AdWords: Quality Score Discussion

[tonynoriega]

I hear all the talk about this, but am i missing something?

Is this "Quality Score" viewable? Can i figure out what my score is somewhere...?

Or is this some mytholocial algorithm, that everyone claims to know, but really doesnt?

thanks

[sailorjwd]

Sounds like G wants everyone to have a handful of campaigns and a handful of open match keywords bidding several dollars each.

Certainly would cut down on their overhead.

Its like msn - i give them 5,000 keywords from an adwords account and they approve 100 of them. Maybe that is telling me something.

[tonynoriega]

inflation to keep up with all of their hardware. i just had to increase a daily two fold! just to keep up with the rest of the junk thats out there.....

5 ad groups and just about every keyword is in the range of $4-5 cpc....

i have to think of this as any other form of marketing....the better the service, the more they can charge....just have to suck it up and realize thats just how its going to be....

although i will update my landing page and do all the little fun google tricks and let everyone know if i see any results....

probobly not...but we will see..

[RonnieG]

Here is why I think too many "long tail" keywords may actually hurt QS: How many of those long tail strings will actually exist on your landing page content and description tag, in exactly the same word sequence? If they do, are the naturally constructed? By that, I mean with natural wording before and after. Or, are they simply free-standing keyword strings, not really connected to any of the other content?

I have found that a middle road approach has been working pretty well for my small number of personal campaigns, and for the campaigns I manage for a small handful of clients, all in the same vertical but different geo markets. I use relatively short phrase match keywords, with 3 to 5 core keywords in a few different sequences. I think of this as the "body" of my keywords, and it allows a "head" and/or "tail", as the consumer may wish. Just what phrase match was designed to do, eh? In each ad group, I also include a couple of broad match keywords, bid at lower than any of the phrase match strings, just to catch any odd combinations a user may enter.

Then, I include each of these "body" keyword strings on the landing page in naturally constructed sentences. In the ad, every word that is in any of the phrase match strings occurs at least once in the ad, with the phrase that is historically getting the top number of impressions in the headline, and the main product word occurring twice, once in the headline and again in the ad text. The main product word also occurs in the display and landing page urls, as a third level domain, with a 301 redirect to the specific landing page of the site, which also has the main product word in its folder and/or file name. This third level domain approach may not be practical for huge campaigns, but for my 30 or so main product words, it is manageable and works well. And, the main product word also occurs in the landing page's title and description tags, with the ad's title line also in the description tag.

I still have relatively granular bid control on each of the phrase match strings, which I would not have with broad matches. The ad text contains every word that occurs in the keyword strings. The corresponding landing page includes every phrase match string variation in the keyword list, all in a naturally written way. In this approach, nothing should be perceived as contrived, masked or fragmented. Everything comes together in a natural, user friendly way. To me, this is simply natural quality and relevance, and what I would look for, as a user, or as a G AdWords algo programmer.

The results since I have been doing this, since early this year, and for each of the geo markets that I manage campaigns in, have been very stable keyword bids and performance, with slightly lower CPC over time as the new ads and keyword phrases gained history. My CTR has improved significantly, and conversions are also up significantly. I also use AdWords conversion tracking, and have NOT seen any impact on minimum bids or CPC costs from that, one way or another, except that it allows me to better tune my ads and specific keywords for better conversion performance.

[QualityNonsense]

Interestingly, I've recently been told by AdWords support that increasing your daily budget can help get inactive keywords active.

I'd love to hear how increasing your daily budget improves the quality of user experience...

[sailorjwd]

Noticed something interesting about the QS today.

I searched for: google adwords advertising.

Saw Googles adwords ad in the top slot for this phrase.

Copied the short cut and went to the landing page.

Notice that the two phrases in the advertisment appear nowhere on the page - 'get a better roi from your website' and 'start gaining new customers today'.

Also - talk about low-content landing page - there are less than 80 words on the page.

Two of the three images do not have alt tags and the one with alt tags has a paragraph of info (stuffing).

Something stinks here.

I bet if I copied the look at feel of the page and changed to topic from advertising to software development that my QS would be horrible.

Nothing about their guarantees or 'return policy', costs.

Anyone else agree with me on google bypassing quality landing page requirement?

[aeiouy]

Might be an interesting expirement to test out. Just do the same thing with same copy and replace it with another topic.

[xdude]

Well according to me, the adwords algo is not so intelligent and can never be. It has to be spoonfed. And when it can be spoonfed, just expect the not so legitimate sites to be spoonfeeding it too.

Ultimately it will be spoonfed so much crap, it will stop relying on everyones spoonfeeding. It will just go into a corner, sulk and then decide that it wont assign any "trust" value to any new site till they have proven their worth. Sounds familiar?

Couldn't agree more. One can only inject that much intelligence into an algorithm. Even with leading edge AI, Semantic Web, etc, a score rating algorithm is still a set of formulas fed with weighed parameters. With patience and computing resource, people can always trial-and-error or reverse engineer to a set of criteria to satisfy the formulas enough, if not fully, to be in the game.

