Buying data to fuel an ad campaign is not rocket science: The real rocket science is buying the right data. If you want to make sure your ad campaign actually works using data -- or if you just want to understand how it's done -- here are some lessons from our experience in this process.
What follows are some key tips to run a successful, online data-driven media campaign. They're grounded in real-world experience and will help anyone run -- or simply understand -- how to engineer these campaigns.
Some of these tips will sound obvious. And if you're doing it right, they are. But you'd be surprised by how few people understand the basic rules.
Shop for data.
If you were buying a diamond, would you run out to the first retail store and plunk down your cash? Of course not. You'd shop around and do your research. Same thing applies to data vendors. Shop around. Get variety.
Research your data vendors
So you shop around and instantly find a few vendors. They all tell you they've got the perfect list for you, chock-full of people likely to buy what you're selling -- for the sake of examples, let's use cell phones. What to do? Flip a coin? Would you do that if you were buying an expensive diamond? You'd ask everything you could about the diamond -- where it was from, how it was mined, etc.
The same rules apply to data vendors. Research means asking a few key questions. Where did they get this data? How fresh is it? Why do they think it's the perfect list for you?
For instance, one vendor tells you he has data from people who just purchased cell phones. That tells you a lot. You're looking for people who haven't made that purchase yet, so that isn't the best data for you. Another vendor tells you she partners with websites frequented by people just in the research stage of buying a cell phone. That sounds better. This vendor is more likely to yield fresh connections to people who haven't yet made the purchase, but are ready to do so.
Finding the right answers usually doesn't always mean doing intensive research. Often, as you can see, it simply means asking the right questions.
Go for variety
You've asked the right questions. Now you have a few vendors you think will deliver the best results. And a few vendors are offering more than one list of data. That's great.
Variety is good. You can't know in advance what's going to work, but your educated guess will probably be pretty good. You should try multiple data types from multiple data vendors.
You can stir variety into your tests by taking advantage of the internet. Test at different times of day, different days, different places etc. Be as experimental as your budget allows.
Test the data
Now that you're ready to roll with some data you think might work, you need to test it. As they say in the intelligence world, trust but verify. Just because a vendor promises a hot list of phone shoppers doesn't mean that list is going to give you an instant return on investment. You've got to try it and find out how well it works.
Once you find data that works, you can use it and see how much it can scale -- but that's the next step. First, test to see if that data carries signal.
Build a healthy test budget and then prioritize it
If you want to run a valid test, plan on spending at least $5,000 to $10,000, if not more.
Why spend so much if you're just testing? Don't think of it as just a test. You're actually making a media buy. It's the only way to find out if the data you've bought actually works.
Think of it this way. If a drug company wanted to test a new cancer treatment, they'd have to test hundreds, if not thousands, of people to get statistically relevant results. Ads work similarly.
If you have a $25K test budget, you can expect to run three or four tests, at most. So prioritize which tests you want to run. If you have $1,000 to spend, don't bother. Your results would be as meaningless as if the drug company only tested five people.
Measure your results
Of course, you say: that's obvious. But you'd be surprised by how many people run tests and then don't adequately measure the results. It's not that difficult. But it does mean you have to be prepared: Make sure you have a clear, consistent way to measure data before you start. Then, make sure you have proper controls. Again, think like a scientist. You've paid good money for a test. Make sure you're measuring results correctly.
Be prepared for failure
By definition, testing means you have the chance of failing. That's why it's a test. Occasionally, you may run into bum data. Don't let it discourage you -- just move on to a different test.
That means when you run a $5K test and nothing happens, don't pour in $5K more. Stop. You tested it. It didn't work. Move on to your next source and be glad that you have the information.
Be prepared for success
Success can mean different things to different people. If you run an ad campaign and it works -- even a little -- pay attention. It helps if you scale your expectations appropriately.
Remember when you launched search on your site? It probably didn't work instantly. It took awhile to fine tune. Ad campaigns also take time.
The internet, business and well, human beings -- are quickly moving entities. Test all the time, and test continuously.
Richard Frankel is president of Rocket Fuel, Inc.
On Twitter? Follow iMedia Connection at @iMediaTweet.