Alphabet Google Falls on Its Algorithms

March 24, 2017

Here in Harrod’s Creek, advertising is mostly hand painted signs nailed to telephone poles in front of trailer parks.

Real Advertising in Big Cities Does This

In the LED illuminated big cities, people advertise by:

  1. Cooking up some keywords that are used to locate products and services like mesothelioma or cheap tickets
  2. Paying money to the “do no evil” outfit Alphabet Google to put those ads in front of people who are searching (sometimes cluelessly) for a topic related to lung disease or flying to the land of milk and honey for a couple of hundred bucks
  3. Alphabet Google putting the ads in front of humans (or software robots as the case may be) who will click on the displayed message, banner, or video snippet
  4. The GOOG collects the money
  5. The advertiser gets leads
  6. Repeat the process.

The notion, like digital currencies, is based on trust. Advertisers trust or “believe” that the GOOG’s smart software will recognize a search for Madrid will require an airplane ticket and maybe a hotel. The GOOG’s smart software consults the ads germane to travel and displays a relevant ad in front of the human (or software robot as the case may be).

goofed for content

What happens when the GOOG’s smart software does everything except the relevance part?

The reaction in the non Sillycon Valley business world is easy to spot; for example, here are some examples of the consequences of the reality of what the GOOG does versus what advertisers and other true believers in the gospel of Google collides with faith, trust, and hope:

I could list more stories about this sudden discovery that matching ads to queries is not exactly what some people have believed.

A n-Dimensional Problem

There are several dimensions to this Googley n-cube.

First, Google was constructed on the concept of “pay to play.” For those interested in ancient history, the idea originated at GoTo.com, evolved into Overture.com, and ended up at Yahoo. The pre IPO Google paid some money to Yahoo to avoid legal acrobatics. Allegedly the GOOG’s engineers were inspired by matching ads to queries. Yahoo thought the inspiration warranted consideration. How did a billion dollars sound? Pretty good to the Yahooligans who watched the GoTo.com model produce the fountain of cash that the innovators at Google trotted out as their own Eureka! moment. But did pay to play work? Yes, it did work when the words in the ad directly matched the words in the query. The GOOG moved in a new direction. Do you remember Oingo, later renamed Applied Semantics? No, well, the Oingo folks figured out how to automate synonym expansion. Stated simply, Oingo could take a query for Madrid hotels and expand it to embrace air travel and rental cars. The magic was that ads without the key word Madrid or hotel in them could be matched to the ads “related” to the concept of Madrid. Everything worked pretty well as long as the GOOG stayed in the text business. Sure, there were some detours around drug ads and others which some folks found in questionable taste. But for the most part, the GoTo.com model as expanded with Oingo type expansion delivered traffic.

Second, Google bought YouTube and entered the video business which is a pretty expensive proposition when anyone can upload videos on any subject whatsoever. Now the question arose, “How can YouTube spin money?” The answer did not require a rocket scientist to come up with, “Sell ads.” Now videos, unlike text, are a bit trickier to figure out. There are images. in my cyberOSINT  lectures, I show an image with a person holding up a piece of cardboard with the url of a questionable Web site. Google looks at the index terms the user assigns to the video of the message and parses the written additional description. The problem is that the words can say “pizza” and the message on the piece of cardboard something else. Google’s smart software is clueless. (Now keep in mind that unless humans process certain types of content, smart indexing does not work particularly well. When humans get involved, costs skyrocket. Google wants to keep costs down, so smart software is given the work.) The problem is that the matching can and does misfire. The cognitive dissonance generated when a video about a controversial topics appears with the ad from a blue chip company. Google’s smart software is unable to deal with the “problem.” For this stumper, rocket scientists are indeed needed, but time and money must be applied. When the expensive YouTube service is losing inflows of ad money, the word “crisis” may be an understatement. As video becomes more important, the Google desktop ad model is under pressure from mobile devices like Google’s own Pixel: Smaller screen, fewer ads. At this moment, the YouTube cost overhang could be a fascinating accounting challenge. Strike “could”. Let’s go with “is.”

Third, Google has emerged from the ashes of the free Web search engine business in the mid 1990s as the technological and financial model for information access. I referenced the importance of what Google taught a generation of innovators to do: Recycle, repackage, and sell access to traffic and sell data in a variety of ways. This is the Google “legacy” I wrote about in my first Google monograph, funded in part by assorted commercial outfits who wanted to know what the Google was and could do. You can see a video summarizing some of this research on YouTube no less at this link. If you want to obtain a copy of the now out of print monograph, write benkent2020 at yahoo dot com. The Google Legacy monograph explains that the GOOG was not like Archimedes, Euler, or Leonardo da Vinci. Google was more like Marconi, Robert Fulton, or Thomas Edison (each of these inventors allegedly borrowed their big idea from another person). The PR miasma swirling around the GOOG obfuscates how the company’s technology works. If you want traffic, do search engine optimization. But, one Googler told me, if you want clicks, buy AdWords. This Googler has since gone to another employer. The Google was in the right place when free Web search engines were chasing the mythical portal to the Internet idea. The GOOG focused on the GoTo.com model and the rest is history. Does the GOOG deliver relevant results? In the Google Legacy I pointed out that for certain queries, Google delivered results which were on point with the user’s keywords. When I wrote Google Version 2 a couple of years later, the relevance of the Google system was pulling in more off point results. The reason? Query expansion. Broader queries mean that ads less and less directly related to the user’s query can be displayed on those boat anchor desktop computers with screens many times larger than on most mobile devices. If we flash word to the present, the present “crisis” is directly related to these business realities:

money

  1. Google has lots of advertisers and has to find a way to display as many ads as possible in order to maintain its revenue
  2. Relaxed queries allow increasingly off point ads to be displayed as evidenced by the grousing about blue chip companies “sponsoring” hate speech
  3. Google’s costs are a very big and growing problem. The death blow to the early Web search engines was the ever escalating, essentially very hard to control costs of infrastructure, hardware, software, and bandwidth
  4. The shift from big boat anchor Web access to miniscule mobile screens reduces significantly how many ads can be displayed.
  5. The fix is to further relax what ad appears where.

If the dimensionality of the problem increases, the complexity of the challenge facing the GOOG becomes similar to repairing a failed subsystem on an F-35 when the aircraft is engaged in an operational mission on the edge of the fuel envelope. Ejection may work for some Googlers. For others, who knows?

What’s the Fix?

Frankly I don’t have an answer to this question. I don’t think the GOOG does either. In the short term, the GOOG will have humans dive into the matching thresholds and spot check what the smart software is doing. But that’s expensive, and it is not clear if today’s crop of Googlers know enough about the jerry rigged relevance and ad matching subsystems to resolve the problem.

The long-term fix is to get back to basics, but on this sunny day in Harrod’s Creek that seems to be a wild and crazy idea. For the mid-term, I assume the GOOG’s legal eagles and apologists will step up their efforts.

Check out the Google Legacy videos. Better yet, write benkent2020 @ yahoo dot com and buy an old analysis of the GOOG, which is quite surprisingly more relevant today than ever.

Stephen E Arnold, March 24, 2017

Comments

One Response to “Alphabet Google Falls on Its Algorithms”

  1. Steve Carr on March 25th, 2017 4:57 am

    The future is alternative search engines. We all need to us another search engine and than we take away the governments power, who have become to powerful, or we just go back to yelling loud try Lookseek com a no tracking search or one of the other alternative searches. Have a good day

  • Archives

  • Recent Posts

  • Meta