The actual conversion process involves testing a landing page leading to a signup or filling in of a form or a thankyou page in the event of a successful sale of a product. With Website Optimizer, you can set up experiments that involve Multivariate testing or A/B testing to track which version of the landing page is pulling in the desired results.

The top ranking of a client’s website on the Google SERPs is often used to justify the success of a SEO firm’s efforts. Clients are also obsessed with their site rankings especially in the face of their competitors flying high on the SERPs which is a pretty natural human reaction. But rankings are not the ultimate way of judging the success of a website. Conversion of visitors into customers or visitors taking the necessary action is an ideal way to measure the success of a site’s SEO or PPC efforts.

The testing I did for a client involved switching between an original landing page and a completely new landing page in a PPC campaign. This is an example of A/B testing which is simpler version of testing with Website Optimiser. It is ideal to start with this test if you have lower traffic volume and gain faster results. The conversion in my experiment involved filling in a form.

Multivariate testing involves using the same landing page with changes to headings, call to action, placement of menu items, logos etc forming the variation of the original page. In effect, you are testing between variations of the same page to see which one results in the user taking the desired action and hence achieving the optimum measure of conversion.

When setting up the A/B experiment, fix the original landing page and a new landing page without any room for doubts as you will switch the traffic between these two pages. Then the conversion page is the page with the form that needs to be filled. It is handy to have a printout of the Website Optimizer help section for reference.

Steps to follow when setting up the experiment:
1) Choose your original landing page (let us denote it as OLP)

2) Create a totally different landing page (this is an A/B test) (denoted by DLP)

3) Choose your conversion page (denoted by CP)

4) Now comes the crucial part of installing the tags. These are pieces of javascript code that need to be installed on the different pages.
a) The control script is installed only on the original landing page, placed just after the head tag.
b) The tracking script is installed immediately before the closing body tag on all the three pages namely, OLP, DLP and CP.

5) Website Optimizer’s validation tool will check for errors in the installed tags. It does this in two ways. You can provide the URLs to your original and variation landing pages and the conversion page. If these pages are externally visible, Website Optimiser will access them and point out errors if any.

Alternatively, if Website Optimizer cannot access the pages on your live site, you can save and upload the HTML source files for your landing pages and conversion pages and allow Website Optimizer to validate them.

Once Website Optimizer validates the installed tags, it is time to start your experiment. Make one final check of your experiment settings. Once the experiment is started, you will not be able to change the experiment parameters. You would do well to ensure that the settings are as you intended them to be.

6) The last but not the least step is the % of traffic you wish to switch between your original landing page and the new landing page. I started off diverting 75% of the traffic to the new landing page and retaining the original landing page 25% of the time. Google advises to have a 100% switch in order to run the experiment faster and hence obtain results faster. It is entirely up to you to make the final decision.

After the activation of the experiment, when you click on Website Optimizer, you can see the name of the experiment (the naming is part of the guided process that Website Optimizer leads you through when setting up the experiment) and the Status column says – Running Collecting Data. There are links below it to edit settings and view the report. (screenshot attached below)

The View Report setting allows you to compare the performance of the original landing page and the new landing page. It shows how well or how badly the newer version is performing against the original. In the early days of my experiment, I saw a yellow colured bar which showed that testing was inconclusive.

There are three other columns namely Chance to beat original, Observed Improvement and Conversion/Visitors. As my experiment progressed, the new combination worked better and finally in 15 days, the green colored bar indicated that the new combination is a clear winner.

The Website Optimizer documentation says that if the Chance to bear Original column exceeds 95%, then you can safely assume that your new combination is a winner.

In my experiment, I had to install the script tags on a site run by WordPress. A great Website Optimizer WordPress plugin created by Filippo Toso makes it a breeze to install the tags.

NOTE:
The switching of the original landing page and the new landing page is applicable to all forms of traffic, be it organic or pay per click.

For example, if the original landing page forms part of your website and your new landing page is created specifically to target a PPC campaign and therefore not visible to users landing on your site through an organic search listing, remember that the original landing page will be switched to show the new landing page according to the traffic percentage value that you set in your experiment.

