new generation of CPC tools
Efficient Frontier has applied Wall Street analytics to the management of a pay per click (PPC) advertising campaign.
Essentially, their Portfolio Maximization technology incorporates portfolio management techniques that were first developed on Wall Street to manage large stock portfolios.
I think it is a very clever way to handle cost per clicks campaigns - focus on maximising the volume of traffic for a pre-defined cost, introducing Return on Investment at portfolio level (or campaign level in Google for instance).
A usual granular approach to keywords is to assign a ROI value to this keyword. If the ROI is positive for this keyword, we increase the bid until reaching position 3, and maximise volume (or up to position 1 if this increases the volume significantly).
The portfolio approach doesn't remove granularity, but let us potentially "loose" money on some keywords, while keeping the overall cost per acquisition under the cap we have set. You end up spending the same amount of money for higher traffic.
This approach is not really instinctive, and cannot be really applied as easily as the first approach, which can be carried out manually. With the portfolio approach, you need to have calculations in the background providing you with the best combination. And that's where using a management software really makes sense.
It's all good for efrontier.com, which obviously maximise clients spending, and therefore increase their revenue (charge a percentage of your spending). A real killer revenue modele.
posted by Emmanuel Idé @ 10:09 AM
6 comments
6 Comments:
Thanks for the write up! That's a great overview of our technology.
As I work for EF, I should add that our clients typically see a 20-50% uplift in their ROI using portfolio maximisation. And of course, we're the only company in the world that offers this approach.
But enough plugs for now - great post.
Thank you for your comment Jonathan - I think the approach is really a good one, but I don't think it would be too difficult to replicate, and would expect other companies to catch up soon.
We're already experiencing other SEMs in the US talking about portfolio management, but none of them are actually doing it. Although the principle of portfolio management is simple, performing it at scale for millions of keywords is a serious challenge.
Interesting conversation. The link you provided doesn't really give a taste of the challenge.
If I look at the patent EF has filled in, there is not much there, except describing what is said on the side, with just a few more techy words, to make it more technical, less marketing (btw: with all respect, if a patent like this gets through, I'll definitely try to get one filled myself! I thought about applying quantum physics to SEM, I have got a PhD in quantum physics..or the French equivalent, this should give me some credibility).
Seriously, I would be delighted if you could share with me the main challenges you are facing in terms of data processing.
I'm not on the technical side, but as I understand it the scalability issues come from the number of calculations you need to perform for each keyword bid. So if the system bids a particular keyword up, it has to look at all the other keywords to calculate whether the investment could be spent better somewhere else. So if you have 10,000 keywords and you're bidding over, say, the top 5 positions, you have 5^10,000 calculations to make.
So tell me, how does quantum physics apply to search? I've always wondered whether thermodynamics could apply to search, but I only have a BSc in Physics, so I'll defer to a PhD.
ahah, don't know Jonathan, but could probably cook something up!
van Rijsbergen’s book "The Geometry of Information Retrieval" would be a good book to read for a start. Provides with some suggestions about how to use quantum mechanics applied to IR. And once all of this is digested, should not be too challenging to patent a way to compute the high volume of computations you mention here.
Post a Comment
Links to this post:
Create a Link
<< Home