Why is it important?
As we mentioned earlier, both Facebook and Google use machine learning to improve the performance of ads.
One of the biggest problems that both platforms have is that they need to accommodate a large variety of businesses, from different industries. As a result, different companies can have very different goals.
Political campaigns might optimise for donations, mobile game companies optimise for app installs, recently founded startups might optimise for leads which will later be converted to customers… It seems impossible to create a single algorithm which would cover all of these cases, which is why both Facebook and Google ask for input on how to optimize the campaigns, and they offer multiple options.
The problem arises when a wrong bidding strategy option is selected. Let’s take a common example with campaigns that optimise for link clicks. The idea is to get as many visitors to the landing page as possible.
But what happens after they visit the landing page? There is no guarantee that these people will interact with your business the way you want them to, they might just visit and leave. And this is often the case with campaigns optimised for “link clicks”: they generate a lot of traffic, for low price, but the bounce rate of that traffic is very high, and conversion rate very low.
The key takeaway from this example is that machine learning algorithms have a sort of “tunnel vision”: when given an optimisation goal, they only focus on that goal, and ignore everything else. This is precisely the reason why we suggest using optimization goals that are directly correlated with the success of your business: optimizing for conversions on Facebook, and optimizing for target CPA/target ROAS on Google.
What if the test fails?
When the test is being performed, the app takes the previous 30 days and looks at how much of total spent has been invested into ads that focus on bringing conversions.
For Facebook, if most of the money is not being spent on ad sets focused on conversions, then the test will fail. It’s best to interpret this as an opportunity, as changing the bidding strategy to conversions is a low-hanging fruit.
Similarly on Google, the test will fail if most of the budget was not invested in campaigns optimised for target CPA or target ROAS.