Running marketing campaigns to the target audience in Google or Facebook, is the duty of every marketing agency.
Things have changed a lot since the times when marketers had to create massive campaigns that would be delivered to every audience equally.
Today consumers are curious, demanding and expect immediate results, most notably on mobile browsing which is where most transactions on ads happen.
Machine Learning happening in our platforms
Every time we create ads, segmented audiences and optimize the budget, we are teaching Facebook and Google algorithms what works better to send out the right message to our target audience.
Initially, we would handle changes and optimizations manually, now the creators of the biggest publicity platforms want to save us the hassle creating machines to register our movements and learn from them to predict behavior patterns and to suggest the best investment and communication strategies.
Now and for the moment, machine learning is not done immediately, in both cases the robot requires approximately 7 days to detect any changes or patterns that could be important to our marketing campaigns and that might escape us (ideally and with KPI registry, we could detect such changes on time).
But what creates these changes? That would depend on the platform.
Google see, Google do
On Google, ads are created by bids, which is how the value of a certain amount of ads is established from day one.
However if there are time constraints, a change in target or strategy, the bid can be modified, that’s when the Google Ads algorithm starts compiling every bit of data to learn the hows and whys of the change, thus learning to offer optimizations better targeted and relevant depending on the performance of the machine.
At the same time, on Facebook the learning period begins when we make changes to a campaign that is already active; when the learning begins, the robots now can determine the public to show the ads to, making it more efficient and automatic.
Does machine learning affect my campaign?
Considering that there are certain movements like the trial of a new bid strategy, changes on the existing bid and changes on the budget that push the learning period, we can relatively avoid that our campaigns enter learning mode, since while this happens, delivery and efficiency can go down, while the cost per acquisition (CPA) rises and the conversion rate falls.
This doesn’t mean that we should just freeze there and not optimize our campaigns to get the best results out of fear of entering the learning period, it means we have the chance of doing two things: understand well the final result of our campaigns or learn from them and avoid changes on the run, or simply allow our platforms to learn and give us a hand on obtaining a common result: for us it’s the campaign’s success, for them is learning all the produced data.
Fuentes
www.wordstream.com
www.blog.google.com
www.bigseoagency.com