SemAngel

Bing – Maybe there really is a Search Engine War

Like the phony war that marked the early months of World War II, the search engine wars have been pretty much lacking in combat. If it takes two to tango, it also takes two to fight a war and Google has so completely owned the field in search that to describe the resulting marketspace as a war has been about as apt as applying the term to when my younger daughter indulges in her habit of playing both hands in a game of cards. War it has not been. Bing may change that.

Bing’s success is contingent on two very different factors – only one of which I have any ability to judge. The first factor is marketing. Whatever Bing’s technical merits, if the marketing campaign doesn’t shift at least trial usage then it simply won’t matter. Google’s lock on search share has been unrelenting; and Google owns most consumer’s mind-share. Personally, I don’t think much of Microsoft’s ad campaign (but that’s not my business). They are, at least, putting some serious dollars behind it.

On the other hand, I’ve been using Bing pretty extensively and we’re also doing some formal measurement on it right now. And on the technical side I think its strengths and weaknesses are pretty obvious.

From a user-interface perspective, Bing is, quite simply, much better than Google. That’s really no surprise – Ask has been better than Google for years without making any real dent in search usage. On the other hand, the difference has never been wider. Bing’s best features include a rollover popup of the page on natural search (I love this feature), excellent guided navigation and search refinement cleanly tucked in the left hand-column, and mash-ups that provide a pretty good alternative source of information about an entry before clicking through. Each of these features is a significant win vs. Google. In fact, I like just about every aspect of the Bing interface better than comparable Google features – even to things I really didn’t expect to be better like the map view and the default settings used there when you switch from a search to a map.

Now none of this adds up to a decision-engine – which is one of the reasons why I don’t like the marketing campaign. If you come out to Bing expecting anything other than a significantly nicer search engine interface, you’re going to be disappointed.

And, of course, search isn’t all about UI – the quality of search results has always been more important than the UI, and it has been the superiority of Google results vs. competitors like Ask that have negated the UI advantages of those systems.

I don’t think Bing is as good as Google – particularly for esoteric searches or cases where you’ve gotten some part of the search wrong. Google is amazingly good at figuring stuff like that out.

I do like Bing’s video search results better than Google’s – which I believe are too weighted to YouTube vs. the owning site. And, as with all the other aspects of the UI, Bing's video search results page is vastly superior to Google’s.

While Bing’s search results aren’t quite to Google standards, Bing is much better than Ask, and in the vast majority of common searches, I find Bing’s results to be roughly comparable to Google’s. On many searches, I couldn’t reasonably decide which result set was better – that’s a fact of life in natural search where it is simply impossible to choose between many thousands of sites based on two or three keywords.

Indeed, it’s the abject stupidity involved in thinking there is one right answer to the vast majority of queries that has always made me believe that improving organic search necessitates improvements in the UI not the search algorithm. Bing is the first real test case of that thesis – and while it doesn’t deliver everything I’d like, it is, in my opinion the best experience out there. I’ve changed my default search from Google to Bing – something I would never have done with Ask or any other search experience.

If Bing does shift share, of course, it’s as likely to be from Yahoo as Google. But any significant share shift will result in something that has never before happened in this young marketplace – Google will actually be under pressure to improve their basic search UI.

Can they do it?

Of course they can. Google has proven they are the best company in the world at building interactive online experiences. So why the pathetic experience you get in their core product and the almost sole source of their wealth? I think the reason is simple and yet not widely recognized.

Google makes money when you click on paid ads not natural search. Natural search results are the glue that keeps users coming back – but they drive no revenue whatsoever. So the more you improve the natural search experience and the more interactive that experience, the greater your risk of shifting share INSIDE your own engine from paid to natural search. From Google’s perspective, that would be very, very bad.

I’ve always believed that as long as nobody could shift share from Google, they had a strong incentive not to improve their natural search interface and heaven knows that’s certainly how they’ve treated it.

So the real threat from Microsoft is that it may force Google to bite their own hand. It will be real challenge to Google if they have to create a system that is better than Bing and that doesn’t shift internal share. Of course, the same holds true for Microsoft. Will Bing hold ads and revenue? To some extent, Microsoft can and must take that gamble. They need share to be relevant – they can worry about monetization later. And, of course, strategically if Microsoft can force Google revenue cuts, they can impede the unending stream of app innovation that paid search on Google currently funds. I have to think that would be a good thing from the view of the boys in Redmond.

Debating the Role for Google Analytics in the Enterprise

Two weeks ago I wrote a blog discussing automated tools for HBX to Omniture migration. In that post, I argued that such tools were essentially deceptive in that they solved a trivial problem (porting an HBX javascript tag to an Omniture SiteCatalyst javascript tag) while leaving untouched the essence of the task (figuring out how to actually take advantage of the Omniture Suite). Part and parcel of most decisions to use automated tools was, I argued, simply the belief that the sites in question weren’t worth measuring particularly well. It also seemed to me that if a site wasn’t worth measuring well, that there was little reason to use an Omniture tag at all – since Google Analytics can do pretty much the same job as a basic Omniture tag will and, of course, GA is free.