Again, depending on the complexity of your experiment, it can take more or less time for Website Optimizer to judge the winning combination.

Ravi Venkatesan is a senior SEO consultant at Netconcepts, an Auckland seo consultancy offering top quality organic search and paid search services to its clients in New Zealand and Australia.

Link bait can be a powerful tool, especially when used in conjunction with a powerful social media network. Like many things however, it generally comes at a cost. Even if there aren’t any direct costs, such as engaging writers or paying for services directly, there will at least be direct time if you or your company creates the link bait yourself, and potentially in-kind costs, such as helping to return the favor or similar for someone else within your social media network.

So at the very minimum, without direct financial costs, the time carries some level of cost…be it $60 per hour, $600, or whatever monetary value you place on your time. With that in mind, it is critically important to at least attempt to estimate the costs involved and measure the impact achieved to arrive at some kind of estimation on the return of the link bait campaign.

Like any kind of off-site effort, it is important to realize that hard and fast numbers may be a measurement luxury. But something is most likely better than nothing, and choosing not to monitor and measure is akin to tossing money out the window while driving down the highway as fast as you can…exhilaration with potentially little or nothing to show for it.

Before engaging in a link bait campaign, it is important to understand the goal or goals of the campaign. Fortunately, goals and measurement often go together, so figuring out one puts you well on the path to figuring out the other.

While the specifics may vary, such as whether the link bait exists on the site or is linked to from where the link bait is launched, some variation of monitoring and metrics can be adapted. Here are three common goals and metrics you may want to consider:

• Traffic – such as direct traffic from people clicking links leading directly to the site, which might be measured based on traffic referrals, or indirect traffic based on brand or topic awareness that leads people to come directly to the site or via searches related to the topic, which might be monitored for increased direct traffic or related search referral traffic.
• Rankings – may be related to the link bait piece itself, an on-site destination based on anchor text signals within the link bait, or combination of both, which might be monitored based on rankings movement over time for the keyword phrases and/or the destination URLs.
• Links – building out the site’s link profile, which is especially important when the link bait resides on the site, and in many ways, probably relates back to traffic or rankings, which might be measured based on tracking overall or specific changes in the link profile.

These aren’t the only goals or metrics, but they are probably the simplest to begin measuring, and probably quite similar in theory to what you already measure. It’s important to understand that the monitoring and measuring of these metrics will need to happen over a period of time. While big bang social media driven link bait campaigns are often positioned to hit critical velocity in an extremely short period of time, even then, the most meaningful impact may well come after the initial spike.

This is of course why a clear set of goals at the outset, a plan for monitoring and measurement, and an ongoing measurement and analysis process is so critical to define. Without it, valuable resources of time, money and focus could be expended for weeks, and most likely months, before understanding the true value and cost.

Out of the gate, you may not have specific numbers to hit, aside from comparing to other tactics, but over time, you’ll begin to better understand what works and what doesn’t, which venues and efforts pay for themselves and which leave you with an empty feeling in your stomach…and possibly wallet.

Can you keep a secret? There is an unfair advantage. Amazon is applying conditional 301 URL redirects through their massive affiliate marketing program.

Most online merchants outsource the management and administration of their affiliate program to a provider who tracks all affiliate activity, using special tracking URLs. These URLs typically break the link association between affiliate and merchant site pages. As a result, most natural search traffic comes from brand related keywords, as opposed to long tail keywords. Most merchants can only imagine the sudden natural search boost they’d get from their tens of thousands of existing affiliate sites deeply linking to their website pages with great anchor text. But not Amazon!

Amazon’s affiliate (“associate”) program is fully integrated into the website. So the URL that you get by clicking from Guy Kawasaki’s blog for example to buy one of his favorite books from Amazon doesn’t route you through a third party tracking URL, as would be the case with most merchant affilate programs. Instead, you’ll find it links to an Amazon.com URL (to be precise: http://www.amazon.com/exec/obidos/ASIN/0060521996/guykawasakico-20), with the notable associate’s name at the end of the URL so Guy can earn his commission.

