Moderated by Jessie Stricchiola
Eduardo Llach from SearchRev. He will focus on the kinds of techniques to use with “early terms,” and john and Sharon will focus on tail terms and some pitfalls. Traditional criteria used for targeting: keywords, ad copy, landing page, and the search network. How to split up traffic based on additional criteria such as location, time of day, etc.
He discusses “multivariable targeting.” How to use these techniques to improve the performance of the campaigns. For example, “online dating.” How to split it up geographically and by demographic, as well as time of day and day of week. Also, they want to optimize the creative. What works better in NYC verus Denver or San Francisco?
Geo/metro targeting. This works using the IP address of the searcher. AOL does pose a problem for this since all seem to be in VA. Country mapping is very accurate, state mapping is now accurate to 50%, citty mapping is accurate to 30%. He shows a spreadsheet detailing some home security system targeting conducted in about 20 cities. They found that it worked to aggregate metro areas by, for example, top 10 metro areas at Google. When they study this, they can find information and determine similar CPC but much higher conversion rates, for example. In these cases, they can use the data to recommend an increase in bids in a certain area.
They also adjust the exposure by sometimes using the search engine only, or adding it’s “search network” to increase visibility. With syndication, they can then compare conversion rates across each of the sites. If not taking the time to do syndication targeting, he would recommend running branded terms across all networks, in the example he shows. Remember that better conversion rates do not automatically mean more orders. Even though Google may have a higher conversion rate, there may be more raw conversions on another engine, for example.
Conversion rates do vary across industry. For example, although normally abysmal results in the content network, they found that a non-branded search in the automotive vertical had a very high conversion rate in contextual placements. He recommends structuring a campaign that works by optimizing beyond the keyword level. This multivariable optimization allows for additional tweaks based on geo and other factors mentioned above.
Time of day and day of week. Focus on conversion rate here…when it is traditionally up, bid up, and when down, bid down. Showed an example of one campaign with much higher conversions on the weekends. This can be 78% more effective if you bid up and down each day. They have noticed widely variant numbers based on the industry. For time of day, they focus on morning versus afternoon. The idea is to measure against the overall conversion rate. They found that some mornings were better than others, and some afternoon on different days were better also. You can use Google and other engines to set rules based on increasing/decreasing bid based on this type of data analysis.
Creative optimization. Each kw and creative will perform differently. Showed one example with a wide variance in conversion rates based on the creatives.
John Kelly from Sure Hits. He will focus on managing tail phrases. They mostly manage clients in the financial services area. So what is a tail phrase? They describe it as any keyword which doesn’t get a lot of volume. Why care about these? Large aggregate volume. Clear intention = better conversion rates. Less competition = lower bids. (Editor note: in some cases, this may not be true, especially with the page relevancy algorithm which sometimes may jack up a minimum bid even if there are no others bidders)
When we analyze the potential target, we need to find out what they want. Then 1 Calculate the click value. 2. Reward word choices, and 3. Watch our for tail dangers. So, how to calculate the click value? [Probability of conversion X Value of conversion] Is small volume a big problem? What if one click every six months? How to estimate the quality and value of the conversion?
Shows an example of a search for “Columbus Ohio car insurance quote.” This includes two great clues: Geographic data and product information. How to deal with this information? A bad way would be to use buckets: States, cities, city and state, car insurance, quotes. The problem is when someone types in quotes, they are twice as likely to convert, so bucketing causes the loss of this type of information. The right way to do it, he feels, is to “tag” tail phrases. You should tag each phrases for the different identifiers, and then use the information to estimate what the probability of conversion is based on the number of tags it has and each of their own historic performance.
How to bring this all together? [probability X value = click value] So in some cases the longer tail will have a much larger value. So what about rewarding the choices made by the searcher? Respond to them with creatives in ad copy as well as on landing page. He found many pages in the PPC listings that did not present the proper information to reward the visitor for choosing their listing.
A couple quick dangers: brand phrases. Many brands are built around cities, for example, “Tampa Bay Mortgage” is a brand. If someone is looking for a Tampa Bay mortgage, are they seeking any mortgage provider or the branded one? Watch out also for “fake” tail phrases, such as “Washington auto insurance quote online.” Too many searches per day on a term like this may indicate that it is being used as a title of a paid link somewhere. Last danger is homonyms, such as “Mobile home loans.” Is this for Mobile, Alabama or for a mobile home?
Last speaker is Sharon Crost from Red Bricks Media. She will speak about avoiding the pitfalls of PPC, or “How to get more ROI by Dragging your tail.” She introduced a couple of examples, but unfortunately I have to leave for a meeting so I will not be able to cover the rest of her presentation.