That’s interesting to read about, Thank you for sharing this informative stuff..

Not sure why it’s unprofessional. Remember, the whole point of a patent is to push information *into* the public domain. In exchange for publishing how an invention is built you receive the exclusive right (the patent) to that intellectual property for a number of years. In many cases if something is truly “secret sauce” (eg, the recipe behind Coca Cola or many financial trading algorithms) are deemed trade secrets and hence expressly *not* patented so that the public will *never* find out. In the case of RM’s predict… you *can’t* just re-build it… because Yahoo! owns the exclusive rights to this IP. Not to mention the fact that this document is four years old and given the pace of innovation probably not relevant to how the platform operates today.

For AppNexus — yes we’re also filing patents and yes when those issue you’ll be able to read all about our secret sauce. Setup a Google alert for AppNexus patent and you’ll even receive an email when they enter the public domain =).

In terms of your question about seat specific optimization — absolutely. There’s a problem of averaging… there isn’t one algorithm that works for everybody and it certainly makes sense to build models that are specific to certain buyers or buyer types. Personally I’d orient more to brand or advertiser specific rather than seat since a seat can represent many different types of buyers but that’s just my bias.

Cheers,

-Mike

We’ve been told RMX had changed their predict few times since then and are not working on customized predict per seat.

Will be good to hear your opinion on how valuable things like that are.