fbpx

Tag: Machine Learning

Facebok’s Infamous Ban Hammer & The Best Case Against AI

Facebook’s robots are great at finding customers. You can create a CBO with 10 ad-sets targeting 10 different things, 5 creatives for each of the ad-sets and the robots will start optimizing on the first day to get you purchases before you’ve even spent $50. That’s great, isn’t it?

When Facebook’s AI gets mad at you though, its quite like Skynet from the Terminator. It tries to erase you and it goes to great lengths to do so. Let me explain that one.

What happens when you use a credit card on Facebook that was issued in your name but by a different country than the one you’re residing in? You’re accused of credit card theft of course. Then you go to great lengths to justify it but it doesn’t work. Then after a few days of efforts, you make it work. But doesn’t matter if you’ve made it work even, cause that’s strike # 1 and AI is going to be tougher with you.

What happens next? One of your ads is rejected for “circumventing Facebook’s advertising policies”. You appeal it, and the ad gets approved. Doesn’t matter though, cause that’s strike #2.

And just like that, before you know it, you lose you business manager. Because you as a person was the cause of trouble, you also lose your personal ad account. In addition you lose your ability to advertise on Facebook or add or remove anyone or anything in any business manager.

If this sounds bad, let me tell you what else happens.

Any pages that you have previously advertised on Facebook get penalized too including the one for this blog that you’re reading. It doesn’t matter if the advertising was done months ago, you still get the hammer. The page gets zero reach syndrome.

Oh Facebook, why does it always have to be this way? Am I always going to feel this way about you?

Visa Restrictions On The Internet

While I acknowledge that internet is the biggest game changer of the human history, it is still not what it was supposed to be or what it can be.

In the offline world, all humans aren’t born equal. They are born in certain conditions where their lives are driven by their socioeconomic circumstances. Their lives are dictated by their place of birth and half of what they can and can’t do is written in the stone. Sure they can break the chains and the barriers to come out stronger but that happens very seldom.

On the internet, everyone has the same opportunities. In theory, though. You could get on YouTube and make a living regardless of where you live. Hundred of millions of people have benefited from such global opportunities that didn’t exist 30 years ago. But these global opportunities are still not provided equally to everyone, although they are marketed as such always.

For example, YouTube first launched their monetization partner program in 2006 for select countries but it wasn’t until 2016 that this program was launched in Pakistan.

Facebook restricts fresh Pakistani ad accounts at Rs 1000/day spend ($6.46 on today’s exchange rate) while a fresh US account is restricted at $50/day spend. Restrictions are lifted more quickly for the developed world, and less quickly for the emerging world. The scrutiny of AI is much harsher for us than it is for the developed world.

In summary, AI is no different than the visa issuing officers that judge us more than our counterparts elsewhere in the world.

On the bright side, in the offline world, if you try hard enough, you can sometimes circumvent these restrictions. You can sometimes emigrate or obtain a better travel document by meeting certain criteria or just by wanting to have it bad enough.

Since internet is almost always better than the offline world in every regard, you can also circumvent a lot of these restrictions if you try hard enough. In fact, on the internet, like everything else, this circumvention is often not as hard as in the offline world.

Perhaps, we’ll talk more of this circumvention in another blog.

Growth Risks With Facebook’s Machine Learning

Facebook has very advanced machine learning capabilities. More often than not, you’re better off reaching your customers for a cheaper cost by reaching a broader audience instead of a narrow targeted audience. But how is that possible? In theory, targeted audience should work better? But with strong ML, the broader audience delivers better and cheaper results provided that the initial customer dataset was correct.

But what happens if you get the initial data wrong? It puts their ML chase your customers in the wrong direction. Let me explain.

When building a Facebook page, growth is going to depend a lot on you first 100s or 1000s likes. Hence getting your first subscribers or customers wrong, can put you altogether in the wrong direction. I can think of 2 reasons why that could happen. Firstly, your upcoming page subscribers are likely to come from the network of your existing subscribers due to sharing and other engagement. And secondly, the engagement behavior of the first data set of subscribers with your content will define how engaging your page is and eventually define the placement of your page in the newsfeed and other Facebook algorithms.

So getting the initial dataset of subscribers/customers is extremely important. It is why I’m generally way more careful in the start when building a Facebook page or an e-commerce store through Facebook ads but later on take the liberty to test all kinds of traffic. It keeps my seed-data clean. The data that is going to be used to build the entire user-base later on.

If you have a question, please feel free to ask in comments.