Targeting Good Behavior

To be six again

My son just completed his first 30 days of first grade and came home with a report card and behavior chart, both stellar. (I’m a proud dad.) While his grades were great, I sense that now in his school acclimation process, the behavior chart is a tad more important. What could be more distracting to a six-year-old boy than to be in the same room with 25 other boys and girls? If he’s exhibiting good behavior, he must be paying attention to the teacher (we hope). To track his progress, the teacher places a smiley face on each day of the calendar with good behavior. Nothing like a month full of smiles!

Let’s dive into behavioral targeting, a topic receiving a lot of press lately in light of announcements from wsj.com and NYTimes.com. Most people think behavioral targeting involves tracking user behavior, then dynamically leveraging the knowledge to serve relevant messaging. I think of it from the customer perspective; how can I reward a potential or current customer with more relevant information, as indicated by their implied or expressed behavior? Or, how can we marketers hand out a few smiley faces? 🙂 (Couldn’t resist!)

You’re tracking my what?

Before the privacy hairs stand up on the back your neck, relax. We’re talking about data collected anonymously and in aggregate, after someone has created a specific, expressed relationship with a site (through a registration process, for example). No Big Brother here.

How does it work?

Say a user cruises through several MSN pages. She checks recent college football scores, then hops over to travel to research hotel specials and flight schedules. Then she clicks to monitor her company’s stock and glance at this week’s East coast weather. That’s your average user session. Perhaps she repeats this a couple days in a row, or over a couple weekends.

If you were a supplier of airline travel, you’d want to ensure your ad shows up in the airline travel section so the message is contextually relevant, right? But since you’ve enabled behavioral targeting criteria, you can also hit that same user with an airline ad while she’s in other MSN content sections because her implied behavior indicates she may be interested in a weekend trip to a college game in sunny Florida (if her brokerage account is looking up, that is).

This might sound far-fetched, but stay with me.

What’s old is new again

Many top publishers are getting serious about behavioral targeting, but one property has done it for years: Yahoo Once called “fusion marketing” (which earned the nickname “con-fusion marketing”), Yahoo has turned its wealth of user data into a tool to strategically optimize less-requested inventory, maximizing its ability to generate revenue (see the company’s recent buff earnings report).

The five ways Yahoo behaviorally targets users:

1. Keyword Search User types in a specific product, say “roller blades.” For the next hour, you can serve creative for the new Aero 9 model skates no matter where the user is within Yahoo’s network.
2. Page View That same person happens to use Yahoo’s directory to drill down to street sports or related content categories. They check out several pages related to skating. Based on that activity, you create a trigger to serve additional Aero 9 ads as they access other content.
3. Ad Clicks If our user interacts with ads matching your product’s category, you can use the activity to serve your message on a future page view. This isn’t Gator-ish. We aren’t hijacking anyone’s site or page view. We use expressed behavior to trigger the ad server to queue your creative on the next page. Think of it as “sequentially relevant ad serving.”
4. Transactional activity Skater girl decides she’ll hop a plane to this weekend’s game in Florida. She’d love to cruise the boardwalk on her blades. First, find a cheap airfare. After she books that low fare, your targeting criteria kick in. You serve ads for southern Florida hotel specials, as we know from the transactional activity she booked a Miami flight. More options: Not only can you target her after the transaction, you can use her shopping and comparison activities even before the flight’s booked. Her flight schedule look-up activity indicates our user is checking out Miami. Why not offer a package deal for air and hotel at one nifty price?
5. Category activity Yahoo’s behavioral targeting includes the ability to buy specific channels for specific behaviors. Skater girl will need a new travel bag to lug those skates. Why not establish a trigger that creates a connection between her travel and shopping behavior?

Take it to the next level

To really get sophisticated, you can create a targeting scorecard for each user using a points system based on a number of the above items. If you’re clearing out the remaining ’03 Subaru inventory, you can create a scorecard based on specific shopping research activities plus search actions for car research sites, referral traffic to other car sites, referral traffic to dealer sites, and Yellow Page look-ups. You can also specify number of days, frequency within days and consistency of the desired actions.

You may just learn the maximum conversion for driving a buyer to a dealer for a ’03 model Impreza is a mix of 25 auto channel page views within two days, plus click-through traffic to at least three auto consideration sites.

Worth it?

This all may sound a bit complicated, but people tend to exhibit similar shopping behavior before they pull the transactional trigger. You must identify that process for your client and map the trigger points for behavioral targeting.

For one airline client, we’ve found behavioral targeting can deliver between 30 and 40 percent more post-click activities and transactions than ROS (run-of-site) or even some contextual placements. Modeled correctly, it works very well.

Potential pitfalls

Items to bear in mind that can negatively impact results:

1. Saturation can create overkill Don’t create a trigger, then carpet bomb customers with every available impression. As with any frequency model, there’s an opportunity to overdo it. Carefully consider frequency caps.
2. Difficulty of determining delivery and inventory management As with any user-generated data modeling technique, there’s no way to measure the exact number of impressions you’ll deliver. Run some decent-sized tests and negotiate the assignment priority of your ad serving. Will your ads always be superceded by other marketers’ category-wide buys? Ask, or you could wait a long time to get the desired level of impressions.
3. Use customer research to determine ultimate impact Many online ad efforts require an offline conversion (e.g. a car purchase). So conduct post-session surveys to refine targeting algorithms and maximize success. You can trigger pre- and post-session research to determine if someone is more or less likely to consider a purchase. Benchmark behavioral targeting versus other buys to determine the lift effect.

Reward good behavior

What are we creating here? Behavioral targeting provides an increased opportunity for touch points between your brand and customers. We know users go online for a specific purpose or mission. Reward them with the information they want, via smart advertising.

Put on a happy face! 🙂