@rakhim: "When an app or a service is described as “AI-powered” or “ML-based”, I read it as “unreliable, unpredictable, and impossible to reason about behavior”. I try to avoid “AI” because I want computers to be the opposite: reliable, predictable, reasonable."
Programmatic Software (and the RTB Bidder-as-a-service as a part of it) is constantly becoming more complex. But bigger does not mean better. Usually, as soon as the new technology appears on the market – everybody considers to mandatorily implement it.
Moreover, no one asks themselves if they need it at all. You don’t have the time to think about it as competitors are already DOING it! Last year it was with ML, now it’s blockchain time.
You can not just say you don’t have ML in the programmatic industry.
A typical dialog between a client and a vendor:
C.: Do you have a predictive buying strategy?
V.: Em… I’m not quite sure I understand you correctly, but probably not.
C.: But why?
V.: What do we need it for?
C.: Well.. everybody has it. I supposed you are a serious vendor.
The right question is: how many of those who say they have ML integrated can actually say it’s working?
Even better: who has received an outcome proving that ML has value for their business?
A vendor who is constantly flaunting their experience with ML, blockchain and all of the other buzzwords is simply unsure of how exactly the offered product works. Indeed, it’s much better to say “the neural network is still learning” than “I have no idea why we’ve got this result”!
The truth is that if you integrate something and you don’t understand what do you need it for – you only harm your project.
RTB bidder as a service is a high-load software. It means it’s always a compromise between performance and complexity. ML learning integration tangles the decision-making algorithm, thus slowing down the application. It requires an enormous growth of the server infrastructure to keep the timeouts low. In other words, 10 times higher service costs will only give a potential 10% CTR boost. Are we talking about efficiency here?
We know what we do. We don’t integrate anything just because everybody has already done it. And yes – we still succeed in reaching the goals of our clients. Wish to see it for yourself? Just contact us – and we will show you how we do this!