October 2008

22 Oct 2008 11:05 am

more The Google Analytics team announced the release of seven features today. The next stage in the metamorphosis of the popular web analytics tool.

Without a doubt the feature that I am most excited about is Advanced Segmentation. This has been a long time coming (can you sense my pushiness!), and in this post I wanted to share with you all how to use this awesome feature.

Along the way I'll share three different segments that you must have in your web analytics tool. Regardless of why your website exists or what tool you use, Google Analytics or an alternative. I'll close with a approach you can use to get answers to your ad-hoc questions / queries faster, in mere minutes rather than days.

But before we go on here are all the features released today:

1. User Interface refresh.
2. AdSense now integrated into GA.
3. Advanced visualizations: Motion Charts!
4. Custom Reports!
5. Advanced Segmentation!!
6. The Google Analytics API.
7. Automatic importing of AdWords cost data into Urchin.

AdSense and API are in Private Beta (access by invitation). Motion Charts, Custom Reports, Advanced Segmentation are all in Public Beta (being released starting today, gradually to everyone in the next few weeks).


If you want to get really really good at web analytics segmentation I encourage you to read these two posts. . . .

The above posts take a strategic view and get you to think optimally, regardless of the web analytics tool you are using.


Now to evolving from being Reporting Squirrels to being Analysis Ninjas!

Why Segmentation?

Analyzing data in aggregate is a crime.

Bold statement, but the reality is that a "monolith" does not come to your website. Your site does not exist for a singular reason either. The core drivers of traffic are magnificently different for each core group of visitors.

gum drops

So your website's really a mix of Visitor Sources, Visitor Behavior and your Desired Outcomes.

When you look at all that in aggregate you get nothing. You think Average Time on Site means something. No! You think All Visits and Overall Conversion Rate gives you insights. Nyet! You think understanding Keywords without drilling down to each search engine will be awesome. Non!

If you want to find actionable insights you need to segment your web analytics data. You need to separate out the various Sources, Behavior and Outcomes.

Then you'll understand behavior of micro-segments of your website visitors, which in turn will lead you to actionable insights because you are not focusing on a "glob" rather you are focused on a "specific".

Need more convincing? Want a specific case study? See this post: Excellent Analytics Tip#2: Segment Absolutely Everything.

Segmentation in Google Analytics:

Google Analytics always had rich segmentation capabilities. The problem is that you needed to be Robbin Stief to use it. She is super bright (went to Harvard!). She is technically super competent (regex anyone?). She is extremely aggressive! (Of course I say that with love!!)

This release of Advanced Segmentation means that the simplest of folks (Me, hurray!) can now perform sophisticated analysis.

Even my seven year old Damini daughter created two segments in ten minutes. Its that simple!!

Click on the "Advanced Segmentation (Beta)" link and boom!

advanced segmentation in google analytics

What you are seeing are a list of "default" segments. The GA team had a nice brainstorming session about what segments might apply to most people. The ones you see were a result of that, and are pre-created and available in every account.

You could just use the ones above that apply to you. You could also choose one of the above as a starting point, just click on Copy and then further customize them. For example I might click on New Visitors, Copy and then customize it to identify New Visitors from Iceland.

You could also simply click on the link that says "Create new custom segment", if you do you'll be on your way. . . .

advancedsegmentation step one

What you are looking at is Step One in creating any segment. On you left are the Dimensions (customer & campaign attributes) and Metrics (numbers, key performance indicators) you can choose to create custom segments of your own.

Ok with that briefest of tutorials let's dive in and have some fun with this puppy.

[The first example is a lot more detailed, just to show you all the steps. After that we'll do the bare bones.]

Idea #1: Segment out your Brand Search Keywords.

Branding baby!

Most search portfolios (Search Engine Marketing or Search Engine Optimization) have a unhealthy obsessions with our brand terms. Yes it is important to care but for many it takes over all our life and thinking. That is not good.

So the first segment is to segment out your brand terms (this thus far was not that easy in Analytics). We want to get a reality check on how big that segment is, how important it is, and is it worthy of our obsession.

