We know that "with great power comes great responsibility": who do you trust enough to have this long chain, and how can we be sure that there are no "bad guys" along it? The answer is that it is really, really hard, if not impossible.
Unfortunately, this also means that all the precious information stored in cookies can become dangerous in the wrong hands. How can you be sure that we are not using them for something you won't like?
We chose the easiest solution: none of our ads leave cookies in your browser. No one can steal your cookies, if you have none.
Each time a user visits a publisher's page, or uses a publisher's app, our distribution server is called, and it decides which ad to send back. It does so by compiling a list of candidate ads, rating them by taking in count a big number of factors: the affinity between the ad and the space (e.g. content or geographic location), advertisers' and publishers' preferences (e.g. avoid political content), and the price (e.g. paying more gives advertisers more space). These ratings are treated as a probability to be sent out.
This means that no matter how low the price is, an ad always have a small probability to be shown. Obviously, an ad with high affinity and above market price, will be sent out almost 100% of the times.
The concept is easy: we "grade" each advertiser and publisher for how "good" they are. We call this grade "PV". Better the PV, better the deals. Also, for advertisers, more contract types are available with higher PVs (e.g. contract CPA).
Obviously, we periodically run anti-fraud checks and bots, so everyone has to keep up with our expectations.
Having a good PV is fundamental for Resove. Remember: we pair advertisers and publishers also based on this value. Winners will play with winners.
Well, it's actually very easy: we trust the advertisers to call us back when a user lands on their pages through an ad, and buys something.
This is also why we don't give everyone the possibility to open CPA contract with us. In order to do that, they have to reach a certain PV value [more on the trusting system], and we will routinely check by automatic means if they are respecting the contract.
It turns out that the whole process is easier and safer for everyone, and every conversion will be accounted for, no matter which software a user is using for browsing.
Thanks to a set of machine learning algorithms running in real time on all the requests, it is possible to categorize future requests without knowing anything about the users. As it was once said:
If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck.
Basically, we don't care about who the single user is: we know in which category the user fits.
We don't track users, and this is a fact. Anyway, publishers who want, can provide (with their users' approval, of course), in the ad snippet, a social media id for the user watching the ad. We will track this id and, based on the user's preferences, try to serve ads which have a potential bigger impact on that person.
It has to be noticed that we won't serve the same "best fitting" ad over and over again. As for any other ad:
If the user hasn't clicked the 49th time, he won't click the 50th.
This is why, even if we could do it amazon-style, we internally refer this practice as soft retargeting: we give every ad, despite the paid price, the best conversion rate predicted by our statistics.