I haven’t really changed my mind about this, but two things caused me to further reflect on the possible role of GA in the enterprise. First, I read with interest a post by my colleague Phil Kemelor that dealt with a very similar issue: an organization finding that many of their sites aren’t using Omniture reporting in any robust fashion (or perhaps at all). Phil examines whether GA is an appropriate solution to that problem and pretty much comes to the opposite conclusion I did. So which of us is right?

On the whole, I’m going to say Phil is. But I am too. Here’s why.

Phil’s essential point is that if a web analytics tool isn’t being used, the fault probably doesn’t lie in the tool but in the training. Training is an essential part of rolling out any Business Intelligence system (web analytics or otherwise) – and it does not end with having an Omniture trainer come out and give a big class to a bunch of stakeholders. Training for BI systems actually involves a process of comprehensive training of potential power-users to create organic structure within the organization. It requires the type of tool training that tool vendors like Omniture and Unica provide. And it requires additional business-specific training - often build around your report set - of the sort that companies like Semphonic will provide.

I’d make a further point beyond Phil’s. The report set you develop is part and parcel of the training you provide. These two are inextricably linked in ways that most organizations simply don’t grasp. First, if you don’t provide a good report set, you’ve taken away the easiest, most powerful, and almost always the first point of web analytics exposure. Second, your report set is a chance to establish a useful and well-thought out framework and language around web measurement in your company. This is critical for raising adoption rates. Third, your training needs to build on top of your report set because that’s what most people will be looking at first.

I’ve posted before on what makes a good report set – and frankly, it isn’t what most organizations build (if you're interested, ping me, I have a whole presentation on this that I think is pretty worthwhile).

Phil is dead-on that people tend to blame the tool instead of fixing the problem. And deploying GA – like using an automated tag converter - is just another way of giving up.

That’s one side of the issue. But I also started to think about a special class of sites – commonly called micro-sites – that may really be appropriate for GA and not in a surrendering sort of way.

The company Phil writes about has 2,000 stakeholders around the world – and goodness knows how many sites. We have some clients with similar issues – and for many of them there are two very distinct types of sites. There are full-fledged sites built around major products or lines of business or geographic regions. These sites are almost always large (hundreds of pages at minimum and often thousands or tens of thousands), have multiple functions, and multiple types of success. For any such site, I continue to believe not only that an automated conversion of an HBX tag is inappropriate but that GA is a poor choice for measurement.

But some of our clients also build and deploy many marketing oriented “micro-sites.” These sites are typically very small – perhaps only 20-30 pages or even smaller. They have a single function. They usually have a single success event. For sites like this, GA can provide analysis almost identical to what you could accomplish with Omniture. If you need to integrate behavior from these sites into a larger customer picture, you might still want to tag them with Omniture (and even use a base tag – but please, not one generated automatically). But if they are essentially standalone, you might be well advised to simply use Google Analytics.

For micro-sites like this, most of Omniture’s additional capabilities simply don’t matter. You won’t be coding tons of variables. You don’t need to do a lot of segmentation. You don’t need fancy pathing. Cardinality limits on reports won’t usually come into play. In short, you don’t need an enterprise tool on a site with 30 pages.

What you do need is good solid reporting on traffic, easy tracking of campaigns, and good integration with your search buys. GA can provide that.

So while I haven’t really changed my mind about anything (god forbid), I’ve come to believe that what I said initially isn’t the whole of the matter. If you’re struggling to get stakeholders to care about measurement, the wrong solution is still an automated Omniture tag. But while it’s cheaper to deploy GA, Phil is right to suggest that it’s hardly a real solution.

And while deploying GA may sometimes be nothing more than an admission that the measurement organization, the stakeholders or both don’t much care about or understand serious measurement, it may also be a very appropriate solution for a very common kind of small enterprise site.

Branding and SEM

There is an interesting thread on Jacques Warren’s blog that is quite distinct from my discussion of branding but has a definite relationship to it. Brand marketing is a surprisingly big part of even the most direct aspects of web marketing. Both the post itself and the ensuing discussion are well worth reading.

Jacques' post comments on the high-propensity to brand search on Google (and not just Google).

Indeed. The degree to which Search Programs are “branded” can be shocking given the industry-wide focus on the “long-tail,” on direct-response techniques, localization, landing-page optimization, etc.  Most of these techniques don’t make much sense or difference when 90% of your visits are people typing in your company name (though if you haven’t tried them maybe you can change that ratio).

There is nothing inherently wrong with “branded” search. But I can’t help but feel that it’s sometimes a bit of a con. Search Marketers often talk as if they are the world’s greatest channel for early-stage prospecting even when the programs they run are little more than glorified signposts to known web sites. This isn't always true, of course. Just true often enough to be scary.

