Does Demographic Targeting Matter?
Moderated by Misty Locke of Range Online Media
Jed Nahum of MSN Search Demographic Targeting and Business intelligence (BI). His team does product planning for MSN AdCenter. How do we provide BI and demographic Targeting? How accurate is the data? Does demo targeting and BI really matter?
Starts with an AdCenter overview. Discusses France and Singapore, then US launch. New release coming out over the weekend that includes Firefox support and a few other changes. So, how do we enable targeting? Pretty simple. Uses a screenshot from registration data collection methods. Windows Live ID. Collects birthdates, gender and aks some other questions. This data is used for both BI and conducting kw research, determining desired audience. Basically tells you age and gender distribution for people searching for a term. So if you know you are targeting specific demographics, you can adapt you kw list.
What percentage of searches are covered? “Not all of them, but some of them.” A sample size that achieves significance doesn’t have to be 100% of audience. “Bid Boost” product helps advertiser target their demographic more specifically. The cookie only uses age group, no specific information that marketers could get to. Because of privacy concerns, it is hard to release data to outside sources. How accurate is data? Internal assessment of data is performed using 2 different databases. Compares US data only. Identifies cases in which consumers overlap. 80% of the ages matched in the two databases. This speaks that there is a consistent match, and tended to be more accurate on the younger side, which is the demographic that is more likely to provide this information. Make of it what you will, but he feels that their internal assessment for accuracy shows pretty good results.
Information discovered falls into two different categories: one is a “duh,” of course it would be known, but other info is actually fairly surprising. For example, the searches for “diaper” are 61% male and 48% male under the age of 36. Research also led to discover that adult diapers are also being sought. Has to be careful to exclude some keywords and emphasize babies. More surprises: “Elle girl” searches 48% male (searching for Elle Macpherson, mostly). Travel: “honeymoon” splits 50/50, but “honeymoon planning” types terms 75% female. Allows you to find what Kevin calls “power segments.” “Ferrari” mostly kids, so you want to avoid catering to them in the ad copy. Process: Reaffirms what you already know. Learn new facts about your audience. Discover intelligence that changes your strategy.
Finishes off with some quotes of SEM’s giving MSN great ratings on ROAS. Also had a case study that he was going to show which shows great ROAS. Check out adlab.microsoft.com for “funky cool stuff.”
Kevin Lee from Did-It.com Wants to reiterate some general best practices with some case studies. (Kevin is in a documentary about SEM that is being produced, so there is a camera on him.) “Better targeting brings us closer to the holy grail of advertising.” First thing to think about with demo targeting and any bid-boosting activity is “what does my regular media plan look like?” Chances are that within those plans there are some good demo target parameters. The question becomes “where did you get the data for that targeting?” You can also regress current customer information and intuit demographics based on analysis of names, for example, to help determine gender. Look at existing customers and see which ones are the best. You can also try to collect demo info during checkout process in ecommerce, for example. Start to append demo info to the user info that you are databasing. This will lead to the eventual ability to bid boost based on demographics. Find the customers that are particularly profitable to you, as Jed mentioned he calls them “power segments.”
Demo segmentation: which demographics actually lead to better conversions? If you can determine this, it will guide you ability to bid boost for them. For example, if the conversion rate of non-identified customer pales in comparison to a particular demo, target them. Does one segment order more over time than another? What is your 90-10 rule? Many questions to consider. Geographic segmentation may be the most powerful method of segmentation you have. Clicks are worth different amounts in diff areas. Clicks from diff sections of the world or country have diff preferences.
Case study #1: car buyers. Dit-it and a major car lead-gen site team use age, gender , and geography based on feedback from customers, the car dealerships. Cherry pick the good clicks with the highest potential to convert to a good customer. Regression analyses will help you understand which are most valuable. Another case: B2B marketer MSC Industrial Supply. Faces a problem many B2B’s face of having consumers “get in the eay” of business leads. They’d prefer not to sell one electric motor, but are more interested if you want to buy 400.
MSN adds increased targeting options by allowing you to raise your bids. One way to target on behavior is control you have over listings allowing you to use day parting (times of day) and days of week. Day parting in Google and Yahoo! Require use of third party system. Shows an example of data from Match.com: different predicted click value is based on times of day and week. Chart goes from very low conversion at 5:00am to very high between 8:00 and 10:30pm. Becomes more difficult for a competitor to monitor your position, since they will not notice if you are bidding higher in a particular area (i.e. the LA DMA). You are up against the two types of marketers: brilliant marketers and total lunatics (a classic line from Kevin that always gets approval). Use power segments to reduce waste, increase profits, and improve ads and performance.
Michael Sack from Inceptor Calls himself the “doubting Thomas.” Decided to do a study to see if demo targeting really matters, or if you can use other techniques to get good results. Starts with an explanation of the concept of demo targeting. Does it matter? It depends – are you trying to drive brand, revenue? Does your product fit a particular demo? What is your budget? The dirty little secret: “half of advertising doesn’t work, we just don’t know which half.” Things to keep in mind: when targeting mass audience, the idea is that the majority meets demographic target. You must expect exposure to far more than demographic target. It is a “Push” concept (vs “Pull”)? Online demographics may differ from offline. Highly dependant upon information (cookies). Privacy issue are abound – the system is entirely dependant on people giving accurate information. He says that he uses several different “personas” when doing research, even to the level of saving cookies in a folder and using different computers. He was somewhat suspicious.
He devised a test. Using MSN AdCenter – test a group of 50 keywords as follows: 7 days with no demo targeting. 7 days with demo targeting. 7 days without, but after creative was optimized for the desired demo target. He used Bid Boost. Results: when he added demo targeting, traffic dropped significantly, and drop in cost. Avg cost per click went from 29 to 37 cents. Conversion rate went up. When he added new creative, conversion rate went up as well. As a result of improved conversions and lower cost of non-targeted optimized ads, they did better. Revenue was significantly better without demo targeting.
Conclusions: does demo targeting work? Yes. Increases conversions and ROAS, reduces cost. However you do sacrifice traffic and revenue. Is it as effective as keyword or creative optimization? Not in the results of this small test. There could be inaccuracies in testing, he admits. Should we employ demo targeting? Depends on objectives. If goal is to acquire as many customers as possible for minimal acceptable return, then “no.” You should be optimizing keywords and creatives. If you have a limited budget/niche market: “yes” demo targeting does make a lot of sense. Online demo targeting at MSN is very diff from offline targeting or Google’s product. You do miss potential audience with demo targeting that could increase revenue.
Q&A speaker is Dana Todd from Sitelab International, Inc., and along with Kevin, a member of the Board of Directors of SEMPO (Dana is President).