Lakhasly

Data Rich, Information Poor
the challenge with web analytics is the very first time somebody plugged in a web server into a wire, it turns out that it spews out data. It's a magical thing that happened You plug in the ethernet cable and lots of data comes out And the approach for the last 10 odd years with web analytics has been that we have all this data, why don't we add, subtract, and multiply it, and puke it over the fence and the people will figure out what to do with it And that model worked for a few years But it does not work anymore Web analytics is incredibly complex, and simply puking data out is not a good way to solve the problem Because what happens is that if you sit in the eyes of the customers and I've spent sort of ten odd years sitting on the practitioner side of things all of these data just simply provides questions and numbers It does not provide you with answers And it rarely provides you with any insights that you can take action on And so the model that existed thus far with web analytics is broken, because data puking b Analytics Challenge> is no longer an option And so, I think that we're Not-So-New Paradigm for Web Analytics living in this nat-so-new

paradigm that I definitely want all of you to take away from today, is to live in a world where data is at the service of driving action

And in the book, I have this three layers of so-what test And essentially, it is that if you look at a metric, anywhere, on the dashboard or in the report, and you ask it so-what three times, and at the end of the third time, it doesn't give you an action, you're wasting your time, no matter what guru or thought leader pontificated about the use of the metric Data should drive action If it doesn't drive action, you're wasting your time And the interesting thing with web analytics, unlike other things in the world, is there's a lot of data to waste time on. So what I want to do today is talk, there are two sort of parts to my presentation
The first thing I want to talk about are The six odd principles about how I think Google Analytics helps you understand your data much more efficiently and drive action And the second part of my presentation I want to talk about this thing I call rules for revolutionaries. If you want to go back and revolutionize your approach, your implementation of data, I have sort of five, six principles that I have learned through my painful experience that I want to share with you today. So the first thing is. Google Analytics does a fantastic job. It does a fantastic job of creating a data democracy And unlike our adventure in Iraq, this one actually works. And it's a good thing.

So notice that it's, I walk into lots of different companies, I do consulting for some companies and I walk in and they'll say, here is our dashboard and it has 62 tabs This is not a dashboard, it's a frickin' report. With Google Analytics, it's actually pretty easy for you to go and create your dashboards It's nice It's pretty You can look at your data You can look at your performance And the nice thing about it, I think as Brett touched upon, is that you can schedule and email this, which is wonderful. You can actually go in and say, you can send a data wake up call. You can say, hey, every Monday you're going to get your data You know, at certain times you're going to get your data. But a lot of people don't realize is you can escape from the 62 tab world because you don't have to accept what is by default available for you in Google Analytics You can go in and say, you know, I have somebody who is purely responsible for doing all merchandising on my website. And it's very important for her to know all of the data around visitors, okay? We can send her this specific report. Or you have somebody who is responsible for all your affiliate marketing or a particular website or your relationship somewhere Just send Laura exactly what she needs Not 62 tabs The thing that needs, she needs to do her job It's easy for you to do that in Google Analytics And a lot of people don't realize that it's that easy. You can sell reports for people. If Andy's responsible for one particular page, just send him that. Why do you want to do death by data?

And so there are lots of different options and lots of great ways in which you can look at data in Google Analytics, but also you can share it with people who need to take action on it If somebody's responsible for the checkout process on your website. Somebody's responsible, in this case, news or excavation. Just send them this little snippet of data. Every time that they need something. You can schedule it. You can email it And they can get on with their life without ever having to come and bother you. A lot of organizations have relied on very few people to actually look at the data, analyze it, find insights, and then take action. And it's really difficult for you to sit in a 10,000 people organization. Or even a 500 people organization and have that model scale. It does not scale Because the people who are running the business have the tribal knowledge that can drive action. But they don't have access to data. In more than one way, Google Analytics makes it easier for you to get data in the right hands of the people at the right time. And they don't even have to log into the tool. The other thing that I think the current version of Google Analytics does a magnificent job at is data discovery Because the model thus far has been that we'll give you lots of tables, we'll give you graphs, and they'll show up at your desk. Or actually, it's kind of funny I was at a Fortune 20 company last month. And there, the person in the company who makes decisions about the website has their secretary print the reports and put it on his desk every week, which is, I cannot believe it. But the interesting thing is. It's really hard for you to make decisions based on a printout or a spreadsheet that's
sitting on a desk. Because every time you look at a piece of data, you want to ask it questions. And you want it easy to get the answers back really quickly. So what GA does is you can say, oh, here is nice, OK.
Page views are dipping over there. At least I'm getting all the context to my data, not just one. Now I know exactly how everything else is going. But I wonder where all these people are coming from. So half a million people came to my website.

