Click to giganticize the Webtrends Great Data Giveaway screenshot

Early on at Webtrends, we decided to run a Facebook contest. That was pretty much the direction: let’s run a contest to see how it works.

So we wondered: what would make Webtrends’ faithful excited? And I came up with this idea of embracing the data nerd element. “Fly your nerd flag high” was an ad headline I remember.

We got sponsors like Wolfram | Alpha and ReadWriteWeb to give away a bunch of datasets our winners could correlate to their own data.

Anyway — I still love this text:

“Because you collect data the way some people collect unicorn snowglobes. (And get just as many puzzled stares.)”

It was a pretty successful contest. A few days into the project, once someone in management realized it wasn’t going to go down in flames, we picked up a new goal: to increase our “quality fans.” Considering we were giving away .csv files of historical Twitter data, these fans were definitely quality. And we went from 500 fans to about 2500 fans in a month.

I also created, bought, placed, and optimized all the ad campaign for this. It was during this time that we started to notice the 3-day rot for all Facebook ads. What that means is that most Facebook ads do pretty well for the first 3 days. After that, they start to get ignored by users.

So here’s a little media-planning tip: if you’re resourcing for a month-long Facebook campaign, you need 10 fresh ads. (And really, if you just use new images and new headlines, you’re fine.)

By doing minimal targeting — regions, for example. Or another variable not expected to influence results — then you could compare the percentage of impressions to the percentage of clicks to tell how varying demographics respond to your messages and offers.

In other words, instead of paying an ad testing company to poll customers, you can use Facebook to extend your campaigns for similar actionable data and yet more impressions. Ad testing can extend your campaign, rather than merely being an added cost.

It’s not the perfect solution. The biggest problem is that you’re only looking at the Facebook population, which biases your data. But since Facebook has grown so insanely fast and is becoming such a ubiquitous presence, I think it’s a decent representative sample.

Visit iStrategyLabs.com for more info on Facebook demographics