ClickEquations Blog

Keywords are one of the false gods of PPC. There’s really no reason to get to hung up on keywords.

The goal of our campaigns is to have our text-ads matched with the most appropriate search queries. Keywords are just the tool we use to get to the most qualified queries.

With that in mind, it’s my opinion that the world of keyword selection and expansion is quite broken.

Keyword selection in PPC – broadly and generally as I’ve seen it practiced and promoted by both ‘experts’ and tools providers – is about finding every possible word and phrase related to the category or topic at hand.

This is a great strategy if you’re a paid search engine looking to make money from way too many clicks with way too little targeting.

It’s not really to your advantage if you’re an advertiser looking to maximize returns.

Waiting For Your Keywords To Bark
Bryan and Jeffery Eisenberg wrote ‘Waiting For Your Cat To Bark‘ several years ago, as one of several books covering their Persuasion Architecture process (now built into their OnTarget offering), and it remains a book I don’t think any online marketer should miss.

Among the many brain-tingling discussions in ‘Bark’ is the idea that people come to the web with a very specific idea in mind, a personality all their own (but categorically like a lot of other people), and a situation that they’re in along with a goal they’re trying to achieve.

This bundle makes up their buying process. I’m a massively geeky tech freak with a strong need to fit in and a brother whose birthday is Saturday so I MUST order something for him today.

You’re online trying to sell stuff. Your mind is on the great price you offer on the new ‘Widget9000′ and the free shipping program you just launched.

I’ll let the Eisen-brothers tell you how to solve this mis-match (ok, a clue: align your selling with their buying, the other way around isn’t going to happen.)

But what does this have to do with keywords?

Up With People
Traditional keyword development and expansion is all about saturation bombing a category or topic. The suggestion tools and brainstorming techniques we’ve all relied on toss in (or try to) anything contextually relevant.

This is too low resolution and comes at the problem from the wrong direction.

Let’s think about it the other way. (IOW: What would Bryan do?)

Imagine a specific person, in their full psychological glory, in a specific situation who wants/needs/is curious about your product or offering. What are they likely to search for? Build the list of words and phrases that capture their needs given the details you’ve assumed.

Start with the most specific and detailed versions of what they might ask, and then slowly narrow it to queries that at least lean in their general direction.

Break down the components of the query – how might they reflect their product desires? How might they reference their urgency? What clues might appear to show that they prefer well-liked and popular products?

Stepping through the range of queries you can imagine, from deeply personal and unique out towards general queries that anyone might do. Taking this deliberate step adds another layer of clarity to each keyword. Some are deeply targeted and precise. Others are vague and broad. Shouldn’t your measurement, bidding, expectations, and text-ads align with these attributes?

Repeat this process for other kinds of people, or other reasons people might have, for visiting your site or buying your products/services. (By now you’ve gone and read the book and have built a full set of user persona’s right?)

Of course, most users won’t load their query with clues to every aspect of their needs, personality, and situation. But some will and more importantly this exercise creates the beginning of an intelligently tiered keyword list we can use to evaluate our campaigns and keywords with a new level of precision.

SEO your PPC
The idea of really thinking hard about the specific queries people are likely to execute is central to good organic paid search optimization.

In the organic world, where broad-match doesn’t exist, a page can only rank for a limited number of keywords, and there is a content+effort cost for each rank, the spray-and-pray approach isn’t practiced and certainly isn’t effective.

Never thought I’d say it, but when it comes to keywords, PPC folks can learn a lot from the SEOs.

A number of people followed up to last week’s ‘Bidding on Brand Terms‘ post and asked how the logic applied to the broader world of buying PPC keywords where you already have organic rankings.

