Misty Locke mods this one up.
Kevin Lee from Did-It.com is the first of two speakers on this session, he starts. Capturing the tail, there are millions of searches every day that are unique. These tail keywords and phrases are highly valuable because searches know what they want exactly, searchers notice the ads that the marketers have written just for them. Millions of unique searches per day, some searches occur once per month or less. Each marketer's graph looks different. The power curve is somewhat asymptotic but flattens out to one unique search per time period measured. He shows a quick example of a long tail search distribution. Campaign goals and objectives should line up with the profile of the searcher. Positive actions vary throughout the buying cycle. Searches at the head have ambiguous desires and needs for several reasons. In addition to the types searches deriving the search inventory at the head, the head contains, link driven traffic from directories, link driven traffic from within the portal, and link driven traffic from syndication patters. Aggressively pursue the tail of the search distribution when the intent of the different tail searches differs widely from each other and the tail searches repeat sufficiently often to justify the unique listing creative, landing pages and bids. Since tail keywords often convert well and customized creative listings capture searcher attention; ROI is high, bids often don't have to be as high. When you have keywords out in the tail, the impression counts are low. You set bids by conversion rates tend to be good, so go in aggressively to start, given the low impression rate and typical CTRs, you may not see significant individual clicks, so consider clustering keywords to get data faster. Google and MSN seem to do all the work when it comes to capturing the tail with broad match, but its not the most efficient way to do this (cost is higher, and quality score is lower). In addition to continuous keyword research, your existing campaign can deliver more tail keywords. Run some Google, MSN or Yahoo in broad/advanced match and watch the inbound keywords with analytics, campaign management software and raw log files. Google DKI (dynamic keyword insertion) comes in handy for long tail keywords. If the search query is too long, the DKI will use the "default creative." Yahoo's standard match always trumps advanced match regardless of bid, that means you need to predict searches as far out in the tail as practical, traffic the listings for those keywords and use advanced match to capture the tail (Yahoo does not have the DKI). Because the traffic at the head represents a huge volume opp but may be early buy cycle, so you may have to treat the head differently. You may want to segment the head in other ways (geo, day part, day of week). This is where technology and analytics become critical.Multitude of simultaneous choices of what to bid and where to focus. Competitive reactions, keyword volumes differ, keyword volatility differs, conversion rate and roi. When do you "kill" a bad tail keyword? Changing the bid or killing off a keyword that has good traffic volume is easy, you need stats to make these decisions. CLuster analysis can help you predict conversion rate. Bidding strategies can also make use of clusters. Stemming relationships within phrases and common landing pages. Targeting the tail and segmenting the head with other targeting parameters allow you to reduce waste, target the best customers, increase profit, improve your messages and offers, be more aggressive when it matters.
Harrison Magun from Avenue A Razorfish
"Taming Your Bid Monkey"
What is the probability of being a twin? 5% (fake numbers). What is the likelihood that the incidence of twins is actually 5%? He brings up a distribution curve, the probability is only 20.5%.
How big a sample do we need to be 90% sure that the incidence is between 4.5% and 5.5%? 5,044 people.
How do we use this data for keywords?
How many clicks do I need? Conversion Rate :: Clicks Needed 1% :: 25,000 2% :: 14,000 3% :: 9,000 4% :: 7,000 5% :: 5,000 10% :: 2,500
Say you have 400 clicks, conversion rate is 2% (you dont know that yet) and you act after 400 clicks, you have a 60% chance your made the wrong decision.
So how do we act on this data? - Don't waste your time on insignificant data - Sedate screaming lunatics (those that yell at you that you have 0 conversion on a hundred clicks) - Create accurate tests - Spread the tests out over time - Understand the factors that are impacting your conversion rates
How can you use these stats? - Excel - Your own categorization - Bid management algorithms and tool sets can save a lot of time but be careful - Add this knowledge into four levers of search -- Bid strategy -- Keyword creation -- Creatives and landing pages -- Business Intelligence
SES NYC Tag: sesny2006