Thus far in this series we’ve talked about the difficulty of getting a clear view of paid search performance, of deciding the most urgent risks and opportunities amidst the volumes of data that you do have.

Now we come to the third issue: the productivity (or lack thereof) in making changes or improvements to your paid search account.

Possible Changes To Improve Search Campaigns

There are a limited number of things you might want/need to do to your paid search campaigns. Most of them aren’t too difficult when required on a small scale. But there’s not much in PPC campaigns that really happens on a small scale, which is where the frustration begins.

You might want to add keywords. It’s not hard to generate a large list of incremental keywords, and there are tools to help you do it. You can even harvest search queries, scape competitor websites, or get lists from Compete or Hitwise of terms driving traffic for others.

But to effectively apply a list of keywords they need to be expanded and parsed into versions and phrases and synonyms and layered across match types and segregated into ad-groups and campaigns and matched with bids and text-ads. The ideal environment for this would both facilitate the process as a whole and provide suggestions based on a learning algorithm which watched your style of division and targeting.

You could see the need to modify match types based on your search queries to build more effective match type keyword traps. This requires versioning keywords, segregating them into Ad-groups, pyramiding bids, and making sure the net is wide and lacks gaps or overlaps. Software could visualize this process and make it ‘drag and drop’ and even ‘bionic’ if someone put a little effort in.

Your bids may need to be changed, and of course this is the one task to which some substantial software automation development effort has been placed. This is a big topic I’ll save for a future series of in-depth posts.

You might need to substantially reorganize your campaigns. This happens for all kinds of reasons, many having to do with the impact of organization on the roll-up summary numbers as presented, some having to do with quality score management, the issue of match-type control and reporting, issues of geo-segmentation, and of course good old logical segmentation.

The technology provided for campaign reorganization today - cut and paste - is getting a little dated and I feel confident that a more elegant and productive solution could be conceived and developed.

The text-ads you’re running may need to be altered. While the idea of presenting four blank boxes and allowing unlimited freedom (with the constraints of available character limits) is powerful, perhaps there would be some advantage in tracking and analyzing the different ‘recipes’ used in various ads, building up repositories of different synonyms for important concepts and then making it easy to re-use effective ones and tracking how they perform both individually and as groups based on their relative position in the ad, in the ad as it runs at different positions or on different days, etc.

Lastly there is a chance that you’ll need to test different landing pages (leaving alone for now the implications of testing various designs within a single landing page). From the typical home page vs category page vs item page variances, it may be wise to consider user personas based on the keywords and queries and other factors as well.

Here again the current ‘type-anything-you-want’ technology could be enhanced by allowing simple meta data to be entered and tracked (how are item pages doing in terms of conversion vs category pages vs the home page) and enabling automated testing of these variations. And it doesn’t have to be limited to just a simple ‘which page’ consideration - performance may vary by the length of the query or number of works in the keyword phrase?, time of day, day of week, visit number, or many other factors. Software could track and optimize this.

Working In A Coal Mine

The common element in the current state of paid search management is that only one of the steps in even the most simplified version of the process has progressed even one iota in the last five or more years in terms of automation.

Tens of thousands of people are being treated as migrant-search-workers standing in the hot sun every day harvesting keywords and clicks.

And for the moment we’re not talking about the chisels and stones they’re given to bang out reports and dashboards.

Where is the Eli Whitney of PPC?

(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)

In my ‘Three Challenges’ Post I wrote the following to describe one of the fundamental reasons why I think the process of managing paid search needs to be improved:

There is a lack of clarity. It is amazingly difficult to get accurate and complete data on campaign performance and results. Much of the data you need to see is scattered across three to five different tools and interfaces. Other data is presented in formats or based on calculations that just aren’t right. (they’re wrong.) Still other information is seemingly unavailable. There is no quick and accurate way to get reports which are satisfying.

Since then I’ve written four posts in an attempt to explain and expand. But I’m not sure I captured it.