However, refresh that page with your browser’s Googlebot User Agent detection turned on, and you’ll see what Googlebot (and others) get when they request that same URL: http://www.amazon.com/Innovators-Dilemma-Revolutionary-Business-Essentials/dp/0060521996 delivered via a 301 redirect script. That’s the same URL that shows up in Google when you search for this book title.

So if you are a human coming in from affiliate land, you get one URL used to track your referrer’s commission. If you are a bot visiting this URL, you are told these URLs now redirect to the keyword URLs. In this way, Amazon is able to have its cake and eat it too – provide an owned and operated affiliate management system while harvesting the PageRank from millions of deep affiliate backlinks to maximize their ranking visibility in your long tail search query.

(Note I’ve abstained from hyperlinking these URLs so bots crawling this content do not further entrench Amazon’s ranking on these URLs, although they are already #4 in the query above!).

So is this strategy ethical? Conditional redirects are a no-no because it sends mixed signals to the engine – is the URL permanently moved or not? If it is, but only for bots, then you are crossing the SEO line. But in Amazon’s case it appears searchers as well as general site users also get the keyword URL, so it is merely the affiliate users that get an “old” URL. If that’s the case across the board, it would be difficult to argue Amazon is abusing this concept, but rather have cleverly engineered a solution to a visibility problem that other merchants would replicate if they could. In fact, from a searcher perspective, were it not for Amazon, many long tail product queries consumers conduct would return zero recognizable retail brands to buy from, with all due respect to PriceGrabber, DealTime, BizRate, NexTag, and eBay.

As a result of this long tail strategy, I’d speculate that Amazon’s natural search keyword traffic distribution looks more like 40/60 brand to non-brand, rather than the typical 80/20 or 90/10 distribution curve most merchants (who lack affiliate search benefits) receive.

Brian

Despite my griping, they rolled it out anyway with this feature unchanged, and they made it impossible to view the data through the old UI as of July 19th. They report adding more requested features, but how about adding back some of the functionality they destroyed? Perhaps they’re more involved in getting the daily data processing issues resolved, and admittedly I’d agree that would surely be a higher priority. I’m just still flummoxed because it seems so unnecessary to revoke good functionality in the first place.

I’ve found yet another irritating change that I consider to be even more serious: you apparently can’t view the data in monthly units – only daily:

(click to enlarge)

Why did they revoke the ability to visually compare monthly periods?!? Most search marketers I know like to compare overall figures from month to month since it tends to reduce some of the spikiness of short-term bursts, and lots of folks are using monthly billing cycles and such.

If I’m mistaken and there’s some where to set the period to display monthly, I hope someone will let me know. I hunted and hunted, and checked their help section to no avail. If they really did revoke monthly display, I can only reiterate further how bad this so-called “upgrade” really was! All glitz with little beneficial substance.

The Analytics team should borrow some of the members of the Google Maps team, since comparatively the Maps team seems to get it right a lot more lately.

One of the main things that I’m pleased about was my effort to be as objective as possible in writing the article — not only did I want to report on what is going on in local-oriented domaining, and who’s involved, but also to provide some concrete conclusions and recommendations which people could take away. I was upfront in disclosing my past negative bias about domaining, and in the course of writing the article I found that I had to revise my assumptions a few times over – in favor of Domaining, actually. Working off and on, I wrote the article over the course of about two months.

While doing the research, I became aware that the Domaining industry seems to have a bit of “younger sibling complex” — as an industry, they wish to be considered a respectable, bona fide line of business. Unfortunately, they have a few things which have been hampering that aim to some degree:

