Evil Google Takes Marketer's Money

I know half the money I spend on advertising is wasted, but I don't know which half.
-- Marketer's lament

If you are advertising with Google, I know which half, and so does Google. There are some easy fixes they could do. But they are rewarded financially to not fix them, and take junk clicks at your expense. This blog is an attempt to make them move in a useful way for their customers.

Our company runs a significant (well, by our standards) Google campaign. We have a limited budget and we constantly fight to control our costs. Google allows you to specify what keywords should trigger a ad, right? You can even specify keywords and phrases you don't want to see. And that should be enough to narrow your ads to your targeted audience.

Wrong.

The problem is that Google will match your request to any search query containing complete nonsense, with a part that contains a "match" (exact or broad). And it is amazing how much trash there is; check CampaignManagement/keywords/details/SEARCHTERMS/All. Surely you have looked and seen the trash. The problem is that you are paying for clicks on this trash, and Google loves this.

Trash Google will match

Here are some trash entries we used to get (we're in the software business):

yieldmanager khttp dumb test net keyword: "test net"
migrate segy keyword: migration software
legacy by vrrjinia woolf dowenloded keyword: legacy software
script coding php&css menu tree keyword: script parse tree
http dsm odnoklassniki ru getimage smileid 3db19bc76c keyword: dms
subaru legacy l 96 specs keyword: legacy software
how i can check toyota avensis fault code 06 diesel keyword: check code
www schuelervz net profile qfomc9pqtkmlphp44b2hrekd_be3zlmpr5gdz4agb2u tid 102 keyword: net profiler
5130 mobile code format keyword: format code
free nokia 6230i security code retrieval keyword: security code check
download chea engine 5.6 1 keyword: search engine
cheat enggine 5.5 keyword: search engine
chet enggine keyword: search engine
cheatengines4 keyword: search engine
search key 7261 b76d 1b25 eb40 a200 206a keyword: search engine

How to eliminate this trash

We've managed to eliminate most of this. Here's some advice on how to avoid much of it.

First, avoid broad match like the plague; Google will match the most astonishing misspelling of your keywords in an effort to get clicks. (Broad match: "search engine" matches every possible spelling that you can imagine of "cheat engine" loved by gamers. There's a billion of those guys, and they're all 14 years old.)

Second, use the Shared Library, Campaign Negative Keywords, and put words in these to suppress pointless queries against your campaigns. I've set up the following negative keyword lists, and basically filter all our campaigns against all of them:

Bad Actors people that want to cheat you somehow, spammers, etc.
Bad Spellers people that cannot spell what they want. Probably not a customer unless you are selling shoes.
Cheapskates people that want "free"; probably not a source of revenue.
Employment people that want to work (selling shoes), not buy.
Geography people that specify someplace far away from you. Shoes in China aren't your market.
Proper Names You aren't going to sell something to somebody looking for John's shoes
Not your stuff If you sell shoes, people interested in horse shoes aren't your clients.
Numbers Queries containing random numbers are likely some machine address or other irrelevant technical junk
Porn/Gutter words I'm not interested in clients that are typing in swear words.
Students These guys search for everything("lectures on shoes") but don't have cash.
Stupid Internet locations Somebody fishing at Alibaba.com probably isn't your customer

We have lots of stuff in each of these keyword lists, but its pretty easy to figure out what you should put in each list by looking at your trash entries. If you do this, it is amazing how much trash disappears. You'll note that none of this is specific to our business; these should work pretty well for you.

The Evil: How Google makes this far too hard

To Google's credit, they offer you the ability to set these up. But for those of you that aren't Google adwords experts, it might not be obvious that you should set them up. It sure wasn't for me, and we paid Google a lot of click-money before I started to figure this out; then it cost me hundreds of hours to fill out the lists. I'm pretty unhappy about this; not only did we pay for wasted clicks, but it cost a lot of personal time I would rather have had to myself. That's where Google IMHO is evil. They are happy for you to not know this, because of the extra clicks they get and you pay for.

What is needed

Since these categories are pretty general, Google should be offering them as standard categories. Google has really smart people, and they are classifying the world. They can trivially offer categories such as "Countries", "City Names", "Geographic Features", "Fancy Buildings". They should have categories for every noun in the Oxford Dictionary; I suspect they already have these categories supporting their "semantic search". Why should I have to invent lists for all this by myself?

Even with such lists, you need ways to combine lists in various ways (this is just like in software!). You may want to eliminate all countries except "Algeria" and "France". Fine, Google should let you make your own lists (MyCountries: "Algeria" and "France"), use lists that are categories (e.g., Countries), and combine them in arbitrary ways, e.g., Country minus MyCountries. This is Information Retreival 101. Otherwise you get to build this list one entry at a time. How long will it take you to eliminate Burkina-Faso?

There are some serious problems with filling in list with just (negative) keywords. Take "numbers" for example; how do you get rid of them? Well, duh... you can make a list: 1, 2, 3... I made a lot of progress by filling in our Numbers list with the numbers from 1 to 1000, entered painfully by hand. (Big dip in trash entries when I did that). But I can't eliminate 6230i this way.

What is needed is the standard computer science concept of regular expressions. [The Google guys are computer science geeks; they know exactly what this is]. Then I can eliminate all the numbers I don't like with a trivial regular expression "[0-9]+", and get rid of things like "6230i" by a simple extension: "[a-zA-z]*[0-9]+[a-zA-Z]*". As a marketer, you may find these a little hard to understand at first, but if this saved you thousands of dollars and lots of time, surely you'd be happy to learn it. But first, Google has to offer it. I have asked Google Adwords people for several years for this, to no avail, which is why I wrote this diatribe. Having regular expressions for positive keywords would be really helpful, too.

Finally, they need a category for bad spellers. People that can't spell are largely those with less cash. Harsh judgement perhaps, but it matters if you are looking for purchasers. I've filled our "Bad Spellers" category with the various ways I've seen our technical terms mispelled. This helps, but this is an endless task; how many ways can "cheat engine" be mispelled? Google knows what words are in the Oxford dictionary. Everything else is mispelled.

Specific features that would remove the Evil

In short, Google should offer as standard features:

Standard categories of nouns ("shoe") containing instances of those nouns ("boot", "sandal", ...)
The notion of "misspelled" as applied to any standard category (including the category "OxfordDictionary")
The ability for Adwords users to define arbitrarily many of their own categories (sort of there already, but not really general)
The ability for users to define entries in their own categories with powerful regular expressions
The ability to combine user or standard categories in arbitrary ways (union, intersection, difference, yes, even complement) to form new categories
The ability to apply regular expressions and other filters to Display Network Placements (e.g., filter out all websites containing "how" in the name, or are vampire sites feeding off mispellings of sites such as "speedtest.com").

(I'm sure Google can improve these ideas with a great category list algebra). These would be useful in forming both positive and negative queries. These can be added to their present "simple" keyword scheme, so nobody loses anything and nobody has to learn anything (to their loss).

Google knows how to do all of this. Until they fix it, I think they are doing the very evil they claim they are not. Yes, it will hurt their profits some (some Google manager is going to turn white thinking about this; I have no sympathy for him). But I'll feel much better about how my online marketing money is spent, and I won't think they are two-faced about it.

If Google won't do these things, maybe Microsoft will figure this out and put these features into Bing. It would sure raise my interest in Bing.