I'm constantly creating new campaigns for different businesses, and constantly in touch with people who do the same. Many times, I see a lot of money being wasted just because the client wants to "be everywhere", or the account manager wants more fun options in doing their AdWords campaign. In many cases the problem is not having a good business focus, and over complicating the campaign in the process.

Here are some tips that I think can help AdWords account managers with their search campaigns, and it can be fun if you have a good start, especially when the maintenance work becomes significantly less than the people who start clueless.

• Always use [exact match]: this match type gives you maximum control, because you know exactly which keyword you are bidding on, so you don't have any bad surprises with match type. But the most important thing is that since it is the most specific, and therefore the most relevant to your users, you are more likely to get a better quality score, and therefore lower cost-per-click. This is especially crucial at the beginning of you campaigns, because the account history plays a role in determining the quality score of your new campaigns. Ideally, you should have all your keywords in exact match, but this is practically impossible, because you cannot possibly capture all the different variations that people can think of. And because you cannot capture all the different possibilities of keywords that your users might use. This is why you should utilize other match types.
• "Phrase match" to discover new possibilities: this match type gives you less control and specificity. It helps you get much more reach than exact match, and more importantly it lets you discover new keywords that you didn't have in mind. This discovery happens when, after having accumulated some clicks, you can go to the keywords tab and click on the "see search terms" button and see which keywords users actually used to trigger your ads. You can do several things in this case:

1. Use -negatives: you will definitely see some irrelevant or bad keywords that you didn't plan for, so you can add these as negative keywords to refine your keyword selection. For example, you might find that users use 'cheap' together with your brand products. If you are not a discount store you can use negative match type for 'cheap' 'discount' or 'free', to make sure you only target the right audience.
2. Use [Exacts]: the top keywords that were searched for should be added as exact match. Learn an easy way to know where to draw the line when segmenting your numbers.
3. Create new ad groups: With some of the keywords that were triggering your phrase match keywords it would make sense to create a special ad group for them, or move them to a different one, because they are more relevant in another place. For example, if you have 'technology news' as a keyword, and then discover that it is triggering 'android technology news' it might be better to place this keyword in the 'android' ad group.

• Don't use broad match: Confession. I completely stopped using broad match. It just doesn't make sense for several reasons. First, you almost never have unlimited budget, so you are better off going with the more targeted options mentioned above. Second, Google's algorithm may go crazy in what it determines to be relevant for the broad match keywords. I once had a broad match term 'sports news' and my ad was triggered by searching for "iraq", which was definitely a 'news' keyword but not really a sports one. I recommend doing more extensive keyword research and gradually growing the account as opposed to getting some bad surprises. Broad match might be useful if yo have an extremely small niche you are targeting, and it would be really profitable to just capture one client. It can also make sense when you are in a completely new industry and keywords are not yet established. In any case, it should be used with care, starting with a small budget, keeping an eye on irrelevant keywords, and utilizing negative match to make sure you eliminate unnecessary impressions.
• Use the broad match modifier: this is a very good option in an industry where you are totally new, and need to know more about keywords. It gives you more control and relevancy than broad match.
• Start with analytics: Many advertisers start with the keywords then write ads and choose landing pages. I prefer starting with discovering what is already working and building on it. I like to look at the site's analytics and discover the best landing pages in terms of bounce rate, conversion rate, pages/visit, or whatever your metric is. Then, after identifying the best ones, I try to see what ads can attract users to this content, and then find the relevant keywords for it. This helps build your account's history, and you make sure you are sending users to pages that are the most likely to convert, compared to other pages on the site. The only case you can't do this is with brand new sites, where you will have to make your own judgment.
• Do the campaigns with the objective of building an understanding: If you approach the campaign as a learning experience as opposed to just 'optimizing', your focus will be on discovering what works for your targeted audience and your website. Whatever you do in terms of keywords, ads, landing pages, will (should) be used to build on your understanding. Start with questions like, "are electronics shoppers more interested in free shipping, or better prices?", "what sells better for sports fans, deadlines that induce immediate action, or deals that are better than the competition?" If you go with this curiosity your mind will be looking for ways to answer these questions, and this will help guide your strategies and especially testing. Speaking of which...
• Test, test, test: Never have one ad per ad group. Let the name "ad group" remind you that it has to be a group of ads. At least two. In order to answer the questions you are trying to answer, you need to be split testing your ads. This means you have two ads that are identical in everything but one element. This element could be the headline, call to action, benefit, feature, or anything you want to test. This way you can attribute any differences to that element. After you identify the better performing ad, you should pause the lower performing one, and create a new one hoping to 'beat' the one you already have, and so on.
• Tightly group keywords: The more focused you ad groups are, the more relevant your ads would be for your audience. The ad group split is the most important element in your account structure, because this is where you decide that a certain group of keywords are relevant to a certain group of keywords, which show very similar user intent. There is no magical number to how many keywords you should have, but here is a way to group keywords for your campaigns. 