1. Many people have been exposed to cybersquatting, typosquatting, and trojan-horse sites which have damaged the public’s perception of domainers. While a number of companies have been springing up which have highly ethical practices and do only high-quality domaining, the industry’s been sand-bagged by the past sins of a lot of anonymous individuals who have exploited domain names for a fast buck.
   .
2. The industry has been disrespected by a degree due to the perception that domaining isn’t very sophisticated. This isn’t really accurate, since the top domaining companies frequently have very sophisticated tools and processes for selecting domains, registering domains, hosting and deploying content on domains, and algorithms for estimating the traffic potential of domain names. Further, on the face of it, that bias isn’t really fair — someone who’s developed an elegant, simpler method for making money should be admired so long as it’s not a criminal method.
   .
3. There’s little independent reporting of domain industry traffic figures. This is a case where there are significant challenges to independent reporting agencies in identifying “domaining” traffic from other types of traffic, since one has to assume varying amounts of user motivations in going to domains directly — this is something that Danny Sullivan mentioned in my article’s comments in reference to the Visual Sciences statistics — and, there are tech reasons why traffic sometimes has referrers stripped out — so, the few figures currently published on direct domain traffic are not clearly all attributable to domaining. In addition, a number of the top domaining companies have valid reasons for avoiding the disclosure of all the domain names in their portfolios since that could reveal some competitive intelligence about their strategies and proprietary processes.

Since the domaining industry is a bit defensive/insecure about how the public perceives them, it wasn’t really a surprise that my article immediately attracted some criticism from folks. Frank Schilling, in his Seven Mile blog, posted a write up about my article called “Search Engine Land of Make Believe“. In it, he mentions a number of positive figures about direct navigation traffic, but none of the figures are very reliable.

He mentions one figure about large amounts of Yahoo ad usage coming in from domain name traffic, though it’s apparently from an analyst’s estimate — and Wall Street analysts typically do not understand how internet metrics are computed, and particularly don’t understand the issues involved in lumping all types of apparent direct navigation traffic in together (no independent traffic reporting agency I know of can differentiate between traffic coming in from a user typing in a domain name versus clicking on one of their favorited bookmarks, and we won’t even go into all the situations where referrers are not being passed due to browser/platform security settings).

Another figure he cites is an Efficient Frontier report on direct navigation, though others have already pointed out a number of issues with accepting that report at face value. The graph I see in the report actually shows a pretty sharp degradation in conversions after a few months of the ads running — a normal enough progression since CTRs typically drop off as users become more familiar with the ads, though not not a progression that I would expect if the majority of the traffic on those sites were from new users visiting them for the first time through directly typing in assumed domain names.

Finally, he cites an anecdotal mention from someone within Google quoted on how AdSense was seeing 5 million unique visits per day from the domain distribution channel, but if we assume that number is still accurate today and compare it with comScore’s estimated number of Unique Visitors to Google for May (120,010,000), it’s not all that significant. It’s also not all that interesting if you compare it to the total estimated number of worldwide internet users. So, Frank doesn’t really convince us that things are fantastic with these figures.

So, the reason why I or any other informed analyst or tech reporter would write a cautious and conservative article about domaining is that there’s just a lack of independent validation of the traffic figures, along with something of an inability for collective assessment of true conversion quality. As Danny noted in his comment on Frank’s posting, advertisers can’t opt out of this subset of Google’s nonsearch traffic, and parked domains are not segmented out in reporting stats.

My conservative assessment is merely based on the amount of independent data at-hand, and not on a bias in favor of search traffic. For all of us who lived through all the hype associated with the radical new business models that were promoted during the dot-com boom some years ago, we remember all too well the implosion of the dot-com bubble. The lesson we learned was to focus on the business model and on good statistical evidence from objective sources.

Frank’s hope that:

“Perhaps one day I will visit ‘Search Engine Land’ (the website) and read a less cautious piece that accurately portrays the domain name traffic industry and gives it much deserved respect….”

isn’t all that likely to happen until there is a greater degree of transparency and more solid, independent reporting on traffic and collective conversion rates.

I conclude in the article that there’s significant traffic in domaining networks of sites, and that there’s good money to be made from it for local business companies. But, I suggest caution in monitoring of the quality of that traffic and its associated conversion rate (real conversion rates – not just CTR). I should have also suggested that not all domainer companies should be considered the same — some are great, while others have portfolios full of scummier sites that publicly-traded companies would want to avoid. I actually am impressed with some of the businesses in the industry such as SEDO (who were quite admirable in their recent efforts to police the domaining of names purchased by the unscrupulous in the wake of the Virginia Tech shootings), and REIT impresses me as well. In any case, I didn’t beat anyone up in the article, and I gave the whole thing as fair a shake as I could.