In the search box on top of Dimensions type in Key and Analytics will guess and find the relevant options for you. . . .

segmenting brand keywords

Pick up Keywords and drag it on to the the Dimension or Metric box. Take a breath. :)

To keep things simple I am going to use "avinash" to identify my brand terms. For you it could be "maxim" or "obama" or "quickbooks" or "sexyback44" or "dell" or. . . .

identifying the segment 1

As I type in "av" in the Value box Google Analytics will do a real time look up and tell me my options to aid me in choosing the word I want. Pretty nice. My favorite "tealeaf avinash". :)

Click on "avinash" and I am done!

Of course I could also add another statement and include "avanish" or maybe even "occam's razor" (which I think of as my "brand" term). But for now I want to keep this simple.

One sweet feature that differentiates segment creation in Google Analytics from that of other vendors is the ability to do real time QA (quality control) and test the segment you have created.

In most web analytics tools (NedStat is a notable exception) you create the segment, process the data, take a nap, wake up and realize you made a boo boo. Then you repeat the process. With Analytics (and NedStat) you don't have to do that.

Its like making sure that you go out on a first date before you propose marriage. Always prudent.

When I click on Test Segment GA will run a life query for the time period I had chosen and bring me the actual data for my segment.

brand keyword segment testing

So of the 44,528 Visits in these weeks I only good 576 Visits from my brand terms?

Heart attack!


Ego bruised!!

Breath. Breath. Look. Ahh… I made a mistake.

Turns out that in the Condition I choose "Matches Exactly" (look in the middle above).

So the segment I have created is how many Visits from people typing in the exact term "avinash", what I wanted was Visits generated by every variation of "avinash". avinash kaushik. avinash blog. analytics blog avinash. amazing avinash awesome author. Ok that last one's made up.

I'll make one small change to my segment:

segmentation conditional expression

I click on the Condition and change from Matches Exactly to Contains.

Notice above that without knowing too much about regex and what not you can easily choose a number of conditions you can apply to your segment.

At the bottom I type in Brand Keywords and click Create Segment (after testing it one time!) and I am done.

Next step, trying to be a Analysis Ninja.

You now have tha ability to apply this segment you have created on pretty much every single report that you have in Analytics.

For example you could go in and see what content (pages) do people who come on Brand Keywords consume. What products do they buy? If I were GM then which product brochures do the download? On a tech support site did they find the right tech support answer? For a social networking site, do people under brand terms show a higher Visitor Loyalty or lower? And on and on.

For me I wanted to start with understanding the size. Recall I have a fragile heart.

So I go into the Keywords report and look at the Visits graph. . . .

comparing brand visit trends

I have to admit it did make a cry just a bit.

That orange line is the number of Visits from my Brand Keywords and the blue line is All Visits. So every variation of my brand makes up such a pathetic number of my overall Visits.

Ok let's put the kidding aside. This proves a very important point I make in my presentations. Search has a very very long tail.

As is clear above I don't got a few visits from my brand terms, even though they dominate the top ten keywords report I look at every day in Google Analytics (or Omniture or WebTrends or CoreMetrics or IndexTools or ClickTracks or whatever else).

I have hundreds and hundreds of keywords that individually drive five or ten visits each.

If I were running a business here then I would use the above simple graph to get my Management team to start paying attention, and creating a solid strategy, to the long tail of search .

Remember your brand terms are important, these are people who know you. But it is equally important that you have a solid long tail strategy because that's where you'll find people new to your franchise.

Next I care about money. Is this segment of Visitors converting (meeting goal targets)?

comparing brand conversion trends


While they form a much smaller % of my overall Visits, Visitors on my brand terms form a nice % of my overall conversions. 18% of the all important Goal 3 (which has a very high $$$ value!).

Looking at the percentage view confirms the obvious. . . .

comparing brand individual goals


When was the last time that you understood this much about a valuable part of your traffic, all in less than 10 mins? This stuff is not that difficult, and now its even more straightforward.

Next I would look at landing pages for my brand terms and see if I could tighten them up.

I would certainly figure how to create a solid search long tail strategy.

And I'll do one more thing to help with that. Create a new custom segment for my non-branded keywords.

Here is one way of doing that. . . .

segmenting non brand keywords

Notice all I did is:

~ Switch the Condition from "Contains" to "Does Not Contain".

~ Added Source Contains the search engines (Yahoo, Google, Live – those are 99% of my traffic).