Generally I think the logic does apply, but with a slightly different set of rules and conditions:

• If an organic keyword is highly profitable, I would assume the paid keyword would be profitable and incremental unless proven otherwise. There certainly could and will be exceptions – if it’s a highly competitive keyword with insanely high prices for example – the competitive PPC bidders may be acting irrationally and you may be better off to take the free traffic and let them kill each other. Another example might be broad terms with a lot of organic clicks and just a few conversions. But determine this via tests not assumptions.
• Cannibalization should be more than offset by incremental traffic. Yes some people who click your paid ad would have clicked your organic ad. But many who click your paid ad would not have clicked your organic ad. I firmly believe that there are paid clickers and organic clickers and a smaller minority who’ll go either way.
• Marginal Net Revenue is all that matters. If you’re making $1000/day with organic alone, and make $1200/day (net profit not gross revenue) with organic plus paid, then organic plus paid is better. The internal fact that some of that revenue could have been had at a lower acquisition cost is irrelevant.

And More Importantly
Organic results are all exact match. When you rank highly for an organic search, in most cases you don’t rank equally well for any/many of the thousands of variants of that query. Every organic result is computed independently.

But when you buy that same organic keyword as a paid keyword, you get to use match types to cover hundreds or thousands of queries – the vast majority of which you’d never win – or perhaps even appear on the first page for organically.

So when you find a winning organic keyword and transfer it to paid, you’re not only buying space on that results page, but if done correctly (by expanding the word or phrase and using match types fully) you can leverage that winning word 1000:1 or more.

Should you have to pay Google to get traffic on your own brand keywords?

Before we answer, let’s define our terminology.

By ‘Brand Keywords’ I’m referring to keywords which center around your company name, which in most cases is your domain name (or a major part of your domain name).

I’m not referring to major brand names that you sell as a retailer. And if you’re a manufacturer of many brand-name items, I wouldn’t even include those product brands.

Just your core company-name brand.

So should you have to buy these keywords in your PPC accounts and pay-per-click for that traffic?

Probably not. But we don’t live in that world.

Why Bidding Your Own Brand Makes (economic) Sense
There are two arguments against bidding your own brand terms:

1. My pages rank well organically, I’ll get the traffic anyway.
2. There is no f&*king way I’m paying for traffic on my own brand.

Yet the arguments for bidding on your own brand terms are pretty simple.

1. You probably don’t rank well, or at all, for every variation and mis-spelling or phrase use of your brand. There are hundreds or thousands of them.
2. Some people just look at and click the paid ads – they prefer them over the free listings.
3. If you don’t buy it, someone else will – and it’s not likely they’re trying to improve your business.

I recommend thinking about it as a part of a much larger expense.

Consider all the money you spend building and promoting your brand. You’ve invested a ton of money into getting people to know it, perhaps even trust it, often advertising in other media which is what generated the search in the first place – all that time and money get them to initiate a search to try and find you.

Almost certainly the money you spend for this ‘last mile’ of the relationship is a tiny fraction of what you spent to get them to that point. Pay the last few % and get those folks to your website.

What sense does it make to spend thousands on branding, trade shows, tv commercials, mailings, social media efforts, or whatever it is you do – all of which ultimately motivates someone to try and Google you – only to have them see and then click on ads for competitors because you weren’t bidding.

This doesn’t mean you shouldn’t do your best to rank #1, or get multiple organic listings, on your brand terms. As Avinash says, those rankings are your God-Given-Right. (Google makes you earn them anyway – but that’s another blog post

Prove Me Wrong
Most tests I’ve heard about, when paid and organic ads were run together and testing was done to turn off the paid ads, showed that while there is some cannibalization of organic by paid, the net effect was positive.

But if you’re really concerned test it yourself.

• Use a reasonably long time frame (with a solid number of clicks, I’d suggest at least one or two weeks to ensure at least a few hundred clicks of data) and run with your PPC ads for one period and then without for another.
• Another important factor is that historically paid clicks convert at a higher rate than organic ads – so even if you just miss a few of your visitors they may have been very lucerative ones.
• Make sure to isolate as much as possible for other factors, like major SEO/organic rank changes, seasonal volume levels, etc.

Check the impact on your organic traffic and overall traffic and conversions.

And when you get your results back, please post a comment with your experiences. I’d love to hear about cases where PPC brand term bidding is purely cannibalistic and a waste of money.