The domainer industry doesn’t need unqualified, glowing reports from analysts and reporters like me — they can continue to laugh all the way to the bank with the money they’re likely generating. But, if they do wish to get more approval and admiration as an industry, there is a way to accomplish it.

Independent traffic reporting companies are currently unable to assess and report on the traffic for the overall industry, or for particular large companies within domaining.

So, Domainers, if you want to prove to everyone the amount of traffic going through your industry, I suggest that you partner with comScores, Nielsen/Netratings, and Hitwise and talk to them about how they might be able to better identify and report upon your traffic. Most likely, this would require that a number of domaining companies would need to team up in order to report all of their domains to those metrics companies after which the traffic reporting firms would be able to group together usage data collectively. Perhaps the domainers could supply their domain portfolio lists to an independent agency, such as the Internet Commerce Association, which could then privately provide the collective list to the reporting agencies.

So, how about it, domainers?  If you want to really make the eyes of CEOs and investors shoot open everywhere, move out of the shadows and get some more independent reporting on what’s going on with all of your many sites!

So, how did this traffic happen?

This isn’t the only time I’ve seen something like this happen. Probably a number of people have had the experience of calling up a partner or colleague to talk about something they see in the Google search results, only to find that the person at the other end of the phone sees a very different thing when they commit the same search in Google. The listing could be shown 9 places down from the top of the page instead of 2 places down, or it isn’t showing up at all for them while it’s showing plain as day for you.

Unfortunately, this is going to become a more and more common experience for webmasters. Google’s diversity of search products and results sets are becoming more and more differentiated for different users, and as this happens, people searching for the very same keyword are going to be seeing completely different search results. Read on for more details.

Web analytics programs commonly report the top search engine keywords that bring users to a site. As the search results for different users become individualized, if you’re a webmaster looking at the analytics metrics you’ll begin noticing that your site may get traffic for search terms when you can’t even see one of your pages listed, or you may not get traffic for somewhat popular terms where you see yourself listed in the very first position in the results page. The lack of clarity in reporting is exacerbated because many of the analytics systems can lump all traffic coming from “something.google.something” or “something.google.com” as being from the Google Search engine. This wildcarding effect might lead you to erroneously assume that all your traffic comes from the US-centric web search results on Google, when in fact you may be receiving traffic from a variety of Google sites in a variety of ways.

Here’s a quick list of some top reasons why your Google search results may appear different from someone else’s:

• Personalization. Google has begun to customize your search results in order to bump up the sorts of stuff that you seem to want so you can find it more easily. This can seem cool in cases where you are clicking on your own company’s pages more frequently, since it may make your CEO think your site is ranking high on his prestige keywords. But, trouble may happen when he is telling his friends about this and they see something different. So, how would you see what most users would be seeing? One way would be to disable the personalization, explained in this helpful thread on the Search Engine Roundtable: Turning Off Google Personalized Results.
  .
• Different Google data centers/servers. Google is big – really big – so it doesn’t run on just one machine – it runs on a great many. Getting all of those machines to be in sync is an ongoing task, and as data trickles across all their machines it results in various servers displaying slightly different search results. So, you may see one set of results in the location where you’re viewing Google, while end users somewhere else may be seeing a different list of results – for the same keyword. There are a few tools out there for viewing the different data centers’ results – here’s one from SEO Chat.
  .
• Google for other countries. Google has individualized search engines for different countries, and keyword searches in them may be filtered down to sites identified with those countries. Ex: Google UK, Google Canada, Google France. Your site might show up higher/lower in some other country for a particular keyword search, and there may be a number of country-centric Googles handling your language. For English language sites, users could be finding you from a number of different Googles: UK, Canada, Australia, New Zealand, South Africa, etc.
  .
• Alternate types of Google searches. While most usage is going through the main Google web search, there are a number of other specialized search engines that web surfers are using: Google Image Search / News Search / Blog Search / Maps / Google Base, etc. Would your site’s content show up in some of these?
  .
• SafeSearch. Google automatically performs a bit of censoring of its search results in an effort to keep objectionable content away from minors and people who wish to avoid it. Google has Moderate filtering engaged by default, but users could be finding/not-finding your content depending on their settings. You can go into your Google Preferences and change the setting to be stricter or disabled. Don’t assume you know for sure if your site or page is flagged as potentially objectionable! For instance, there are perfectly respectable plastic surgery sites which appear/disappear from the SERPs, depending on the user’s SafeSearch setting.
  .
• Language preferences. One can also change the desired language returned in the SERPs in the Google Preferences. Changing this option will cause pages matching your desired tongue to show up much higher in the results order.
  .
• Gadgets / Widgets /Feeds in iGoogle. People who are using Google’s customizable homepage service, AKA “iGoogle“, can add all sorts of “gadgets” or widgets to their homepage, along with content feeds. So, if your site content is accessible through one of the many gadgets out there, or if you have RSS/Atom on your site, users who have subscribed to those services might be coming to you through those avenues.