I save it and I am done. 99% accurate but good enough. I can get the last 1% accuracy if I can afford to spend a couple hours. In my case it is simply not worth the investment of time. You'll make your own choice, just make sure you are balancing time invested with the reward you get.

I recommend looking at your key Visits, Traffic Sources, Content reports to identify how to start monetizing your long tail.

Trust me it is not that much work, and it is just so much sexy fun.

Idea #2: % Visits with a high "Degree of Engagement".

Yes I did use the word engagement! But atleast I am following the simplicity and guidance from Theo Papadakis.

From the dimension selector choose Page Depth. . . .

website engagement

What I am looking for is the number of visits where more than a certain number of pages were seen. To convince you of my greatness you have to read atleast three posts I have written. Yes it is not instant impression! :)

In your case it could be that you have a shopping site and the cart, checkout it 4 pages, then to get to the cart it takes another three pages of product pages and cute babies on pages. How many people on your site are going through the torture?

Or how many people are you impressing enough to stick?

Pick Page Depth. Then pick Greater Than. Then type in the number (be honest). I choose three. Here are my results (remember to click on the super sweet Test button). . . .

visitor engagement

2,916 visits out of 44,519!

Honestly that surprised me. This is after all a blog and most people come to read the latest post and nothing else (which is fine). But almost three thousand read three pages. Yea!

What's your website's number?

The actions I would take would be to understand what websites / affiliates / keywords / campaigns are sending me Visits that have a positive "degree of engagement".

I would also investigate what pages these people land at, what content they consume (compared to my other segment, the terrible people who bounce! just kidding!).

Learn from these Visitors and then apply them to other Visitors / Pages / Campaigns that have a poor "degree of engagement".

Excited? See I am.

Idea #3: Outcomes! Segment Precious Customers.

So you sell on your site. Good for everyone!

One good idea for you to maximize revenue (and admit it everyone's scaring you about how bad the economy is and what not). So your HiPPO's are pressuring you to make more money.

What you want to do is identify Valuable Customers (after they buy from you they move from Visitors to Customers, yea!), and learn something from them.

One easy way to do that is to find the "whales". :)

I am going to use the number of items that Customers purchased on my website as a proxy for "whales". . . .

high value customer segments

I want everyone who bought more than three things.

The actions I could take it to again try to understand a lot more about the Visitor persona that ends up converting to a "whale" customer.

Quantity is just one of the ecommerce segmentation possibilities in Analytics. There are, if I am not mistaken, another 15 to 20 attributes about your ecommerce experience that you can track, and hence segment by.

It has never been easier to understand what makes people purchase, what items go together, what calls to action / content works etc etc.

Idea #4: A Surprise: Use Segmentation to do Ad-Hoc Analysis.

I have a lot of ad-hoc questions I want answered. I can do this using reports in the web analytics tools. But it takes time. And usually is a pain.

The addition of the Test feature in Google Analytics Advanced Segmentation allows me to answer those questions much much more easily.

I'll give you a example.

I was doing the page depth segmentation (Item 2 above). I got distracted and noticed that another post from my blog got "stumbled upon". This always warms my heart because it brings lots of people to the blog.

But the question I had was: Do these people from Stumble Upon have a high "degree of engagement"?

So start with page depth. . . .

degree of engagement

and then add my source of traffic, Stumble Upon. . . .

stumble upon segment

and bada bing bada boom (!). . . .

degree of engagement stumble upon

653 people came from Stumble Upon in these couple weeks. Not bad at all.

But sadly only 17 of those met the "degree of engagement" test. There were a bunch more if I reduced degree to 2. But still not as many.

It reiterated to me that it is nice to get stumbled upon, but you build audience one person at a time.

In 15 seconds I answered a question. Prior to now it would have been really hard to do it that fast.

Now I can dump this and move on to the next thing my mind flies to.

Soapbox: :)

In Web Analytics there is too much report creation and too much trying to do things permanently and forever. Yet most interesting questions come ad-hoc and answers that require you to submit a ticket or justify the creation of a new report are answers that come late and are useless.

Of you use Google Analytics use the Test feature to get your ad-hoc questions answered in seconds, not days. If you are using Omniture or WebTrends or other web analytics tool, then use this feature in those tools.