The general principle is not to get lost in the features of keywords, ad groups, and landing pages, and simply focus on the business needs of the website, the users who can benefit from them, and how to connect the two together. If you get that figured out, choosing which technique to use become a much easier and fun process.

We now finally have an agreement among the three main search engines on a standard for structured data, and how to mark it up. Schema.org was released and it documents how publishers can standardize the structure of their content.

The potential effect can be huge on users, where machines now are getting a huge boost of intelligence simply by us feeding them with knowledge instead of having them become so intelligent that they can understand things the way a human being can.

None of this is actually new, and there is no technology breakthrough. It is simply a human system that we agree upon, and feed the computers with data and "meanings" about these data. Hence, the name 'semantic web'

Imagine how much process and statistics a computer or a search engine needs to do in order to understand when to treat apple as a fruit and when to treat it as a computer company. Imagine something more dramatic like "banana republic", two unrelated words, each in a different field, and yet the phrase is something completely different.

What will happen now is that when someone writes an article, they will mark it up with the related meta tags, so that the computer understands how to treat this string of letters. The users reading, will see the same content, but the computer will see structured data.

The effect can easily be seen on search engine results, and this where search engines graduate to the next level of usefulness and become, as Bing claimed to want to be, decision engines.

They will give us meaningful information about the search queries we are looking for in order to better decide where to go.

A big category of content is recipes:

Traditionally search engines would try to figure out what part of the page is the most relevant to your search query, and then provide the closest thing in the search results' snippets.This required the search engine to understand what you meant by that query, and accordingly provide you with something useful. With semantic markup, and if the user includes in the query something helpful like "recipe", this tremendously helps the search engine, and gives the users relevant results in two ways:

1. Displaying relevant information: this is usually presented in gray, right under the headline or title of each result. Because the search engine figured out the relevant pages you might be interested in, it gives you information from that page that would help you better decide which result is the best for you. In the above case, you are probably interested to know about the calorie value and preparation time to select which option is the best for you.
2. Special search options: Since each type of content has its own set of attributes and uses, we need a different set of tools to refine results for each type of information. In a different example, after Google figured out that "laptop" is probably a shopping query, it gave me different options for shopping. It asked for my location, so it can give results "nearby", and it also presented me with a whole bunch of attributes, based on which I can filter the right laptop for me.

The really important thing here, is that the search engine doesn't really need to deeply know the different options a laptop has, and figure out how to display them. That would require a lot of processing and intelligence.

The content publishers, the sellers of laptops in this case already did the homework for google and when they published their products, they semantically tagged each attribute. Now all the search engine has to do is pick up these attributes and display them for the user in a structured way, where the filter is a useful one for the user.

Instead going to ten different pages, going back, refining your query and finding the best way to phrase it, you can do much of the filtering and choice before you go to the page, and this dramatically increases your chances of finding what you are looking for. 

This is a crucial step in how we find information.

It all stems from offline thinking. Many people think that the home page is just like the cover of a magazine, and therefore it gets the most "exposure", and therefore you achieve "reach".

One of the worst places to place your ads on is the home page, especially if it's a portal.

Here's why:

1. Targeting: Since the home page is usually the most general page on a website, you have no idea what kind of user is visiting it. Let's take an example of one big portal, MSN. Currently the navigation bar includes the following links: news, entertainment, sports, money, lifestyle, locals, autos. The purpose of that page is to route users to other pages. Even if they don't know it, there is nothing to do or read on the home page but links and very short teasers of articles, so the user will quickly end up going to another page to continue whatever task they came to do.
   A full discussion of the categorization of the different types of pages can be found on Semphonic's Functionalism, and in a more specific white paper.
2. Mental State: Since this is a spring board kind of page, and the user in a browsing mode, their hand is still on the mouse, and their eyes are moving quickly, scanning, and evaluating the different options. They are trying to find something interesting, and as mentioned in point 1, you cannot know what that content is, because it varies from cars to sports, and from entertainment to money. If you take a look at the submenus, you will easily realize that this page can be interesting to anyone on Earth!

Since the user is not identified to be interested in anything in particular, and since they are in a scanning state of mind, the home page is not where they would be most receptive.