Where are they all coming from? So it's very easy. You click on a button. Boom. There's your visualization. It's really easy for you to understand. Then you can say, oh, I wonder what's happening in Africa. Maybe I want to run AdWords campaigns in Africa Now you can do that. Here's another great one. You want to make the most optimal content for your websites. So boom. There's your report. This is exactly what's going on when it comes to content But notice, it's really easy for you to have all of these other options. I'm going to show you a couple of options on the right that help you figure out, what should I do next? What is it that is interesting to me? So it's like, oh, all these people came to my website I wonder, oh. We do have good errors. I actually got this the other day. It's very cute. It's nice. Please try again. Thank you for your patience. It was very cute Let's get back to the story. But the nice thing, I can say, oh, I wonder where do people come from to a website?
One of the things that a lot of people don't realize, I really love this one, is a lot of people obsess and have their egos tied in homepages of the websites.

And the thing that people don't realize is that you actually don't have control anymore about what the homepage of your website is. 80% of the people start surfing at a search engine An astonishing number, by the way. Scared me. But 80% of the people start surfing at the internet. So Yahoo, MSN, and Google decide what to do. They decide what the homepage of your website is, not you Based on what people are typing into a search engine, the search engine will do its best to figure out what's the right page on your website to show them. And that is the homepage of your website. So as I browse around and look at content reports, I'm on the root directory.
I wonder how many people enter my web, any given page. And that's the first page they see. It's a great number to know. And it's very easy. You go in. You say, you know, I want to know all of the entrance sources.

I want entrance paths. Entrance keywords. I get all this data with an easy reach that helps me understand where are people coming

from, what is driving them to my website, and do I have relevant content on that page for these people. Don't let your egos be tied up to the homepage.

No one cares. So you can see, okay, where are these people coming from? Oh, 80%, 87, 83% actually entered the site on this page. It's very important.

I wonder what keywords are driving them to my page, or actually what sources they're coming from. Sure enough, Google, Indeed, Yahoo, right? These are all the places they're coming from. Oh, nice.

I wonder what keywords are driving them so I can make sure that this is the right page for them to be landing at. Oh, all of these pages. And look, there's a wide diversity at it. I can go back and say, whoa, should this be the case? Should I have a different strategy?

But to do all of this, this, this, I've dived really deep into the data. And I've only pressed three buttons. And that's the interesting thing.

It's much easier for you to just start at any level you want of your data and simply explore the data and find insights.

This is really hard for you to do with Excel spreadsheets that might come from any tool. Doesn't matter what tool you're using.

The other one that I think is really interesting is a lot of people cannot make decisions based on their data. Because they don't have context.

One of the most wonderful things about the internet is that if you're half-competent at your business, all your web analytics numbers go up and to the right over time.

Even server errors go up and to the right over time. If that's the reality of the world, how do you make decisions?

I think people can't make good decisions on their data. Because they don't have context. And context is extremely important.

I think one of the things that, the team that builds GA is sitting at the back over there. So I am simply showcasing their work. So big round of applause for the team over there.

So one of the things that I think they've done a really great job of is provide you context as you look at your data. So for example, rather than looking at one solitary number, you can actually get a nice performance overview.

So you can see how many pages are there, how long are people spending, what's the exit rate, get a dollar index, sweet.

But what is actually better is no matter where you go in Google Analytics, by default the team is trying to give you context from your data.

In this case, if you look at this trip, it's actually telling you, it's comparing the numbers to site averages and actually giving you a hint of where you should be focusing on. And that's very nice.

You get this by default in most reports in Google Analytics, you get context. So you no longer have to worry about is that good or bad. It's giving you an initial indication that you're doing kind of okay over there.

But the nice thing is it's extremely easy for you to go in and say, I want to compare two different time periods. I want to compare this June in this year to June in last year. I want to compare this week to last week, last three months to the three months before.

