technique for search.  One interesting study came from Microsoft back in 2007. The interactive query expansion (IQE) variant, Real-Time Query Expansion (RTQE) can lead to a better quality of initial queries; however Google Instant Search emphasizes that more care is needed in how query expansion can be presented in the interface. Ted Ives: While it's an interesting idea and impressive technically, I think the thresholds they have set leave something to be desired.  If you type for instance [how to fix]  Google immediately brings up results for "how to fix a leaky faucet"; I can tell you with some authority that the probability of someone wanting those results, just given that they've typed [how to fix] is EXTREMELY low. Why bother bringing results up with such a low probability of satisfying the user?; the net effect is mostly to distract all the other users who don't care about faucets. Try "where can i" and it brings up "where can I buy silly bandz" (whatever those are).  Silly bandz may be the most popular query, but it's not worth distracting all those people that want to buy Britney Spears Albums, N,N-Diethyl-meta-toluamide, or gold coins!   I think the thresholds they've set for this to bring results back are ridiculously low and they will have to revisit them down the road. ________________________________________________________________ Thanks to all our agents for sharing their initial thoughts on Google Instant.  With less than 24 hours of experience, it’s fair to say we have more speculation than prognostication and observation than insight. We’re just getting the dialogue started, so leave a comment and let us know what you think.   Is this a game-changing innovation for search marketers?  Or a cool, but over-hyped feature that won’t materially impact your approach to SEM or SEO?

Tags | , , ,

12 Responses to “Google Instant Insights”

  1. George says:

    Ted, could you please explain “some authority”? :)

  2. If you’re wondering how Google selects the ‘auto complete’ queries for Instant

    100% hands off – algorithmically determined

    http://www.google.com/support/websearch/bin/answer.py?answer=106230

    @Ted if “where can I buy silly bandz” is a mystery, you obviously don’t have a 13 yr. old daughter :-)

  3. Ted Ives says:

    My authority stems from having fixed many things, none of which are faucets.

    No, seriously – I had occasion to examine all queries in the “how to” space a few years ago and it is an extremely flat long tail space.

    To illustrate, a quick run through the Adwords Keyword tool shows these global search volumes for last month:

    how to fix a leaky faucet – 8,100
    how to fix an xbox 360 – 49,500
    how to fix a computer – 40,500
    how to fix an xbox 360 – 60,500
    how to fix windows xp – 27,100
    how to fix a toilet 14,800

    8,100 out of the whole is somewhere around 4%; but there are *thousands* of queries that start with “how to fix”, every one we could add into this analysis would drive the percentage down until it’s next to nothing. How-To queries in particular are an extremely long-tail category. You would quickly see this number driven down to far far less than 1%.

    Showing results for this particular query when the probability it’s what the user wants is only << 1% doesn’t outweigh the distraction it gives to the other 99++% of people doing this type of query.

    Now, why didn’t they pick some Xbox topic? Clearly they think I’m more prone to want to fix a leaky faucet for some reason – it’s probably some personalization or regionalization going on there.

    I notice for “how to cook” a bunch of clam cakes and lobster items came up – local food for Rhode Islanders like myself.

    The same argument applies though, even if it’s a little more targeted, making the percentage a little higher…making a wild leap that I want to cook clam cakes based on my typing “how to cook” seems more in the realm of palm reading and crystal ball gazing than machine learning and behavioral targeting.

  4. David Chung says:

    Another area I’d like to possibly explore further is the impact of Google Instant on SEM campaigns that have strict geo-segmentation requirements.

    According to Google, in the hierarchy of precedence, query-keyword matching trumped the campaign’s geo IP target settings. If someone were to search for “life insurance california” (purely an example), traditional Google would parse the geomodifier out of the query, and consider it an exact match to your [life insurance] keyword sitting in a CA geotargeted campaign.

    That being said, what if Google Instant predicted “life insurance california” as the top suggestion as you typed l-i-f-e i-n-s. This isn’t the exact example, but I’ve found a few high volume ones like this.

    Google Instant’s “preview” does NOT parse the geomodifier from the query. Therefore, if I am in NY, searching for life insurance, I will start seeing ads for CA Life Insurance.

    Managing separate campaigns by state, this is starting to show some unwelcome colors in my data.

    My CA targeted campaigns should not be serving those ads on my computer in “preview” mode because it will open the door to invalid clicks.

    I checked some before and after data on the % of geomodified query parsed clicks, and % of clicks originating from a state outside the geotargeted region, and I saw huge spikes of 100-200% additional clicks.

    When I tracked those clicks to the actual keywords, lo and behold — they were high volume phrases that Google Instant threw a state term at the end of before I even finished typing my token keyword phrase…

    While I have always been cognizant of the reason for Google’s utilization of geo query parsing, there certainly seems to be a need for a control to “turn off” query based geo-targeted serving, especially with Google Instant throwing random state queries into high volume search tails.

    Try typing “keyword in” and see what you get — lots of people bid on keywords like “plumbers in los angeles” and may now get served ads for Indiana Plumbers (by advertisers specifically targeting IN only) if Google predicts plumbers indiana to be the top suggestion.

    By the way, first time commenting here, but a long time reader. I love the research and studies you do and leverage them all the time in my analysis. Keep up the awesome work.

  5. Alec Green says:

    David – Thanks for the comment, and thanks for being a regular reader of the blog.
    You make some great points about geo-targeted campaigns. Our team is taking a look at this as well, and we’ll report back any interesting findings.
    You may also want to check out our new white paper on Google Instant. Would love to get your feedback on that as well.

  6. Paul Cherepy says:

    At the moment Google Instant may not have a huge impact because it’s only available to those with modern browsers who use Google home page as their default search method. Most people tend to us the integrated search boxes within the browser, although I’m sure that will change soon.

  7. Wow, good post thanks for posting !

  8. Waleed says:

    Am I the only one not able to increase my # of Results to 100, ever since Google Instant? It would make sense, from a bandwidth optimization perspective, but it’s frustrating as a user who likes to see all 100 results on one page. Let me know if there are any work-arounds. Thanks

    • Alec Green says:

      I noticed that as well. Based on the sheer number of comments in this Google Help Forum, I’d say we’re not alone. And based on their tone, I’d say we’re not the only ones that find this very frustrating.
      My guess is Google assumes the vast majority of users don’t even know how to customize the results page. So limiting it to 10 results under Instant wouldn’t seem like a change at all.

  9. @TheButterPeanutJelly yes

  10. Fabiola Koeppl says:

    You are my breathing in, I have few web logs and sometimes run out from to brand : (.

Trackbacks/Pingbacks

  1. Google Instant Insights | Christopher Auman

Leave a Reply

Follow Us on Twitter

Authors