I have yet to see a search program where the “branded” component didn’t have fundamentally different behavioral and performance characteristics from the non-branded part. This is common knowledge, of course, and it is pretty much accepted practice to treat and measure “brand” and “non-brand” components as distinct elements. So if your SEM program isn’t making the “brand” component of your search program visible and treating it distinctly, you need to think about changing your SEM program.

And, to me, it’s just another indication of how important brand often is. There are plenty of direct response people who sneer at branding at the same time that the success of their program is largely driven by the strength of a brand.

Everybody’s Talking at You

I have tons of stuff to write about queued up (and not much time), but I wanted to quickly make note of the fact that I've begun podcasting my blogs. The still incomplete series on Form Abandonment and Conversion Analysis is available as is my recent X Change post. Here is the link for iTunes or your Reader:

http://www.semphonic.com/webrss/garyangelonwebanalytics.xml

The podcasts are all around five minutes (okay – around six minutes) and are essentially just the blog slightly transformed and read in my dulcet tones. If you are an avid reader, I wouldn’t really recommend both listening AND reading - the content is pretty much identical. But I’m going to be doing nearly all my blogs (at least the real, informational ones) as podcasts as well. So if you’d prefer to listen while you drive or exercise, then taking the podcast feed might be a better path than straining your eyes each week to absorb yet another 2000 word post.

Talk to you later!

A Perfect Storm

Last week was a sort of "Perfect Storm" for me as I flitted about the country speaking. I do a modest amount of speaking, but once or twice a month is about my norm. So having four different presentations in four days back-to-back was quite an experience.

The fact that every one of these presentations was fundamentally different made it more challenging but also more interesting. I get bored giving the same presentation over and over.

I wanted to quickly mention a couple of those presentations and if people are interested in getting the actual Powerpoint presentations, let me know. You’ll end up on the Semphonic mailing list, but I think they are worth it!

I kicked off the week with a presentation based on my current ongoing series on Form Abandonment. If you’d like a nice Powerpoint summary of the information I’ve been going through on the blog, this would be it.

On Tuesday, I spoke on how to organize the SEM Function. This presentation includes thoughts on the organizational and vendor capabilities required to do good PPC and SEO. This is a topic that’s gotten me in trouble more than once (Agencies seem to hate this) but I think it’s pretty interesting stuff.

Wednesday I was in DC and presented on advanced management reporting. This presentation sums up a lot of our current thinking on reporting. The opening section discusses our overall philosophy of reporting and some of the myths around doing web analytics reporting. This is controversial stuff and I think it’s better heard than read off bullet points. But the presentation also includes stuff on Analytic Reporting, focusing on getting the right numbers and getting them right, and placing reports into a good conceptual framework. To me, this was the most interesting presentation I did and I really enjoyed doing and talking about this one.

Finally, on Thursday I was back in San Jose at SMX (beautifully run, by the way) talking about Web Analytics Tool Evaluation and SEM Analytics. This presentation is really just a distillation of the White Paper I wrote on Tool Evaluation a while back that is on our web site. In truth, I thought the Panel discussion there was more interesting than my presentation - tool evaluation just isn't that interesting unless you happen to be buying a tool. But there’s some new stuff on good SEM reports and capabilities and if you’re looking for an easily digestible version of the Tool Evaluation White Paper or are just interested in some thoughts on web analytics SEM capabilities this one would be nice to have.

If you’re interested, just drop me a line to let me know which one you’d like and I’ll send it out!

Go Ahead and Flush

In my post on the Microsoft acquisition of Yahoo, I mentioned that one of the areas where I thought Google was vulnerable to attack was the Content Network program (AdSense). Like most content network programs, there is a lack of transparency about what you’re buying and a misplaced reliance on content matching technology that often leads to very poor program results.

That’s partly why I was interested in the recent Click Forensics study that showed extremely high rates of click fraud on content networks.

Here’s a link to an article describing the study on MediaPost – and I’ve excerpted the key passage for this discussion here:

“The average click fraud rate of PPC advertisements appearing on search engine content networks, including Google AdSense and the Yahoo Publisher Network, was 28.3% in the fourth quarter. That's up from the 19.2% average click fraud rate for the same quarter in 2006, and 28.1 percent for third-quarter 2007.”

That is an astonishing number. Hardly believable. Probably not even true.

The content networks generally admit to a much, much lower level of click-fraud. And, of course, Click Forensics has almost as much of an incentive to overstate fraud as the networks do to understate it.

I’ve read posts by Google and posts by Click Forensics, and all I can say is that neither make a whole lot of sense to me. I’m just not in a position to judge the reasonable accuracy of these competing claims.

I can say this. I don’t think any significant brand should ever spend advertising money without knowing where that advertising is going to appear. I can’t think of another channel where the sort of large and highly-sensitive brands that we typically work with would even contemplate such a thing.