I’ve probably neglected to add a few other routes which can result in differentiation in how users may come across your pages from Google, and this will continue to change as Google adds and alters their various products and features. But, these are some of the top ways that users may come to you from Google, and these are the primary places to look if you’re trying to figure out why some other people may be seeing different results from what you’re seeing when performing keyword searches, or when you’re trawling through your web analytic keyword referral reports.

Google Analytics is updating their user interface and report presentation, and just as I feared, some of it is a downgrade in usefulness. I’ve been using metrics/analytics packages for ages now. At my old company many years ago, I helped set up and use NetGenesis NetAnalysis product to generate reports from our log files. We later used SurfAid and Coremetrics and Omniture. We also built our own, in-house reporting system to supply stats that we couldn’t get via off-the-shelf packages, and I personally programmed some of those and managed other developers who worked on them as well. So, I’m pretty familiar with analytic reporting systems, and I know what’s possible in designing them. I don’t like some of Google’s changes to their service.

Google Analytics has done what I’ve seen so many other analytics companies do: they’ve dumbed down the reporting presentation capability of their service, apparently gearing it primarily toward less-technical marketers and people who run Google ads on their sites or who advertise through Google. The trend with all these analytics companies seems to be to evolve solely towards generating report charts for marketing departments, focusing mainly on conversion statistics and prettified reports that have extracted the ability to easily see quantitative amounts over timeperiods, lulling the brain with pretty colors and obsessing more over slick Ajax/browser interactions than delivering statistical content in a meaningful way.

In their blog, Google even highlights/links to a ClickZ article which says that the new version was built for Search Marketers.

There are other people who have other interests in the reports than just marketing! These reports are vital for usage analysis and usability research. How about just trying to gage the popularity of a given page? Why is there sometimes such devotion from divorcing people from the raw figures found in the usage tracking logs?!?

One immediate downgrade between the new system and the old is that you can’t view trendlines of Visits simultaneously with PageViews. I want to see the raw numbers of both, in conjunction with each other, not just PageViews per Visit! The new interface lets you only view one or the other, not at the same time.

They’ve messed up the labeling of Referrer metrics, too. In my opinion, it’s misleading to label a report “Referring Sites” and not have it include all. In this report, “Referring Sites” doesn’t include referrals from search engines – you have to go under “All Traffic Sources”. I see merit in having a separate report just for Search Engine Referrals, but not for “Referring Sites” with SEs extracted. Regardless, the label is misleading for anyone with any length of background in internet IT.

The listing of keywords which referred traffic through individual search engines is very cool!

In like vein, why not provide top keyword ranking numbers, similar (and expanded) to what is reported in Google Webmaster Tools?