Don't settle. Demand answers fast.


Regardless of the web analytics tool you use it is extremely important to use the power of segmentation to understand Visitor Sources, Visitor Behavior and Outcomes.

There is no other alternative to actionability.

One last, perhaps even more exciting, thought. If you combine the power of new GA Custom Reports feature with Customer Segmentation then you'll never have to look at a standard generic data again. You can get precisely the data you want for your business. Check it out.

Want access now?

If you want access to the public beta features (they are slowly being rolled out as I write this) then you can send me a email with two items. 1) Your Google Analytics account id (UA-something) and 2) Your email address that has access to that account.

On your behalf I will submit it to the GA team and request them to activate the accounts, for the first xx people who email me. Important: Please do not submit this via comments below. Please only send a email (blog at kaushik dot net).

[Update : Several hundred (!!) of you wrote in for early access. Thanks to the team at Google, by end of day today, 23rd, you will all have access to the new public features. Now off to reply to all your emails individually!]

Ok your turn.

Have you tried segmentation in Google Analytics? Yes it is in beta and there is scope for improvement but what do you think of it? Have your absolute favorite segment you use all the time?

Your ideas, suggestions, critique are welcome.


13 Oct 2008 12:57 am

togetherI don't know of many other words quite as abused as innovation. Maybe the word strategy, but that's about it.

It seems to mean everything and anything. [Reminds me of "engagement"! :]

I have personally struggled with internalizing what innovation truly means. I mean really truly. Until recently.

In preparing for my talk (Self Sustaining Innovation: Ideas for Creating A Data Driven Organization) at the JMP Summit, I found the perfect way of thinking about innovation. It came from a old Google talk given by Douglas Merrill (ex-CTO of Google). I loved the framework so much that I adapted it for my JMP presentation, and wanted to share it here with you all as well.

In his talk Douglas defined three types of innovation. I think it provides a great deal of clarity beyond the standard definition of innovation as "improving something".

#1: Incremental Innovation.

A great example of incremental innovation is. . . wait for it. . . yes you in Kansas. . . . Evolution.

incremental innovation

Life rolls forward and over time there are small improvements that add up to something significant when looked at over time (and of course usually in hindsight).

A more every day example is Kaizen popularized by the Japanese car manufactures, for example arranging parts so that they are closer to each employee improves the process ever so little bit. Enough of these improvements add up.

For me personally Microsoft's "ribbon" menu in the Office 2007 were a nice example of incremental innovation. It makes a complex product much easier to use, in turn causing a bump up in revenues for Microsoft.

Another example of a incremental innovation is "related searches" you see on the Google search results page. Or more recently the suggestions that come up when you search on google.com.

googlecom avinash kaushik

Or the "onebox" in Google's Chrome. Each small by itself, but keeps moving the ball forward.

#2: Incremental Innovation with a Side Effect.

Opposable thumbs. That's a great example of incremental innovation.

incremental innovation with a side effect 1

At some point during our evolution we (and not everyone else competing with us – woo hoo!) got opposable thumbs. An otherwise mundane development that allowed us to hold things better. This lead to tools. That lead to other delights that you now see around you.

Life rolls forward and along the way a side effect that causes significantly higher improvements.

A non-mundane example of this is Apple's creation of the iPod. Its normal Mac innovation machine produced something that redefined who Apple is as a company today.

Another example is Google's AdSense. Prior to AdSense there were other ways to monetize content on the web. But AdSense was a great side effect because it allowed any blogger, student, housewife, or big site to add text based ads to their site with greater ease and monetize content. While certainly not perfect it remains one of the most widely used monetization systems on the web. Good for Google and good for its customers (big or small).

#3: Transformational Innovation.

You'll understand why there aren't too many examples of transformational innovation in nature. That's not how, fortunately, nature works.

transformational innovation

Over the course of human history there are lots of examples of transformational innovations. The invention of the wheel. The printing press. The interweb. Lots of examples like that. Dramatically changed life for all of us. One thing, huge impact, new curve.

A much smaller business example from Google is AdWords.

Advertisements on search results pages existed before Google. Not just banners but text based results existed as well (in various forms). Google's transformational innovation was creating a auction based model for ads where the price paid for the click on the ad was directly tied to the economic value created for the advertiser.

