Netflix offers me the film Lethal Weapon because I like 48 Hours, and recommends the TV program Burn Notice because I like Mad Men and Arrested Development. The predictions are sometimes pretty close to the mark, sometimes seriously off-target, but usually entertaining. [caption id="attachment_3169" align="alignleft" width="320" caption="Rush"] [/caption] In contrast, MOG thinks, "Since you love Rush, I bet you'd like to hear some Blue Oyster Cult." Nope. I wouldn't. Sometimes apps miss. And that's okay. Netflix and MOG are among an emerging class of applications that are working to understand our unique tastes by uncovering the delicate influences that determine why we want what we want. These apps are supposed to get to know us; bluntly customizing to our tastes is how the app proves its value.
Ads have to play it cool, or risk being creepy.
The shock of seeing a personalized offer in an advertisement, without understanding how the magic trick was done, can make people nervous. Consumers are correctly wary of advertisers taking notes and collecting data behind their back. However, there are a number of things advertisers can do, to personalize ads without crossing the line.
The key to personalizing without being creepy is to be up-front about where the data is coming from.
Don't hide anything. Advertisers new to personalization are sometimes tempted to make it seem like a magic trick. This is a bad idea. The payoff just isn't there. It's possible to make personalization feel like service instead of a privacy breach. Be open about why the ad is personalized and where the data came from to power it. If it's done correctly, consumers will appreciate what you have done for them. Ads that appear in personalized apps can typically be personalized without making us feel violated. When ads are personalized in an environment that's also personalized, it lends valuable context to the ad. This is why Facebook can offer up sponsored stories, and why Groupon can give you personalized deals. Consumers understand and are coming to expect this behavior. The data and the ad appear inside the circle of trust.
Be sure to consider letting people opt out of it.
Hugo Liu, Chief Scientist of Hunch, says that personalizing without giving people a way to opt out is like constructing a building without fire escapes. I support giving some audiences the ability to opt out of the personalized experience. Obviously, this depends on the audience.
Reject accessing stale data; just use the data that's right there in the open. It can yield the best results and help avoid creepiness.
Surprisingly, the most valuable data to use for personalization - the dataset that actually matters most - is usually the data that's right there in the open. Accessing past purchase history and Experian data might seem like an exciting opportunity, but it's usually not very relevant. Data becomes stale quickly. There's usually little to gain from using this sort of data. The most valuable data to make ads more relevant are:
1) Device type, OS and SDK.
Optimizing an ad for the device experience is critical to making the ad as relevant and actionable as possible. Sometimes we call this targeting, but it's just a form of customization. If a user is on Android 2.3.6, eliminate transition effects. If on iOS 5.1, offer a click-to-buy via iTunes.
Where is the user right now? What's happening around the user right now? Is she shopping, or at a concert? Is she in a hospital or a museum? What's nearby? Location data can make ads extremely actionable. GPS data is available in HTML5 and advertisers should not be afraid to ask users to share it.
What is the user searching for? Car dealerships or car parts? Coffee or tea? Have they abandoned a shopping cart? Listening for subtle clues here is important, especially for re-targeting. Partnership with a data management platform can really help with this type of customization.