That’s why I’ve always wished that advertisers bought in content networks using an exchange model – where reputable publishers (sellers) are matched with reputable buyers and the exchange takes a small cut on every transaction. I think it’s better to have transparency around placement and to select appropriate sites than to get whatever minimal boost in relevancy that content-matching technology actually provides.

I also know that our experience analyzing campaigns on content networks had pretty much convinced us some time ago that they were usually a poor investment for clients. We tried valiantly to exclude bad and underperforming sites, but new bad and underperforming sites always managed to find their way right back into our buy. In the end, we pretty much gave up. No amount of good analysis can compensate for a bad system. It turns out to be a lot easier to find the few sites that generally drove what good traffic there was and have our clients buy sponsorships directly from them.

This underscores a critical point - even where a content campaign appears to work, it's often the case that a few good sites are responsible for nearly all the good clicks. Why not just buy inventory on those sites directly?

That doesn’t mean that levels of fraud are anything like what Click Forensics claims. It may just mean that the system itself is over-bought, doesn’t work all that well, and forces you to buy against lots of dumb money overspending for clicks. That isn’t quite fraud but from the buyers perspective, it might as well be.

Most of the very sophisticated buyers I talk to don’t buy on the content networks at all. And yet, the total dollars spent on content networks continues to grow. Which makes me wonder: who is spending all that money?

I don’t really know, but I have two theories.

First, I think a fair amount of content network buying is from unsophisticated small buyers. As a small buyer, you don’t always have a great deal of choice. Sponsorship of a significant site is unaffordable. Search on niche terms often provides access to very limited traffic. Banners are harder to buy and take more creative effort. So content networks may provide the only reasonable avenue for scaling a small, niche online buy.

It’s my guess that another fair amount – and maybe the bulk – comes from program buyers anxious to spend every dollar they can. Let’s face it, even with the “long-tail” you can only buy so many search terms that will produce any clicks. It often surprises companies how small the click numbers are on even seemingly logical terms. But no such problems exist with content networks. Content networks are a vast sponge that can absorb almost any budget – no matter how large.

I can’t tell you that ¼ of all your content clicks are fraud. I don’t really believe it. But I do believe that managing a content network program takes lots of special care if you are going to do it at all. So if you are a significant investor in PPC and don’t manage your campaigns, here are a few pieces of advice.

• Find out what percentage of your buy goes into content networks.
• Find out how that buy compares in terms of performance to your search buy.
• Make sure than any content network buys are measured only against absolutely real conversions (purchases and valid lead forms) – not against traffic or even conversion proxies.
• Use your web analytics tool to find out where your content network ads are appearing and measure their effectiveness by site.
• Make your buyer eliminate continually the sites that you don’t want your brand on.

If you’ve ever gambled at all seriously, you realize that one of the most important facts to know about any game of chance is the percent the house takes off the top. The larger that percentage, the harder it is to make money. In Las Vegas, casinos typically field games where the house takes a cut ranging from less than a percentage point to as much as 5-7%. We all know how well that works out for them.

Because of the total lack of transparency in content network advertising, nobody knows what the house percentage actually is (Google, Yahoo and whatever click-cheats there are constitute the house). For the most part, you don’t actually need to know the house percentage. You just need to make sure you aren’t one of the suckers spending all that money without making sure it actually works for you!

So here’s my best advice if you’re truly tempted to use the content networks in Search Marketing. Don’t do it at all unless you are an expert buyer. And if you aren’t going to take the time to measure the results with all due care, then you should just take that money and spend a weekend in Vegas instead. You’ll have a better chance of getting a return on your investment and you’ll have a hell of lot more fun.

Covario - SemDirector and the InflectionPoint 2008 Conference

I just returned from the InflectionPoint 2008 Conference down in San Diego.  What’s InflectionPoint? It’s Covario’s (until this week SemDirector) User Conference. Covario has a solution set for SEO and PPC reporting that is particularly appropriate to large enterprises trying to manage programs that are spread across divisions, engines, channels, countries, etc. It’s a high-end solution that provides a level of enterprise-wide search reporting (Paid and SEO) and online channel management that just isn’t available elsewhere.

This last week was a big one for Covario. They had their annual conference. They changed their name. They announced a 16M funding round. They rolled out some cool new products. And, drum-roll please, they announced a partnership with Semphonic.

Well, maybe that wasn’t the lead. But the partnership is an interesting one – especially since some people might think we are competitors. Fortunately, neither they nor I really see it that way. Here’s the thinking behind the partnership.

Covario’s solution requires companies to feed information about online success back into their reporting solution from the web analytics tools. Makes sense, right? But a lot of their clients (and keep in mind these are mostly very complex multi-channel enterprises) have a hard time doing this: not only do they struggle with getting the web analytics set up right, they struggle with figuring out what the proper optimization goals need to be.