To manage something effectively it’s necessary to see cause and effect. The paid search networks use such complicated rules and hide certain key data elements which make this impossible.

Search queries, which are the primary driver of search success, are a key example. But it’s really the full relationship between queries and keywords and match types and quality score and text-ads and landing pages. The truth lies in that matrix somewhere, but nobody is letting you see it.

You see a pile of queries over here (partially, sometimes). A bunch of keywords over there. Some ads further off in the distance. Want to understand the relationships? Put them together in your own head.

Clues are great in a mystery. Not in a business transaction.

In the earlier posts in this series the point was made that it’s hard to get clear and complete information on the performance of PPC accounts.

This is true, in part, because some important information is either unavailable or plays hard-to-get. Examples mentioned included search queries, information about missing clicks, and results in terms of true profitability.

This post drills down on search queries; the others will be covered in future posts.

Finding Queries
Let’s start with an assertion: It is not possible or reasonable to competently manage paid search campaigns without full access to search query details.

This isn’t a nice-to-have. It’s required.

Managing without query details is like managing a baseball team without being allowed to know what happened at the plate. Suppose you’re told who gets on base and who doesn’t, but nothing else.

How do you rate or make changes to your batting lineup without knowing who strikes out, who hits deep long fly balls miraculously caught on the warning track, or who gets hit by pitches?

The analogy may not be perfect, but the point is that choosing to add, delete, change bids, add negatives, not add negatives, or modify match type without knowing queries is a bit of blind-folded juggling.

And yet, most people do manage paid search without full and detailed query reports - they have to because the data is not available to them. With very limited exceptions, you can’t get it from the engines, in web analytics software, or even in specialized paid search management tools.

Why Search Queries Matter (the short version)

Queries are vital because they can contain insight into the desire or intent of the user. If you sell tennis racquets, for example, and buy the keyword ‘tennis racquet’ (using the standard Broad Match) then your ad might be shown to someone who wants ‘tennis racquet restringing’, or ‘New Prince V14 Tennis Racquet’ or  ‘used cheap tennis racquet’ or even someone looking for ‘tennis racquet art’.

Are each of those people relevant to you? Are the ones that are relevant equally relevant? Can you write a single text ad that speaks directly to each of those people and persuades them to click and take action?

If you know the search queries that people clicked on when triggered by the keywords you’re buying, you can answer these questions and take action to improve the targeting and results of your account.

Without knowing, you’re left without the ability to fine tune your campaign, so you waste money and miss revenue opportunities.

Where The Queries Are(n’t)

When talking about this, I’ve found that people usually have either never thought much about the difference between queries and keywords, or have the impression that they do have access to that information but don’t use it aggressively so they haven’t realized the limitations in the little bit of query data they can access.

Let’s review what search query data is available in some of the most widely used SEM analytics and reporting tools:

• In Google Adwords the Search Queries report lists queries at the ad-group level, but it does not tell you which keywords triggered which queries. And they notoriously hide a massive percentage of them in rows marked ‘Other Keywords’.
• In Google Analytics does not display search queries at all, at least by default. It can be hacked to display queries, but from what I can see in the ones I’ve used you cannot see/link the queries to specific keywords (or even bucket them into adgroups).
• In Omniture SiteCatalyst & SearchCenter offer great query support if you purchase the optional ‘db universal’ VISTA rule (typically $5K). With this enabled you gain fairly complete search query reporting and it’s a metric you can use with the full power of SiteCatalyst reporting, meaning you can use the ‘break down by’ feature to subsort by query relating it to keyword, product sold, or just about anything. You can also access it via their Excel tool in powerful ways.
• In most stand-alone paid search management tools (like Clickable, Acquisio, SearchRev, SearchIgnite, Efficient Frontier, and others), search queries do not exist. They’re completely unavailable. These tools rely on the search engine API’s for data - they don’t have their own page/URL tags - so they just can’t get query data. Which means their customers don’t get it either.