Until Then, Bid On Your Brand Terms
I believe that not bidding on brand terms is cutting off your nose to spite your face. And your face looks funny without a nose.

Average Position occupies an important place in the mythology of paid search.

Many people covet or chase higher positions, and there are several possible reasons:

1. The assumption that ads in higher positions get more clicks simply because they’re in higher positions.
2. As we all know ‘higher is always better’ – especially when it costs more.
3. And of course, eye tracking studies prove, um, er, that people look higher more often.

(The empirical and anecdotal evidence I’ve seen suggests that the power of higher positions is much less than most people seem to imagine. In a future post I’ll go into this in great detail. This is not the real subject of this post.)

As a result, there is a lot of attention paid to the Average Position metric. And a LOT of money is spent on upward bid changes made because of the number this metric reports.

So how good is this number? Probably not very good.

Let’s take one tiny little case study to demonstrate.

The keyword is ‘cat treatment’. And on Saturday Jan 3rd it produced about 25 clicks in one of our accounts. The average position for the term (in broad match) was listed as 4.55. This is the average of all the positions in which it appeared during the 1543 impressions it enjoyed that day.

Now be honest, despite all you know about averages (including the fact that it could have appeared in position #1 760 times, and in position 8 783 times) when you see that 4.55 was the average it makes you think it spent the day bouncing between position 4 and position 5. Right?

But did it?

Let’s look at Google Analytics’ handy Keyword Position report for this keyword on that day. This shows the position the keyword was in when it earned its 25 clicks.

Yowza!  This keyword covered more ground than Paris Hilton in NYC on Saturday night. (I always wanted to see how much Google traffic a single Paris Hilton reference caused.)

It was in all three top positions, and everywhere on the right side from position 1 to position 6.

Keep in mind that this is a map of clicks, not impressions. So maybe the impressions did cluster closely around the 4.55 average and the few stray impressions way up to top 1 and down to position 6 just all got clicks. Or maybe the actual impression distribution was extremely broad and the 4.55 average, while it is true, is really not useful to us in terms of analyzing keyword performance or making bidding decisions.

At this point only two things are really clear;

1. We really need better information. If the search engines won’t provide the actual impression and click position distributions, and/or make the the position-at-time-of-click a macro that can be delivered in the target URL, they should at least provide the standard deviation for the average position so we have some idea of what it really means.
2. We should resist the urge to put much faith, or make too serious of decisions, based on the reported Average Position of any keyword.

What one piece of advice would I give to help improve a paid search campaign?

That was a question asked of our panel as SES in San Jose last week.

My answer: Make sure your brand keywords are fully segregated from all others.

Brand keywords – any keyword with your company name or variations in them – have completely different cost and performance characteristics than category or other other generic or product specific keywords.

These differences completely confuse the reporting for any campaigns and Ad-Groups if they’re co-mingled.

Separating Keywords and Queries
The first step is easy – every keyword you buy, regardless of its Match Type, should be in an Ad-Group if not a Campaign with only other keywords that contain the Brand name too.

Preferably, the brand terms are bucketed, with the ‘Pure’ Brand keywords in one group (those that represent just the name and variations itself), the navigational versions in another (www.brand.com, brand homepage, etc.) and the Brand-Plus keywords (Brand Sweatpants, Brand Coupons, etc.) in yet another, and so on.

In these brand focused Ad-Groups, you have to use Broad and Advanced match very sparingly and carefully, and eventually almost entirely eliminate them. If you leave them, you’ll get too many non-brand queries matching and diluting the intent of these highly focused Ad-Groups.

The other side of this Broad/Advanced Match coin is that you’ll also want to add your brand as a negative in all the remaining non-branded Campaigns and Ad-Groups. Otherwise the engines will match brand-inclusive queries against your non-brand targeted keywords.

This can be and feel dangerous, if you’re not completely sure that your Brand campaigns are complete, bid properly, running the full range of Match-Types (with of course the Match Type Keyword Traps fully configured and loaded.)