That’s our bread-and-butter, of course. In addition, Covario’s clientele heavily overlaps our areas of focus. They have more technology and probably less Financial Services and Media than Semphonic, but the types of clients (multi-channel, international, complex) and the focus on sites without traditional eCommerce events is very similar. So we’ll be providing these implementation services to Covario clients to make sure they get the best advantage out of all that slick enterprise reporting!

And while Semphonic does have a SEM reporting tool, it’s nothing like Covario’s. CampaignTracker isn’t an enterprise data integration tool – it doesn’t even remotely tackle that function. It isn’t an Enterprise tool at all, and it isn’t for PPC or SEO channel management. It has a fundamentally different purpose, different price-point and different set of users. These two products may live in the same galaxy, but they don’t co-exist in the same solar system. And let's face it, Semphonic is a consulting company - that's the heart, soul, core (and revenue) of our business. So when we started talking together, I think both sides quickly realized a partnership was possible.

Best of all, I really like the people there. They are very easy to talk to and do business with. They are smart. And their customers obviously like them a lot. I think the whole corporate culture was apparent in InflectionPoint; it’s a culture that I like and suspect most clients will too.

Their similarity in style to Semphonic was very obvious at InflectionPoint. In fact, there were many aspects of the conference that reminded me of X Change. Since we started X Change, I’ve developed a whole new interest in and (grudging) appreciation of Conferences. I notice lots of things I never thought about before.

This increased attention is a bit ironic, because I mostly don't like Conferences. I’m not much of a schmoozer; I don’t really enjoy speaking (funny, because I love talking!); and I’ll be honest and admit that I find most conference presentations a bit tedious. Then, after a long day of being kind of bored, I’ll return home to the obligatory mountain of emails we all get and have to do my work.

This complaining is mostly a way of getting to the fact that I thought InflectionPoint was pretty damn good. A couple of notable wins:

• Extremely limited self-promotion
• No speakers over breakfast, lunch, dinner
• Very polite involvement by sponsors (no time-wasting fluff)
• Outstanding attendees

Unlike most company conferences, the chest-beating was amazingly minimal – they let their customers do almost all the talking. A great strategy since the customers had really good things to say. This reminded me of X Change – but since X Change isn’t designed as a Semphonic customer-only event I think their restraint is even more surprising.

Like X Change, they didn’t pollute breakfast and lunch with speakers. I hate having to speak while people eat. And I hate having to eat while people speak. They served good food, too. Major kudos!

They gave partners sponsorship slots but didn’t waste everybody’s time with long-winded sponsor plugs. We don’t do any sponsor slots at X Change, but this was as nicely handled as I’ve seen. Good job by the sponsors as well.

Most important - they brought together a really good group of people. Part of this is the fact that their products play mostly to very large efforts in very large companies. So they get really serious people involved. The panel I was on included truly senior interactive marketing folks from iVillage, Lenovo and Siemens plus David Falls from DoubleClick. And that’s pretty typical of the whole Conference make-up. Outstanding.

I was deeply impressed by the sophistication of that group. Not something you hear me saying every day. I especially enjoyed chatting with Barry Kresch – whose particular challenges at iVillage are very common in our media client-base. It seemed like we’ve done a lot of similar work and come to similar conclusions about how to effectively measure media sites.

I’m still not that big a fan of panel presentations. They feel a bit like political debates – more sound-bites than real discussion. That’s why I love X Change’s Huddle format. But I’m glad X Change is special that way – we need some kind of special sauce!

Ultimately, what impressed me most about the conference was the emphasis on providing value to their customers. As a consulting company, that makes me feel very comfortable with Covario. Not every software company spends more time listening to their customers than talking at them. Not every? Not many.

It must have been an amazing, exciting, draining week for the people at Covario – with so much news plus a big conference.  Congratulations guys. I hope this is the beginning of a beautiful friendship.

Search Engine Marketing (SEM) Optimization and the Politics of Change

With so much happening in the web analytics world it’s a bit difficult to even remember that I had promised a capping post in the SEM series on the political nature of SEM Analytics and how it might best live in an organization. Since then, I’ve written extensively on the VS acquisition and Score. But I haven’t even touched major issues like the recent purge at WebTrends.

I don’t know the people at WebTrends particularly well, and we have fewer clients on WebTrends than on VS or Omniture. So don’t look to me for an insider’s take. My suspicion is that negotiations with Omniture and subsequent VS acquisition unearthed significant differences between management and investors regarding exit strategy. It’s easy to see how that could happen and easy to understand why it would be fatal to working together. I’ve worked with VC’s before, and while you and your investors don’t have to see eye-to-eye on every aspect of running the company, you MUST agree about exit strategy. If investors think management isn’t on the same page about selling the company or yielding control, then change is inevitable and necessary. That’s just speculation, of course. It’s a bit like guessing why a marriage failed – sometimes it’s pretty obvious but sometimes it’s a thousand little things you never see from the outside.