The best place to be on is what some people call "deep pages", which are pages of articles, videos, or any content where the user sits back, releases their mouse and start reading. A quick correction, there is no such thing as a deep page. The web is a flat place, and since most traffic starts from a search engine or social media, you are usually sent directly to that article page or video page to immediately view and many times you continue on a journey in the website without even seeing the home page.

I previously wrote about the power of contextual targeting to inform us about the user reading the content, and following that logic, the home page is not effective at all.

Going back to our offline magazine advertising friend, I would argue that even in magazines, advertising on the cover would not be as effective as advertising a product next to articles related to that industry. The cover will definitely give you more reach, but at a much higher cost. Most of that reach is irrelevant audience, because everyone reading the magazine will have to start at the cover. The reduction in price, appearing next to your article, and using the remaining budget to advertise on more publications with a more focused reach would get you much better results, because you are reaching people who are in the market for your product vs. reaching the masses. Of course there is no way to prove it offline, but when done online and the cost-per-acquisition is calculated it is clear that the targeted option performs much better.

When might being on the home page work? Very rare cases where your brand is highly known, used by almost everyone, and depends on impulse. Carbonated drinks for example want to just remind you how thirsty you are, how hot it is, and how much you need that extra coke. Most brands are not like that, and you usually have a limited budget and you are much better off spending it based on performance.

Targeting the right audience is clearly one of the most important things to do in your campaigns. This can be achieved in several different forms; demographics, search keywords, behavior, and others. Contextual targeting, mainly brought to mainstream by Google's AdSense program provides one of the most powerful targeting tools out there. I would say it is the second most powerful targeting tool after targeting by keyword.

What does it mean that a user is reading a certain page? How is this page related to their interests and what does it have to do with my campaigns? How do I target my contextual campaigns?

The short answer: people don't just 'end up' on a page, they generally know where they are going, and therefore are interested in the content of the page they are at.

Let's take a look at the different ways in which a user can end up on a certain web page:

1. Directly: The most straightforward way to go to a page is typing in the URL in the address bar, www.website.com. If you know the URL by heart you know what that page's content is about, and you are interested in it (or interested enough to find out what it's like). The user knows where she is going.
2. Clicking on a Link: A user might click on a link while they are browsing to go to a new page. Naturally, there is a description of that link and an expectation of what kind of content the user might expect to see if they click. Of course some sites create redirects, or have misleading copy, but we are assuming you want to advertise on a legitimate website that doesn't do these things. Therefore, the user knows where she is going.
3. Search Result: The search result snippet is a description of what users can expect and again, the user knows where she is going.
4. Advertisement: Clicking on an ad is another way someone can end up on a page. Again, the ad is promising something, and therefore the user is interested in that thing and goes in the hope of finding that thing. She knows where she is going.

Even if the site is legitimate and the content is created with the best of intentions, several things might go against the above reasoning. Misunderstanding of text, clicking by mistake, are two examples, but we can still say that in most cases people are on a page because they are expecting something. And since the four ways of ending up on a site are all voluntary, then we can safely say that the user is interested in that content.

From an advertising perspective, placing ads on pages (not websites) that have content about a certain topic is highly relevant to the audience on these particular pages. Higher relevancy of course means better placement, and a higher probability of a relevant audience that will be interested in whatever offering you have.

This is completely different from the practice of placing ads on websites, or sections of websites. In the contextual targeting case, you are targeting by page. You are placing ads on any page that has the keywords you want. Although the site might be completely irrelevant some pages might be talking about your specific topics, and therefore, it will make sense to advertise on these pages.

It would be inefficient to go to 2,000 websites, and make 2,000 deals for specific URLs. That would be madness. But AdWords' contextual targeting capability solves this in a great way, and is able to connect advertisers with a very specific audience.

Having dealt with many traditional marketers entering the online world, and having gone through the transition myself, the best advice that I could share with anyone joining this world (and everyone will, eventually) is the title of this book. Whatever You Think, Think The Opposite is a brilliant book that doesn't have to do anything with online or offline, but the advice makes a lot of sense to me because almost everything that we do online is counter-intuitive to the thinking of traditional economics. Paul Arden presents his ideas by using the medium he is best at, advertising. The pages of the books have the structure of an ad. A big image with a headline and some descriptive text. Each page can be read separately without having to read other pages for context. The only difference between these pages and ads are brand and logos. So, me easy advice would be to think the opposite of the way you think. Most of the thinking and planning works the other way round; big becomes small, one becomes many, decide become discover, etc. Here I outline some of the main differences in doing marketing online compared to the thinking offline.