It’s probably a good idea to skip this step of adding the brand as negatives in the non-branded campaigns for a few days to ensure that there aren’t certain query formulations that your new Brand targeted Campaigns are missing.

Watch the query reports carefully, and add variations to the brand campaigns, and ultimately more negatives to both the brand the non-brand campaigns.

The Payoff
Immediately upon starting this process, especially if your campaigns had brand terms and lots of broad match scattered throughout, you’ll see radical shifts in your search reports.

• You may be amazed how much revenue is coming from and and how little cost is going into your pure brand campaigns. That’s the good news.
• You may be shocked at how much money and how little revenue is coming from your now-strictly-non-brand ad-groups. That’s the bad news. Or the opportunity, depending on how you look at it.

In any case, you’ll have a new level of clarity about the performance and activity in your PPC campaigns.

Coming Up
I’ll share more thoughts on the execution of full brand segregation, and the implications of the changes it makes to your reported results, in future posts. This is another one that may take 3-4 posts to just scratch the surface of.

In the meantime, questions and comments are encouraged. Are your brand terms separated into ad-groups? Does that help you better understand the way your PPC budgets are spent? What problems have you seen trying to control brand via Match Types? Any other ideas?

One thought I wasn’t able to put in the last post about missing and misleading click data, was about keyword click-through-rates.

Do keywords really have click-through-rates?

Objectively they do because the engines report them. But does that make sense?

If A Keyword Falls In The Forest, And The User Doesn’t See It…

The user doesn’t even know the keyword exists. The user typed a query (which in some small percentage of searches was exactly matched to the keyword, but far more often was only related to the keyword) and was shown (if they even saw it) a text-ad (containing some specific copy) in some position on the page in relation to a number of other text-ads (not to mention the organic search results.)

What portion of the influence in that click, or lack thereof, did the keyword have?

• We know different text-ad copy produces different CTRs.
• We know different positions result in different CTRs.
• We know that the presence or absence of specific competitive adds produce different CTRs.
• We know different queries that may match to the same keyword in broad or phrase match type have different CTRs.
• We can assume that CTRs vary by time and the geography of the user.
• There must be a couple of other factors I’m not thinking of right now… (comments?)

So does the keyword really have a CTR, or do the combinations really have CTRs? Clearly the Keyword CTR is the average of a range of different situations and conditions.

The Average Average is Only So-So

There are a lot of averages presented in search analytics. That’s necessary as we can’t handle all the granules, but close attention must be paid to the composition of these averages, lest they be less than clear or useful.

If the campaign is reasonably constructed in terms of organization and match type application, and are being reasonably run (meaning the text-ads and bids have both logic and dilligence being regularly applied to them), then the average CTR as reported for keywords can be useful. If any of these elements are missing, the utility dwindles rapidly.

As with most averages in PPC reports, if you aren’t sure dive down and look at the components – the more performance diversity you find inside the less weight you should place on the average.

Know Your Metrics

Just another example of the fact that even the simple metrics of paid search have more to or behind them than you might realize, and how some understanding and healthy skepticism can help you get closer to truly understanding what’s happening in your campaigns.

(Credit where it’s due: The idea of questioning KW CTRs, and many other ideas you’ll find in this blog from time to time, was first suggested by Bruce Ernst)

(Upcoming Events: I’ll be at the Semphonic XChange Conference in San Francisco on Aug 17-19, and am Speaking on “Identify, Analyze, Act: SEM by the Numbers” at Search Engine Strategies in San Jose on August 19th)

Another lawsuit aimed and preventing the use of trademarked keywords was dropped this week. This time it was American Airlines who had filed a lawsuit against Google for allowing other to use their name to trigger the display of competitive ads.

According to Bloomberg:

American claimed Google violated its trademark by allowing competing airlines to bid on keyword searches that generate “sponsored link” ads on search-results Web pages. The ads take advantage of the American brand’s popularity, even if the name isn’t used in the ad, the carrier said.

