News

Google Leaks Quality Score Variables (Pscore, mCPC and thresh) in Search Results

Earlier this morning a colleague of mine was conducting a search and called me over to show me something interesting. Immediately below each of the sponsored search results (AdWords) were three separate variable names and values. The numbers that have been shared publicly, and documented here on this post include those for major corporations including Honda Motor Company and Hewlett-Packard.

    The advertiser values shown include:

  • Pscore
  • mCPC
  • thresh

Here are screenshots of each ad and their corresponding stats…

honda011 Google Leaks Quality Score Variables (Pscore, mCPC and thresh) in Search Results

honda021 Google Leaks Quality Score Variables (Pscore, mCPC and thresh) in Search Results

I’ve also included a Full Screen Shot of Ads and their Values for review (opens in new window).

The next step for me of course was to try to search for these variables. When I conducted a search for “mCPC” I saw the following example again for an advertiser’s stats being shared:

mcpcad Google Leaks Quality Score Variables (Pscore, mCPC and thresh) in Search Results

My initial thoughts here are that mCPC is the maximum cost per click for an ad. For example, the first ad shown appeared to be from Honda Motor Company. The second AdWords ad was for a small, lead generation web site in the automotive industry. My suggestion here is that with deeper pockets, Honda is in a position to maintain a higher maximum cost per click — which appears to be $4.71 compared to the competing CPC ceiling of $1.09.

Additionally, when researching Pscore, I found a document published by the European Bioinformatics Institute that may be related. That document states:

P-score represents minus logarithm of the P-value. P-value measures the probability of achieving the same or better quality of match at a chance, i.e. at random picking the structures from the database. “Quality of match” is a complex characteristics, which accounts for RMSD, number of aligned residues Nalgn, number of gaps Ngaps, number of matched Secondary Structure Elements and the SSE match score. The higher P-score (the lower P-value), the more surprising, or statistically significant, is the match.

Either way, this is a very interesting leak of data. I want to give thanks to Paul from Search Marketing Corporation for sharing this initial find with me as well. Please share your thoughts on this via the comments area below.

You Might Also Like

Comments are closed.

22 thoughts on “Google Leaks Quality Score Variables (Pscore, mCPC and thresh) in Search Results

  1. Its really strange!!! And Kudos to both you guys for such a lightning fast find and proving us with screen shots. Expecting a comment from “Google” himself!!!

  2. Searching for “pscore” isn’t going to tell you anything. I’ve written dozens of programs through the years that have used that exact variable name and many others like it. No one is going to publish a paper on what a variable name means, so anything you find on the Web is irrelevant to whatever those little bits of data mean.

    This is a whole lot of kerfluffle over nothing.

  3. How about “P” equals Placement. Based on all other things being equal.

    A view of the whole screenshot with all the scores would help to decipher the various parts of the code.

  4. “This is a whole lot of kerfluffle over nothing.”

    One man’s opinion. I know you’re fixing that on the pScore element, but as an advertiser, would you want your maximum cost per click public domain? If your competitors used that information against you, would you not feel as though you were at a disadvantage?

    Chances are it’s only kerfluffle when it’s not your money being wasted.

    This of course, is this man’s opinion.

  5. I remember a while ago something similar or the same being noticed by someone. I forget where, but in that example it was framed as being a tool for google’s adwords sales team.

  6. Delicious find! But it’s probably more useful in comforting our web developers that even Google makes gigantic slip-ups every once in a while. There’s not enough data to parse through…

  7. Scott Fish: “I remember a while ago something similar or the same being noticed by someone. ”

    Michael: There have been a couple such code burps in the past. The SEO community does a pretty bad job of analyzing them.

    Eric is right to point out that advertisers may be concerned about whether their maximum bids have been disclosed but no one has shown that is what the “mCPC” value refers to. Nor has anyone shown that the data is associated with the specific ad under which it appears.

    Code burps are unpredictable and generally only make sense to the programmers who wrote the burping code (and sometimes it takes the programmer a while to figure out what is going on).

    Eric has good company with his speculation (Barry Scwartz and Jordan McCollum) but it’s just speculation and isn’t really constructive.

  8. Can we get someone to decipher this:

    “RMSD, number of aligned residues Nalgn, number of gaps Ngaps, number of matched Secondary Structure Elements and the SSE match score. The higher P-score (the lower P-value), the more surprising, or statistically significant, is the match.”

    @Michael – If you don’t think it’s significant, no need to comment and be a downer for those people who do, a second time. You’ve made your point initially. You just discredit yourself by playing devil’s advocate and ignoring the balance of probabilities with something like:

    “no one has shown that is what the “mCPC” value refers to. Nor has anyone shown that the data is associated with the specific ad under which it appears.”

    CPC is an industry wide acronym. What are the odds it means something besides max CPC? At worst it means minimum CPC, but in the context above, that’s highly unlikely.

    And while your point on pscore may turn out to be accurate, Eric has found a pretty likely match for its meaning. From what I could understand of the biotech folks’ jargon, it means ‘how relevant the ad/lander are to the keywords’ .

  9. “would you want your maximum cost per click public domain?”

    I remember a time when bid amounts where public and shown next to each ad. Rest in peace good ol’ GoTo.com hehehe.

    Hey Michael, why are you always so negative? :)

    You are right, this thing in itself is not moving Earth a single inch and no “sensation”. However, it could be a piece of information that is the only thing missing for somebody else who also has a small and limited amount of incomplete data to make sense out of it.

    Don’t tell me that it never happened to you that you found on the Internet a tiny and seemingly irrelevant peace of information, that is maybe not even about the subject really what you were looking for, which was exactly what you needed to make a break through for a problem that kept your brain cells smoking for some time already without being able to make sense out of things?

    Puh, long question. It became a test in itself. :)

    Cheers and thanks for posting this information Eric!

    Carsten

  10. Very Interesting. I would have loved to see the Google results when it actually happened ! It would be great if you can post a screen shot of all the ads.

    I’d love to study these values for various ads on different positions to find out which variable among the three made the biggest difference in ad position! Anyways great info mate.

  11. Wow this is so strange! I wonder if anyone has seen this at all? I do remember the good old days of GoTo and Overture…bid jamming used to be so much fun! :o)