The worst gaffe of the new UI is in the display of geographic visit data. In the old reports, you could glance at a world map and little dots would show how much traffic came from what areas in the world. Those dots could vary in size to give a quantitative indication of how much traffic came from each area. The new maps dumbed things down so that you can’t easily see how much traffic came from what area in the world without mousing over areas, nor can you tell how it was distributed in each area without zooming down to areas of the map. Compare:

Current Analytics Map UI

New Google Analytics Map UI

Using color scales is really hard on users to quickly scan and understand the geographic charts. Can you easily tell what each country’s shading goes with which color scale shade in the map key?!? No, most users wouldn’t be able to. Mousing over to get the number doesn’t help, either — the point is to be able to easily see amounts across the board and be able to compare them all at once — not one at a time. Can you remember the numbers from Oceana, Africa, Asia, Europe, North America and South America — all at the same time?!? This isn’t a game of Concentration.

Check out this closeup map of the States:

Google Analytics Closeup Map – United States

By viewing this, can you easily view this and tell how many visits came to your site from the various cities throughout the country? No. This is an even worse game of concentration than the worldwide map. Instead of a few continents, try remembering the quantitative amounts for all 50 states at once! Not to mention, can you tell which shade goes with each scale box?

Over time, a user might be able to memorize the scale shades, except — the scale itself is variable, according to the site’s traffic! So, one week the second-to-lightest box MIGHT be representing 1,000 and another week it could represent 10,000. Oh, and it’s not set at even numbers. Oh, and I could just WISH that the scale fell into nice, even numbers! But noooooo! They tell you what the lowest and highest numbers on the scale are, but NOTHING in between! News flash: I’m a pretty clever/capable guy, but I cannot MENTALLY INTERPOLATE each of the numbers in a scale between say, 48 and 1,853! Who can?!? Why should I? The whole point of using charts is to make data readily understandable through visual presentation, so why should I have to work harder to comprehend the data?

I liked the early/current version of Google Analytics, and this new version comes up a bit short for me, particularly with the geographic charts, so I’m going to have to pan it.

I hope they’ll continue to support the legacy interfaces quite a while longer, until they improve some of these issues I’ve noted herein.

It’s free and open source. Download it now for your WordPress blog or site. Enjoy!

Most of the major SEOs and top companies keep such figures as closely-guarded secrets. Even the search engines keep the numbers of searches by various keywords secret, using various techniques to hide actual values.

The much-touted Eye Tracking Study conducted by Enquiro and Did-It show that the first listings on Google SERPs are looked at and clicked upon the most by users. Most pros already concluded this through common sense, but it’s difficult to get actual traffic amounts associated with the rankings of listings on SERPs.

I’m going to change this situation right here, right now, thanks to new data that Google has graciously begun providing to the public, and thanks to a brief reshuffling of rankings on a top keyword for one of the sites that I manage. Read on, and I’ll elaborate.

Recently, Google had a sudden and inexplicable reshuffling of keyword rankings for one of the most hotly-contested keyword terms on the internet, and this seemingly-arbitrary reshuffling affected one of the sites that I manage. Now, due to my work I have to play the boring old “proprietary information” card just like everyone else, so I cannot reveal the site nor the precise keyword that was affected. Suffice it to say, though, that some companies have paid literally millions of dollars in an attempt to gain first ranking for this key term, and suffice it to say that my longterm work has helped to keep our site in the first position for this term since at least the year 2000. (We optimize on millions of keyword combinations, and come up first on a great many of those, so I’m not revealing any proprietary information here.)

Even though I can’t reveal all those “proprietary” specifics, I can give enough specific info to make this very interesting, and I can reveal traffic specifics about some top keywords that have been hitherto Secrets held by various companies.

For a couple of days, our rank on the hotly-contested term suddenly dropped to second slot across most Google data centers. Nothing major was changed on our site/environment nor on the site that popped up to the first slot, and our analysis showed no clear reason for the sudden shift. We thought it could be that Google tweaked some thresholds on the various factors they use in rankings. The change was inexplicable, because it seemed very unlikely that the competitor site could be considered more relevant for the keyword, compared with our site, for a variety of reasons.

After a couple of unhappy days of obsessive checking for SEO mistakes or other environmental issues, the rankings started reverting back to their former order, and it seemed clear that there was some sort of odd and temporary mistake which took some days of processing in Google before the rankings for these sites and this keyword returned to their previous equilibrium.

Now the downside of seemingly-arbitrary reshufflings is that you may lose some traffic. The cool side of it is that it provided very interesting metrics to compare with the historical data.