Google settled similar suits by other U.S. companies before the untested area of trademark law could be addressed by a judge or jury. Foreign lawsuits still pose challenges to the advertising practice, part of Google’s AdWords program.

Courts in France have held Google liable for allowing advertisers to select trademarked terms as keywords, according to U.S. regulatory filings. Google, based in Mountain View, California, said it is handling or recently resolved similar cases in Germany, Israel, Italy, Austria and Australia.

Google had argued that its “invisible” use of trademarks isn’t technically “trademark use” under U.S. law. Google compared the program to practices such as placing generic drugs next to name brands in pharmacies and buying billboard ads next to those of competitors.

This last point is the reason I’ve never understood the merits of these suits. Trademark law is designed, in my very simple understanding, to prevent one company from confusing customers with a name that is similar (or identical) to another company.

Does buying a trademarked term as a keyword provide one company benefit from the name and reputation of another? Certainly. But isn’t that why all the car dealers rent space on the same block? Doesn’t it happen when magazines review all the products in one category together?

Every company in the world wants to steal customers and prospects from their competitors. Their efforts to do so yield better features, better pricing, and loads of other consumer benefits.

Using trademark protection to limit confusion benefits consumers. Using it to try and limit consumers knowledge and awareness of competition harms consumers, and should itself be illegal. Great to see a lawsuit go the right way.

NOTE: This is part of a post series. It’s available as a single post for easier reading: The Match Type Series.
===============================================
The first and second posts of this series introduced and expanded on the concept of the Match Type Keyword Trap. In the first, it said:

Over time, by watching the queries that each keyword attracts we can tune this system quite precisely, not only filtering unwanted queries with new negative keywords, but expanding our total volume through quality score and bidding improvements and tailoring the ROI of different query classes.

And promised to expand and elaborate. So here it is:

Tuning Options

When you buy the same keyword at different match types, or different keywords and phrases at different match types in a coordinated effort to properly target and value queries, your initial settings will be less than perfect.

You don’t know what queries you’ll see, the price you’ll pay for each, or how they’ll perform in terms of conversions.

But you can monitor and measure each of these over time, and make adjustments to create a more effective trap.

There are three controls you’ll primarily use to tune the trap:

• Negative Keywords - Without question your campaigns will see (and you’ll pay for) queries that are either undesirable or prove to be poor performers. You should continually review query reports and add keywords as negatives either to all appropriate ad-groups, or to those which have bids above the keyword value.
• Add or Move Keywords - As you review the queries caught by each keyword and ad-group, and the performance of both queries and keywords, there will be interesting or well performing keywords which should be moved up the match type & bid hierarchy.

  If a query is performing exceptionally well against the Phrase Match option, for example, you might want to create an Exact Match copy of that keyword and give it a higher bid. This should cause that query to be grabbed by your new Exact Match and yet let other matches to that Phrase Match keyword keep matching there.

  Well performing keywords in the Broad Match group (which is usually bid particularly low) are especially good candidates to be ‘promoted’ into the higher-bids & more targeted environments of the Phrase Match or Exact Match ad-groups.

• Raise or Lower Bids – Based on your goals (revenue or CPA or ROI or whatever) and reflecting the measured performance of the purchased keywords, find the right shape of the pyramid by bidding good Exact Match performance up and cutting Broad Match bids as you negative out losers and promote winners.

I should have pointed out somewhere earlier, that by far the best way to configure the MTKT is to separately each keyword group with different Match Types into separate Ad-Groups. This makes reporting and measurement easier, and allows you to control negatives at the right level.

As a naming convention , we end each Ad-Group name with a (E) if it holds Exact Match keyword, (P) if it holds Phrase Match keywords, and (B) for Broad Match. This makes is much easier when visually inspecting reports or making account changes.

Measuring Progress

Success with your MTKT is achieved when you’re attracting only desirable queries and have maximized ROI by setting bids according to conversion profitability.

