> Take an audience of X people. Divide them in two. Show ads to your test group, don't show to control.
That works for the simplest of advertiser campaigns, but doesn't scale.
How do you break down your audience across platforms? I.e., how do you divide your audience in half across your ads buys on Google & Facebook, to make sure someone excluded from your ads on Google is excluded on Facebook?
And that's not even the hard problem of how you do it across digital, television, billboards, magazines & so on.
Even if you could get that data, if your campaign is flighted, you don't have enough time to act. If you have a big movie opening or product launch, you can't wait to see the initial response & then optimize.
These approaches can work for someone taking advantage of the latest Instagram Dropship scam, but they're going to work for where the bulk of advertising dollars are (big brands, movies, cars, consumables, etc.)
> That works for the simplest of advertiser campaigns, but doesn't scale.
Actually, the examples I'm referring to are at pretty massive scale (>$100mm in spend) all the way down to <$100k. It works across a pretty massive range.
>. How do you break down your audience across platforms? I.e., how do you divide your audience in half across your ads buys on Google & Facebook, to make sure someone excluded from your ads on Google is excluded on Facebook?
This is a challenge for sure and it depends on how you do it, and its also where vendors like LiveRamp and Oracle do so well. As they have very high quality data sets that work across platforms they can do these sorts of tests and exclusions.
However! That isn't actually necessary. Typically when you do a test, you will isolate the test to certain ranges and hold other campaigns constant. As long as you have enough data to create a solid predictive model you can pretty easily isolate your effects between different channels.
> And that's not even the hard problem of how you do it across digital, television, billboards, magazines & so on.
See my above answer :)
> Even if you could get that data, if your campaign is flighted, you don't have enough time to act. If you have a big movie opening or product launch, you can't wait to see the initial response & then optimize.
Now this is a challenge for sure. For big launches, its much harder to create this kind of testing during the campaign. This is why testing ahead of time is important and for big launches media buyers are running off what they already know about a channel vs. testing during.
> These approaches can work for someone taking advantage of the latest Instagram Dropship scam, but they're going to work for where the bulk of advertising dollars are (big brands, movies, cars, consumables, etc.)
Completely disagree.
If anything, this doesn't work at all for small scale advertisers unless they spend real $'s on their campaigns (>$30k/month at least). Technically you could do a test like this at small scale, but unless your CPA is super low or you get a metric ton of clicks and impressions, you won't be able to get a key learning from it.
That works for the simplest of advertiser campaigns, but doesn't scale.
How do you break down your audience across platforms? I.e., how do you divide your audience in half across your ads buys on Google & Facebook, to make sure someone excluded from your ads on Google is excluded on Facebook?
And that's not even the hard problem of how you do it across digital, television, billboards, magazines & so on.
Even if you could get that data, if your campaign is flighted, you don't have enough time to act. If you have a big movie opening or product launch, you can't wait to see the initial response & then optimize.
These approaches can work for someone taking advantage of the latest Instagram Dropship scam, but they're going to work for where the bulk of advertising dollars are (big brands, movies, cars, consumables, etc.)