14 December 2008

Select Language Please...

zomg, blogging from work.

The Google Adwords debate. Various commentators have criticized, and Google has all but admitted, that it controls ad pricing via its 'Ad Quality' metric. The cost of an ad is directly related to the Quality score assigned to it by Google, or more specifically the bid and quality score of the next lower-ranked ad. Different websites, including Google explain how the system works, and point out either the positive or negative effects of this system. The positive effect is that it means ads have more relevance. The negative effects are more complicated...

Ads are placed and bidded on in different categories. Keyword searches are very different from targeted placement on 'content' websites, which is what many people associate when thinking of the distinctive 'Google AdWord' program. Cost-per-Impression bids too are different from Cost-per-Click. So what's exploitable?

Ad Quality for CPC bids depends, among other things, on click-thru rate (both the individual ad and historical trends for the advertiser's entire account). This encourages ads that are relevant to the marketed content site... but it also encourages ads which only SEEM to be relevant. Clickthru rate only measures appearance, not how that appearance translates to sales, but Google's Ad Quality program means that deceptive appearance may translate directly to lower costs for the advertiser. This is one reason for criticism of Ad Words.

The other metric, used for both CPC and CPI, is 'landing page quality'. This is probably a subjective quality, unless it is measured by network links, but even then it's subject to mistakes and flaws in Google's link-measuring algorithms, where mistakes by Google lead to discriminatory higher costs against affected innocent advertisers. This is another reason for criticism of Ad Words.

However the real problem is not that a relationship exists between Google's algorithms and the number of customers that buy products on an advertiser's website; it is that it is a direct relationship that can vary instantly based on Google's actions, where a random variation in an advertiser's Ad Quality can immediately lead to a higher Effective Cost-Per-Click or Per-Impression for the advertiser. What it should be is a reduction in impressions, because an 'irrelevant' ad should not be seen so much. You might think that since Ad Quality determines ad ranking and only a limited number of ads can be shown on each page, that Ad Quality would translate directly to fewer impressions, but in fact this often isn't true because the system is based off of budget saturation of higher-ranked advertisers, which is why Google characterizes a better Ad Quality as directly translating to lower costs per click.

((Ugh... doing something else for a while never helps logical coherence of blog posts...)) The ideal solution, of Ad Quality reducing impression rate and only impression rate, is related to the last problem of the size of market demand for the advertisers' products. The CPC and CPI bid metrics present a simple and understandable approach to the cost of advertising, when the market size is unknown or unreliable; but it leads to these problems of Google being perceived as directly controlling the return on advertising expenditures based on 'arbitrary' quality measurements, which are supposed to prevent exploits in the system but require advertisers to trust Google even while the system will never be without critique or possess the transparency that many would like.

Basically, Google ((I had Blizzard there, and didn't realize until I finished this paragraph... >_< zomg and in the last paragraph too!! -deletes it-)) attempts to use the 'uncertainty of form', advocated by Sun Tzu and the same strategy which I attempt to use myself to reach or.. maintain my existence, in their case used to attain an optimal operating position due to the compromises which they would otherwise have to endure operating in a world or environment of stupid people... but they do not do it perfectly, because there are still critics.

/sigh, what was I saying... too many Blizzards, now I don't know what I'm talking about and this is why I'll never have a serious blog lol.

Summary: Google's Ad Quality directly influences ad cost, and indirectly influences impressions/clicks because it encourages advertisers to use up all of their daily and monthly Budget via offering fuzzier ad matching options. What Google's Ad Quality SHOULD do is directly influence impressions/clicks at the same cost per impression/click, and INDIRECTLY influence cost when this encourages advertisers to raise their bid to increase their Ad Rank and indirectly the number of impressions.

The reason this can't happen under the current system is that it requires the number of impressions, or number of clicks advertisers get to be slightly fluid. Not in the sense of varying hugely based on Google's assigned Ad Quality rating, but rather on the slow and usually imperceptible currents based on market changes and on varying increases and decreases in competition due to overall average Ad Quality in the entire advertising market. I'm certainly no expert, but I would imagine this would mean transitioning from a system of variable $X per constant Y results, to a system of constant $X per variable Y results. The point that must be conveyed is simply that the variation in clicks or impressions is the same for all customers, regardless of Ad Quality or keywords or content placement of an ad. Provide more information about the expected effects of a certain expenditure for a website with a quality product and landing page, as well, to help new customers comprehend it all (a key complaint!).

It does still penalize newcomers to the system, because confidence in the system is dependent on it not collapsing from not being able to find sufficient placement in content and relevant search results... but that is as much as may be, and removing the perception of arbitrary cost penalties by Google is paramount, and this is the only way to do it.

No comments:

Post a Comment