Even the addition of manual review is not going to guarantee good quality assessment as human judgement can also be biased. More importantly, the guidelines these human reviewers are using are only as good as what the underlying principles of the guidelines allow. Bear in mind these guidelines are themselves products of Google's keyword ad business strategy.

For example, if Google determines to blur the line in between content-oriented search listings and sales-driven ads, the guidelines will have to be tweaked accordingly to play up content. In essence, QS could be as objective or subjective as a formula created by any John Doe with certain agenda in his head.

[Great White Shark]

Hi, Haven't seen any new postings for awhile now for this topic.

Question: If google is discouraging exact match, then why is it still in existance?

Also, why would they recommend using match options in their guidelines?

Cheers

[QualityNonsense]

Interesting post here regarding a big UK AdWords advertiser's conference call with Google on landing page quality.

[edited by: eWhisper at 7:02 pm (utc) on Dec. 11, 2006]
[edit reason] Please no blog links. See TOS. [/edit]

[tonynoriega]

so this big deal with exact matching.....i just dont understand why and google rep would recommend against it? you would think that they would prefer exact matching to get a better CTR?

here is my question regarding keywords and the QS...

if i take my ad groups that have about 20 keywords in them, and chop that down to a more precise set of the best performing keywords, say....5..or so...

thus, most likely improving my CTR, would that affect my QS?

for example, if my keywords are A, B, C, D, E, F, G....and i really only do well for A, B, C,.....and my landing page only really talks relevantly about ABC....

am i on track to boost my QS?...and keeping ABC "exact match" keywords?

[pdivi]

My take is that exact match doesn't give Google the same ability to build a profit-maximizing set of ads as broad match. If I have 50 advertisers bidding on (broad match) widgets, I have total flexibility in arranging them however I like in phrases like blue widgets, red widgets, Kansas City widgets, ... I can arrange them using bid/CTR data in a way that brings the most return by phrase.

Exact match restricts Google to certain advertisers and certain bids by KW, which is bound to suppress revenue somewhat. That doesn't answer any questions about QS...it just explains why G might like broad match.

[tonynoriega]

so its their own way of peddling their product to benefit them.....gotcha..

im sure they could use the extra (*cough*) revenue...

so that in turn, benefits them, but not people like me who dont want all the unnecessary clicks..?

[shorebreak]

There was a great summary article on QS written by David Andrews and posted on ReveNews (I won't link to it 'cause of WebmasterWorld TOS but you can find it easily enough on their site). One key snippet:

"DETERMINANTS OF QUALITY SCORE
The determinants of a Quality Score include Landing Page Quality, Ad Copy Relevancy, CTR on Google, as well as the CTR History based on an Exponential Moving Average. An exponential moving average places more emphasis on recent history or CTR data than long term CTR data which means that your Quality Score will be impacted much more quickly than it would have been if a standard Simple Moving Average was used."

We use exponential moving averages as well for both our click and revenue recency models. The question we need to find out is *what half-life does Google use for its CTR history Exponential Moving Average*?

Anyone have any thoughts?

-Shorebreak

[magicdan]

is CTR history the only exponential average though, If I were google I would also have that kind of recent trend average used on the landing page CTR back to google results too. Its a good way of stopping previously good ads and landing pages being switched into bad ones slyly. I reakon the amount of time used comes down to the deviation from that average so if it abrubtly changes for the worse strongly then you invite more heat quicker. It has to be at least a 3 - 7 day check on those key indicators perhaps.

[glenv]

I saw elsewhere there is a theory that Google only re-evaluates landing page score once amonth. Any verification of this? Any other ideas? I would think if you completely re-wrote all ads it would trigger it but now I am not sure.

[tonynoriega]

Well, i just revised almost all of my Adgroups over this past month. 9 out of 10 of them actually...i used keywords on the landing page to match my Adtext, totally revised the landing pages layout, content, added some drop downs, images, more links to information...etc.

but i have no idea if my page has been re-evaluated, becuase i can not find any measurement that would reflect an adjusted QS....

this is why i find the whole scheme of QS so bogus....especially without any method of measuring success.

Plus i had to boost all of my Max CPC's to keep up, so i really have no clue as to wheter or not my adjustments to the landing page had any affect....becuase of course my money sure will boost my Avg. Pos. but i will never really know if my adjustments had any affect...

didnt really answer the question, but i heard that as well....

[davidzhawk]

Good article find Shorebreak. What I find most interesting about the claim that there are "multiple quality scores" is that this factor basically makes it impossible to predict position and optimal bid. Even if you have information on Google's exponential moving average (EMA), the EMA by itself is not enough to really achieve perfect information (as was available in the old GoTo model and to a lesser extent in the pre-QS AdWords model).

Basically, at the end of the day, you are going to need to apply a Bayesian analysis to estimate your bid. Thus, this basically makes the Google system even more "blind" than it already was. And blind auctions always lead to one thing - inefficiency, which leads to increased bids and more revenue for the search engine.