There are other examples all around you of companies, people, organizations that have, by leveraging ideas big and small, created transformational innovation. Please share your examples via comments.

Net net. . . .

Next time someone sprout's innovation on you then my hope is that you'll think of these three types of innovation and promptly ask them what kind of innovation are they hoping to accomplish.

They'll be confused.

Draw the graphs for them.

Then ask them again which type.

And don't forget to ask them if the investment in "innovation" will be worth the ROI.

A company we all love and know spent $6.2 billion in R&D in fiscal year 2005 and $7.8 billion the year before. Yet for all that money you would be hard pressed to find even one example of Transformational Innovation from them between 2004 and today (they spent north of $5 billion each year since 2005). The best I can think of is minor incremental innovation.

For the CEO of that company I offer, with the deepest humility, this framework so that he can make this exhortation when he signs next year's R&D chq: "… and dudes when I say innovation, I mean transformational! Here's a graph from a nice blogger that shows what I expect. Note the red line!!" : )

What kind of innovation can Web Analytics empower?

web analytics tool This is a blog about web analytics, so what about us? What can we deliver?

While we all hope to achieve transformational innovation in our web based efforts using web analytics data, I have come to the realization that for the most part that is not possible.

Transformational innovation is possible on the web, and it is all around us. I am unsure that it is because of the use of web analytics data.

I feel sad saying that but unfortunately we have to be truthful and acknowledge reality.

A big reason for this (we sadly not powering transformational innovation) is because for 99.9% of the people who have access to data out there web analytics still means slapping a javascript tag or processing a log file.


That simply does not have enough juice.

A far bigger reason is that there aren't enough smart Analysts around. Companies are complex bureaucracy. Web Analytics, in spite of best intentions, is still an afterthought. In a silo. Online Marketing is largely still a faith based initiative.

It is hard to admit it but it is true.

Ok if not transformational then what?

All that is not to say that web analytics can't power some forms of innovation.

If you only going to rely on your clickstream tool then really it would be hard to even to incremental innovation. If that's all you are using to make decisions, then. . . . well. . . . not much is going to happen.

To get in the innovation game on the web using data you need to get inspired by the Web Analytics 2.0 mindset and approach. . . . .

web analytics 20 2

You can improve the chances of causing repeatable incremental innovation in your company if you use a couple of the components above.

Say using Google Analytics, Omniture, CoreMetrics, NedStat etc AND using the Website Optimizer. You analyze the data and do A/B testing (or even Multivariate) and you increase the chances that you'll keep your company moving forward.

Or use the clickstream tools and also 4Q. Those two together will help you understand both the What and part of the Why. Again customer ideas flow in, you execute and bam (!) incremental innovation.

Now here's the great news.

I think that for each of our organizations (big or small) we, the Awesome Analytics Actuaries (A3 baby!), can actually empower incremental innovation with a side effect.

For that we need to kick it up a notch and truly execute against the Web Analytics 2.0 mindset. Fire all the cylinders!

innovation with a side effectThat means using analytics tools but also using Surveys (to get ideas from our customers), using Competitive Intelligence tools (to learn from the ecosystem), using Experimentation and Testing (to prove the HiPPO's wrong, and try new ideas), using robust Outcomes analysis (to know why what works). All that massively improves your chances that you'll find "side effects".

But it's not just tools.

It means following the 10/90 rule, it means setting up your people to be Analysis Ninjas and not Reporting Squirrels and other such cultural changes.

But it can be done. We have the power to bring about incremental innovation with side effects.

Yes we can! (Now where have I heard that before?)

Ok now its your turn.

What do you think when you hear the word innovation? Does this framework help? Would you propose an alternative? Care to share your favorite innovation? Web or otherwise.

Please share your feedback, ideas, critique. Thanks.

I had the privilege of leading the WAA Standards committee last year. Its hard working pretty Members have produced an awesome metrics definitions document covering 29 new terms.

The committee is asking for your brilliant feedback. Please share it with them on the WAA Site.

If you prefer email, you can write to me, I'll be happy to forward. If you prefer snail mail, let me know. If you prefer morse code, we can accommodate that as well. Please share your feedback any way convenient.