It's unwise to ignore the political dimension in most real-world decisions. As analytics practitioners, we tend to believe in the power of an “objective” view of reality – one in which rules of reasoning and evidence are used to structure decision-making and eliminate the most common sources of error. Certainly I believe that most organizations and most decision-makers can and should work to improve the quality of their decision-making. Not the least because business and marketing decisions are quite often mainly – if not entirely – political.

And nowhere is this truer in our world than in the Search Engine Marketing Analytics space. Here’s what I wrote in my last SEM Analytics post:

“SEM Analytics lives in a peculiar place – both in terms of web analytics and Search Marketing. As a web analyst, you aren’t generally responsible for optimizing your SEM program – either PPC or SEO. And there’s a pretty good chance that the people that are in charge of that optimization aren’t going to welcome outside measurement. That makes life a lot more difficult and challenging than it really ought to be. “

The heavy reliance on vendors for both SEO and PPC and complications around brand have created silos in many organizations that are extraordinarily difficult to cross. I have seen companies where online managers were not able to get information about their own PPC buy because their agency didn’t want to release it to them and was responsible to a higher power or a different organization.

That’s silliness of an extreme sort. But it’s not particularly uncommon. And what passes for normal in most organizations is a nearly complete segregation of SEO, PPC, Brand Marketing, Channel Marketing and Web Analytics into autonomous and often uncommunicative fiefdoms.

I’m going to talk about Brand at some later date (one of my planned posts dating back to X Change huddles!). But what about Search Engine Marketing – what’s the right way to structure your organization?

In a previous series of articles and rather controversial posts, I asserted that it was perfectly reasonable for an organization to bring Search Marketing in-house. I continue to believe that. It’s also perfectly reasonable to use an Agency. We see so many mismanaged programs of each type that it is difficult to choose which option is worse.

One of the advantages to going in-house is that you will typically get much better cooperation between the SEM team and the analytics team. This isn’t universal, but it is frequently true – especially if a measurement team is already in place when you start building your SEM team.

But whether you are using an Agency or going in-house, it’s important to understand how your web analytics team should interact with your buyers. The web analytics team is not going to be in a position to perform day-to-day optimizations. That’s the buyer’s job. 

The web analytics team should be responsible for setting the framework within which the buyers work and tracking the larger picture of how PPC and other marketing programs are interacting and performing.

It should be the job of the web analytics team to determine the appropriate optimization points for the PPC buying program; to help make decisions about new resource allocations (SEO, PPC, Banner, other); and to play a watch dog role on the program by insuring that gross mis-optimizations don’t exist and that channels are playing well together.

In particular, I think it’s important that the web measurement team be responsible for management reporting on SEM (and other online marketing) channel performance. Allowing teams to report up to management independently will practically insure that there is no way for management to compare their true effectiveness.

It’s these last two functions (watch-dog and reporting) which are likely to create resentment. Many a program buyer will insist that if they are meeting their goals they should be left alone. Maybe. There is no one right way to manage – most people and vendors respond better with oversight, others don’t. However you choose to manage, though, the necessity for STANDARDIZED performance reporting across channels is absolute.

It’s essential to keep in mind that web analytics practitioners have tools and data that usually aren’t available to your buyer. A good buyer, realizing this, should be open to outside measurement that increases their knowledge of what’s working and what isn’t. If your buyer resists this, perhaps you need a new buyer. I hate to say it, but often, the less competent the buyer, the more resentful they are of measurement.

With SEO, the situation is usually more straightforward. Unlike PPC, SEO doesn’t have real-time optimization and SEO vendors don’t deploy their own tags for measurement. Instead, you’re more likely to have to deal with a general lack of understanding about the need for measurement in any form.

Certainly one of the most common mistakes in SEO is to have vendors measure their performance by tracking position on search terms. This strategy creates serious mis-optimizations of SEO effort.

You should be managing SEO to total quality of traffic provided. And to do this, you are responsible for providing data back to your SEO firm.

Remember, SEO practitioners don’t have alternative data sources. It’s your job to make sure they know what the optimization goals are (ala PPC) AND how well they are achieving them. No vendor, however astute, can somehow intuit this data or know what’s working and what isn’t. Like they always say in marriage counseling – you have to say what you want, your partner isn’t a mind-reader.

It’s also true that SEO engagements often stall after a very productive first round. Using analysis techniques like the SEO Holes study, your measurement team can help focus SEO efforts as well as track their subsequent success.

Unlike PPC buyers, SEO vendors are unlikely to push for their own measurement solutions – but perhaps even more likely to resist the need for measurement at all. Make sure your vendor (or in-house team) buys into your measurement paradigm – not the “# of terms in high positions across engines” that otherwise passes for SEO measurement.