There are many other analytics and paid search tools of course, and I don’t personally know the details of many of them. (I believe Marin Software does have their own tags and can gather query data, but I don’t recall the level of reporting, for example.)

If you know the details of available or unavailable query information and reporting, please leave details in the comments.

Missing Data 1, Good Search Reporting 0

Based on this review of the popular platforms people use for paid search reporting, it seems safe to say that the vast majority - probably at least 90% and maybe as many as 98% of search managers do not have the ability to look at which queries drove clicks (and spent their money) on a keyword by keyword basis.

Imagine if your sales records only told you what categories of items you sold, not which specific items or SKU’s were sold. How would you decide on inventory re-orders or future promotional plans. You couldn’t with any level of accuracy so you’d have to just guess and play the averages.

This is what the search engines want you to do. Your inefficiency is their profit margin.

It’s hard to understand why the web analytics and focused paid search software companies place such a low priority on this vital information. I have some theories, which I’ll share in future posts.

A Fair Shot

If it’s the search query/keyword combination that triggers ads, causes your money to be spent, and dramatically clarifies the ‘why’ of who clicked and converted, why should paid search advertisers have to manage their accounts without this information?

I suggest you ask your search engine account managers, or analytics / PPC tool providers that question.

Yesterday I noted that paid search managers face three challenges in trying to effectively manage paid search campaigns:

• A lack of clarity (reporting problems)
• Difficulty defining priorities (strategic and planning problems)
• Horrible inefficiencies (mechanical and processes problems)

I believe that these problems need to be solved in order to improve paid search management, both the profession and the results.

First you need to see what’s happening, then you’ll want to decide what needs to be done, and then you can hopefully get it done with a reasonable amount of effort.

That doesn’t sound like too much to ask.

But 4-5-6 years into explosive growth in paid search and we’re hardly out of the starting gate. Today I’ll expand on the issues regarding reporting and clarity, and in future posts dive more deeply into the problems of setting priorities and executing paid search tasks.

What Paid Search Reports Don’t Tell You

Paid search is about answering questions. People type queries and search engines return results, which are lists of possible answers to the questions they believe are being posed. I want to structure my campaigns as tightly as possible around those search queries.

Every search engine tells you how many impressions your ads had, and how many clicks you got. They have to I suppose, since the CPC is what drives your billing. What I really want to know is what did I miss? And why? Then I can set goals and define strategies or tactics (or at least design tests) to do better.

Each conversion hopefully generates more revenue than it cost to cause that conversion, which is reflected in the rather innane ROAS metric. Being impressed with a good ROAS seems akin to believing you’ve saved money by buying something you didn’t want when it was on sale. Goods or services have costs (COGS) and the only metric that matters is ROI taking account (at least) both direct-marketing and goods/services expenses.

When my clicks do generate revenues, I’d like to know which ones. Then I can make wise decisions about future investment and effort around certain keywords and queries.

Unreasonable Demands?

So I’d like to know which search queries generated which results, how many clicks I didn’t get and why, the actual amount of profit made on each transaction (and from each keyword, query, and click).

Do any of these sound unreasonable? Far-fetched? Demanding?

Yet these desires are not generally or specifically fulfilled through the paid search reporting capabilities provided by the search engines, popular web analytics software, or even specialized PPC management tools.

Surprised? The devil is certainly in the details, and some of the information defined is available in some packages/places, but generally with huge compromises and limitations that disqualifies or invalidates them as actual or sufficient information.

Really? Yes to the best of my knowledge, as the next post will review in somewhat excruciating detail. I’m happy to learn new facts or discuss this further in the comments - significant corrections will be appended to that post.

User search queries, accurate revenue & expense allocation and matching, and ROI reporting are just three of the ways that the current generation of PPC reporting generally fail paid search advertisers and managers.

The fact that these problems/limitations are seemingly not well known, frequently discussed, and therefore clammored for as improvements is one of the things that has to change to move the business/market forward.