Reviewing the queries on an ad-group by ad-group basis is the cornerstone of the process.  The Exact Match keywords should be clear and profitable. The Phrase Matches should be on target or quickly either promoted to Exact or made into negatives. And the Broad Match should also winnow down in many cases (but not always) through promotion or negative creation.

In some cases the Broad Match ad-groups are ultimately turned off, or left running with extra low bids just to capture any potentially new and interesting queries.

Results are harder to summarize, although as pointed out in the previous post, what you normally shouldn’t see is great variation between the ROI (or ROAS if you must still use that horrid metric) for the different Match Type divided ad-groups.

There are exceptions, but they should be positive ones where Exact Match, or more rarely Phrase Match groups are extremely profitable while others are just normally so. But very low or negative returns are a sign that either the queries being attracted just don’t have potential, or else something later in the chain is wrong – ad-text, landing page, offer, checkout process etc.

Next Up

In the next post, unless something else comes up we’ll finally cover that Rock Scissors Paper game I promised to disclose.

NOTE: This is part of a post series. It’s available as a single post for easier reading: The Match Type Series.
===============================================
The previous post introduced the idea of building a Match Type Keyword Trap. This layering of keyword & match type combinations provides control over which, where, and how queries are attracted, and therefore their cost-per-click.

In the simplest case, you’d buy one keyword (say ‘Whaazooh’) three times in one campaign – once on Exact Match, once on Phrase Match, and once on Broad Match.

The goal is to catch all queries which are literally ‘Whaazooh’ with the Exact Match keyword, all queries which are ‘Whaazooh’ plus some word(s) before or after it with the Phrase Match, and all other related queries with the Broad Match.

Why?

Because in almost every case where many different queries exist for a single word or topic, some of those queries are very valuable, some are mildly valuable, and many are not valuable (or at least not valuable enough). We want to segregate these queries by their value to us so we can pay highly for the high value ones and less so for those less valuable.

In the simple cases (I have to keep saying that because not all cases are simple, there are many complex variants of this) we’ll do better by trapping the best ones with the most specific Match Types (Exact if possible or Phrase if not) and using Broad Match to harvest winners and losers which are acted upon accordingly.

Winners are promoted (to Phrase Match or Exact Match). Losers are demoted via lower bids or even made into negative keywords.

We do better not because of the place they’re trapped, but because by segregating them we control the bid (as well as the text-ad, landing page, etc.)

Forcing The Stack

Buying the same keyword three times at different match types does not itself bait the trap. If the same word is purchased at both Exact and Broad, and has the same bid and earns the same quality score, chances are good a related query with be matched sometimes to one and other times to the other.

To force the trap to work you have to stack the bids – higher for the Exact Match versions and sequentially lower for the Phrase and Broad Match versions. This gives the Exact Match keyword multiple reasons to attract and win the Exact Match queries; it is a better match and it is bid higher (which is good in itself and factors into quality score).

When you do this, leave enough room between the various bids. The Average CPC the engines report are averages, so expect a range of bids in each and leave enough room so the ranges don’t overlap.

In this example, we bid higher for several terms that have proven great performers, setting them on Exact Match and bidding $1.25. Several others that are good performers and perhaps come in some variations are set at Phrase Match for $0.65. A larger collection of phrases and concepts are bid Broad Match at $0.15. Over time we shift, add, put in more negatives, and generally take control over how we pay for and catch queries.

Equalizing Return

How do you know if it’s working?

In theory you’ll normalize the ROI (or ROAS if you must) for your Exact, Phrase, and Broad Match keywords. In other words, you’ll raise bids for your Exact Match keywords to maximize profits. You’ll set accordingly lower bids on Phrase and Broad Match keywords until they produce the same return as the Exact Match does – so their lower conversion rates and ROI are compensated for with proportionally lower bids.

They get the bid they deserve.

To Be Continued

Again there are many exceptions and details left out of the above descriptions for the sake of time and length, but I’ll move into examples in future posts which should illuminate the concept. In the meantime, if you have any questions about this leave a comment and I’ll elaborate.