For the sake of convenience, let’s refer to the keyword in question as Term X.

I can tell you that our site receives approximately 30k of visits on average per day from Google, just from keyword searches for Term X. There’s typically one or two Sponsored Links just above us on the SERP, and a few Sponsored Links on the right side column, too.

When we dropped to second slot on the SERP for Term X, we lost approx 18k of visits per day. So, there it is: the difference between the number one slot and the number two slot for a major keyword term comes to about a 60 % change in visits!

Now, to make this even more interesting, we can now relate this data about a keyword that we do know Google referral traffic for, with some keywords for which we don’t have direct data. This can be done by comparing search traffic for terms using the recently-launched Google Trends interface.

By submitting a few keyword terms to Google trends, separated by commas, Google will return back a graph which shows the amount of searches over time for each of the keywords. If you know how much traffic is produced by one keyword term, and then find some other keyword terms which have histograms which are near to it on Google Trends, you can deduce that their search traffic in Google is very similar, and you can further deduce that their clickthrough values for sites which share the same keyword rank slots on the SERPs will be closely similar to your known value as well.

For example, I found that the keywords “restaurants”, “nyc”, “miami”, “seattle”, and “television” all have histograms that are very close to my Term X during 2006:

Because I know how much traffic I receive from Google on Term X, I now have a very, very good idea of how much referral traffic may be associated with the top slots for Restaurants, NYC, Miami, Seattle, and Television! Occupying the top position for each of those keywords likely would bring a site about 30k in visits!

It’d be nice if Google would actually spell out the units for the vertical dimension on their Trends graphs, to make this utility even more useful. But, we can achieve the next best thing through deduction and interpolation.

Now, how do you estimate the referral traffic for a keyword which has a significantly different vertical value on the Google Trends graphs, from your known keyword value? We can’t immediately assume that the pale gray bands on the vertical dimension are consistent units, since Google likes to obfuscate actual values. (For example, Google doesn’t display actual search values for their Zeitgeist comparison of popular search terms, nor do they show actual numbers of searches in their advertising estimation tools. And, they also don’t tell actual PageRank values for sites.) Just as with the PageRank indicator in the Google Toolbar, it’s possible that their Trends graphs are made on a logarithmic scale. They aim to just show us vague, quantitative values for the volume of searches by keywords.

I don’t know why they insist on obfuscation when it becomes easily possible for us to estimate some true values with the data provided, as I’ve demonstrated. (This reminds me of how the US Government used to block one of the three values provided by the GPS satellites in order to make it harder for terrorists to target missiles. Yet, a few researchers demonstrated that it was no hard work to triangulate the coordinate data from a few different satellites in order to estimate the blocked values with fairly good accuracy.)

All we need to do now is to compare referral data from a few different keywords with significantly different vertical trend lines. That would give us enough data to estimate the vertical units of the graph scales. (I know I’m assuming a lot here — after all, they dynamically change the vertical scales according to the quantity of searches for different queries submitted to Google Trends. And, I’m assuming that referral traffic would be similar for the top keyword position for various different terms. But, once you know the vertical scale for a particular term, you can easily figure out how to resize the scale when comparing with other keywords, and it’s not unreasonable to figure that referral traffic for the top position sites may be fairly consistent, regardless of the keywords, based on the results from the Eye Tracking Study.)

I know a few nomographic tricks from back when I was a scientific illustrator, and compared with some conversion work I did for AASHTO engineering guides used by highway engineers, this should be a walk in a park!

We may still not know the raw number of searches on Google for particular keywords, but we will be able to accurately estimate how much referral traffic can go to a site if they occupy the top one or two positions on the SERP. And that’s what we mainly want to have, isn’t it? A good way of estimating traffic for different keywords and ranking positions, right?

Stay tuned, because I’ll likely have sufficient data to match up a vertical scale for those trend graphs after I have a chance to properly analyze them.

36% of search engine users associate top rankings with brand leadership

In addition, 88% of users will change engines or search terms if they don’t find what they seek within the first three pages of search results, up from 78% in 2002.

Brian