Building effective communication between your web measurement team and your SEO and PPC teams ought to be a significant priority for anyone serious about online marketing. That means carving out a clear role for web measurement and making sure you have teams or vendors willing – even excited - to accept those roles. And don’t neglect to back up those decisions with hands-on management to insure that the behaviors become institutional. Left to themselves, most organizations will quickly revert to the comfort-zone of “leaving well enough alone.”

I see far too many bad programs to believe this works. Measurement is your safeguard against complacency; your defense against both institutional and vendor arrogance. Use it.

Me, Me, Me

Under normal circumstances I wouldn’t have posted the comment from Emerson Hartley on my last SEM Analytics post since it is more of a commercial than a comment. That’s an obnoxious practice in general and I try not to encourage it. But I decided to post it anyway because I wanted to comment on it. Here’s Emerson’s comment:

Multivariate testing and optimization should take place outside web analytics. what's more important is that optimization leverages crm, approval and other data rather than traditional WA data. Only in this manner can we optimize throughout the customer lifecycle. what i'm saying is, as a customer of memetrics and a user of their solution, it's more important for us to know that a particular ppc ad and landing pages converted people who were approved for a credit card, activated and used it than any type of integration into our omniture package.

Take away the memetrics commercial (about which I have no opinion pro or con) and you have an attitude that’s actually fairly common and has led to the siloing of multivariate testing and landing page optimization from the rest of web analytics. So what’s wrong with this argument?

The first part is unobjectionable. If you need approval and back-office data to optimize correctly then you need approval and back-office data to optimize. You’ll never hear me arguing against using the right optimization points. But this raises an interesting question – if you need those variables to optimize landing pages then presumably you need them to optimize anything else as well. So what in god’s name are they doing with Omniture? Reporting page views? And why don’t they move that data to Omniture so they can optimize something besides Landing Pages?

If you can move your customer data into some other system, you can move it into your web analytics package. And if you absolutely refuse to ship your data offsite, then get a web analytics package onsite.

The idea that landing pages are somehow distinct and require different data to optimize than any other aspect of web navigation is absurd. The all too common idea that if you’ve moved your key conversion optimization data into your PPC or Landing Page tracking then you don’t need to move it into your web analytics package is just plain stupid.

There is, as I mentioned in my post on this topic, a very good reason why multivariate testing HAS taken place outside web analytics. The web analytics (and CMS) vendors have done a terrible job of supporting it. But that isn’t an argument about WHY it should be that way. And the argument Emerson makes in terms of data integration is not reasonable either from theory OR real-world practice.

Landing Pages, SEO and PPC

Part 12 in a Series on Analytics for Search Engine Marketing

This is the last analytic topic in my extended series on SEM Analytics. In the course of this series, I’ve covered topics ranging from getting setup for measurement, selecting the proper optimization points, measuring the cross-channel effects of PPC and SEO, and a variety of different techniques for measuring and optimizing specific parts of a SEM program. A good many of these analytic techniques are somewhat different than the traditional view of web analytics as a hunt for solutions to particular site or campaign problems.

There’s a good reason for that. Much of the basic optimization in SEM programs – especially PPC programs – happens outside the web analytics solution. Your bid management solution is the natural locus for most of the basic optimization. Web Analytics is often supplementary. This is especially true since web analytics tools often lack the essential data (cost per click, click rate, impressions, position, etc.) necessary to make sense of the actual marketplace. Even where this data is available, it’s not clear that the web analytics solution provides reports and analysis that are particularly useful for really understanding and optimizing this data.

The topic I’m taking up today is similar – though for slightly different reasons. A great deal of effort in the past two years has been focused by serious PPC buyers on the optimization of the Landing Page. It’s become something of a specialty – with whole companies devoted to this particular art. Multivariate testing, in particular, has grown up as a discipline largely driven by the particular needs of Landing Page optimization.

What’s somewhat ironic is that multivariate testing is done largely outside traditional web analytic packages. That seems ridiculous, but if it is ridiculous, it’s not the fault of users – who really haven’t had much choice in the manner. Multivariate testing capabilities have tended to be siloed solutions that integrated their own tracking capabilities. This made it easier for MV vendors to sell their stuff to companies – and it has been hard to argue with since WA companies lacked credible capabilities in this space. Nor was it possible to generate good MV tests with most CMS solutions and then piggyback measurement on top.

This may all be changing, of course, with Omniture’s recent acquisition a very positive sign. But then again, it may not be changing as much as we all hope. It seems to me that the proper place to put MV serving capabilities has always been the CMS – not the WA solution or any 3rd Party solution. Ideally, I’d like to have good MV serving capabilities in the CMS and then be able to integrate my analytics solution easily on top of that to track the testing. This would subsume multivariate testing into the same basic content generation/serving/analytics paradigm that is used for every other part of the web site.

Any other type of solution seems jury-rigged.

But for now, that’s not the world we live in. Most companies Semphonic works with have either turned to a 3rd Party Landing Page optimization vendor deploying custom multivariate testing or have deployed the Google Site Optimizer and are managing multivariate testing there.

In either scenario, you won’t be using your base web measurement reports to track landing page success – you’ll be using your multivariate test suite to both serve content and measure outcomes.

This doesn’t mean, however, that there aren’t some interesting aspects to Entry Page analysis using web analytics.

One fairly interesting type of analysis that is part of understanding the overall relationship between PPC and SEO is to measure the incremental value of controlling the Landing Page. With organic search optimization, the pages that are most relevant (according to the Search Engine) become the de facto landing pages for your sites. These may not, however, be pages that are particularly strong in driving to the core business goals of your site.

There’s a pretty good chance that nearly every page on your site will show up somewhere in the Google rankings for key terms. So, in that sense, nearly every page is a Landing Page for many key terms. But in the real world, only pages ranked in the top 30 or so listings will get any volume. And only listings in the top 10 will usually get enough volume to matter.

When you buy PPC ads, on the other hand, you control the landing page [Can you multivariate test on heavy organic Landing Pages? You can, and it’s certainly an option worth exploring. I can’t make head-nor-tail of the Google rules when it comes to this practice and I have no idea what would constitute abuse – if anyone can explain this I’d welcome the information!].

To really understand the impact of cannibalization, you need to understand the differential in effectiveness between your organic and paid landing pages. If there is a significant differential, then you might want to ENCOURAGE cannibalization. In fact, it’s conceivable that you might be better off REDUCING the organic position of the page. Otherwise, you can’t keep people from going to the wrong place on your site.

This isn’t an issue I’ve heard discussed – and I know of no SEO professional who has ever been asked to reduce a page rank (for an owned site, at least). But the logic is impeccable.

The measurement of differential effectiveness is reasonably straightforward. You can simply measure conversion (or appropriate optimization measure) for PPC sourced visitors to the PPC Entry Page vs. conversion for the Organic Entry Page.

As with so many web analytic techniques, however, there is a danger of self-selection in this analysis. It may be that visitors who use PPC or organic listings have a built-in differential in quality. If so, then your straightforward comparison of conversion will miss the actual significance of controlling the Landing Page. Here’s where piggy-backing this analysis on top of a cannibalization analysis is interesting. Because when you measure the shift in traffic from cannibalization, you can also track whether or not there is any change in measured conversion effectiveness.

One aspect of PPC and SEO integration that shouldn’t be overlooked is simply taking learnings from your PPC optimizations and applying them to popular SEO landing pages for the same keywords. This isn’t a task that demands any high-level of analysis – but it is the kind of thing an analyst should always be prepared to do. Not all of web analytics requires a web analytics tool. If you know that certain messages and link paths work well for PPC-sourced entries, there’s a very good chance they’ll work well for SEO sourced entries as well. And many times, this type of knowledge can be applied to Entry pages without significantly altering their SEO profile.

In a similar vein, studying the differential effectiveness of SEO Landing Pages (Bounce Rate and Conversion Rate for SEO Entry Pages Compared) AND the SEO to non-SEO differential effectiveness of Pages (Click-Thru and Conversion Rate when SEO sourced and when accessed in any other manner) can yield considerable insight into two classes of SEO misoptimization. First, you may find pages that simply don’t perform very well relative to other SEO entry points for the same or similar keywords. When you find these cases, you can try to either figure out design cues (links or content) that you can borrow or you can focus SEO efforts on the better performing page.

When you find pages that perform much worse when accessed via SEO, you need to consider whether the design template is appropriate for a Landing Page and when types of content might be necessary to improve it. Keep in mind that the bare fact of a page performing worse when accessed via SEO isn’t all that interesting. What’s interesting is comparing the differential between Page X accessed via SEO and otherwise and Page Y accessed via SEO and otherwise.

In general, I’ve found that SEO is less carefully measured, less optimized for conversion and less tested than PPC. So the web analyst is quite likely to find numerous places where SEO optimization is misguided. The frequently monomaniacal focus on traffic generated by position in SEO has created an atmosphere where vendors get rewarded for driving traffic whatever its quality. As I’ve mentioned before, this practically guarantees bad traffic or – in the case of Entry Pages - bad site entry placement or bad page performance.

This last point sums up so much of what the web analyst needs to keep in mind when it comes to SEM analytics. The simple fact of source is one of the key behavioral facts the analyst has available. And it turns out that it regularly makes a big difference. But SEM Analytics lives in a peculiar place – both in terms of web analytics and Search Marketing. As a web analyst, you aren’t generally responsible for optimizing your SEM program – either PPC or SEO. And there’s a pretty good chance that the people that are in charge of that optimization aren’t going to welcome outside measurement. That makes life a lot more difficult and challenging than it really ought to be.

I’ve covered many aspects of SEM analytics from a practical perspective. In my final post in the series, I’m going to address this peculiar political position and what organizations can do to make web analytics a more effective part of their Search Engine Marketing Effort.