And that gives you even more specific context for yourself. In this case, I'm comparing May and June. Simple. This is the dumbest thing to do. Compare two different months. Am I doing better or worse?
In this particular case, whoa, that's kind of nice. I got 300% more page views on the website. And look, I kind of did not suck.

By the way, this kind of growth is really hard to manage. You get so many more new people coming and consuming content. It's really hard for it not to actually impact your other numbers adversely. You'll see this more later in the day. So this website, spectacular success.

Context is the thing. Very quickly, you open a report, you look at it, you have an idea if you're doing well or not. But there is also one other subtle difference between these three things.

When you show most senior executives the first strip of data, just this data with no context, the question that they ask you is, what is the definition of time on page?

Because they have no idea if the number is good or bad. Say, tell me how you captured the data. A useless conversation, by the way.

If you give them these two pieces of data, nobody cares about what the definition is. You know where you're good and where you're not good. And it moves the conversation along farther quickly.

And that's what you want to do. You don't want to argue about things that are not important. You can do this in your tools. You can do this in Excel. You can do this in Excel if you want.

Very simple strategies in which you can give context to your data so you can actually make decisions from it, rather than arguing about what the definition is. Here's another way. This is nice.

Everybody wants to make money. You can set up your goals. If you don't have goals in Google Analytics, that's very suboptimal. I'm a huge fan. There's a whole chapter in the book that talks about how outcomes are very important.

If you don't know what the outcomes of your website are, no amount of analysis will tell you what you should be doing. So goals are good. Sometimes you mess up and you create a goal that you don't want to track anymore, then it's fine. But create goals.

And the nice thing is you can say, oh, I see conversions. It kind of sort of looks like it's going up, but I do the same exact thing, compare two different time periods, I get a much better read on performance.

Not doing so well over here, doing a little bit better over here. Right? Very, very simple trick, but it gives you context to your data. Exact same thing here. Isn't it more fun to know?

To know that you're up 7% rather than knowing you've got a million visits on your website? Like who cares about the absolute number? What's more important is to move the conversation along. Same thing here. Search keywords.

Oh, sorry. Where are the people coming from? What traffic sources are working better for me? Very, very easy for you to understand.

And the next thing I wanted to share is data in service of action, just to sort of reiterate the theme of what I've been talking about so far. And it's nice.

You go to your content, this exact same screenshot I had a few slides ago. It's nice. I know overall story, nice page view trend, I have my strip, I have my context, and now

I'm ready to move beyond the summary level data to know at the nitty-gritty level what's going on. And when you go to that level, it's kind of overwhelming to say, this is cute and, dare

I say, sexy, but it's a lot of data. It's a lot of data. It's nice. Page views look good. But where should I look? How should I think?

In GA, you press one button, that button up there, and it actually allows you to compare two different metrics quickly.

So not only do I know what pages are more popular on my website, but I can actually compute how well are those pages performing in terms of sucking people into my website

and getting them to spend more time so I can show them my banner ads. Quickly. Notice some really wonderful things here happening compared to over there.

One of my new principles, PETA, People for Ethical Treatment of Animals, that's a great organization.

And so I came up with a new acronym, and it's called PALM, P-A-L-M, and it stands for People Against Lonely Metrics.

And the reason I say that is because lots of people look at one number. By itself, we all want a husband, and a wife, and a girlfriend, and a boyfriend. Maybe all of them at the same time.

By the way, that's this screenshot. But if you want to make effective decisions with the data, always find the best friend for whatever metric you're trying to track.

If page views is what you're trying to drive at, find the best possible metric that would give you context with the data that allows you to take action. In this case. For me personally, I want to know what pages are important.

Actually, this is much more important. So I know what stories are more important to people, what people react to, so I can write more. If I did not look at this metric, the boyfriend metric,

the girlfriend metric might be telling me something wrong. So for your metrics, as you analyze the data, never ever on any dashboard or report look at one metric by itself.

It will not give you the kind of insight you need to take action that will impact your bottom line. And the nice thing is there are a couple different ways to do it. Is this good or bad? I can make some judgments.

But if you want to short circuit the time it takes to make the decision, click one more button and you get this. Now this is a wonderful view. Exactly the same data, exact same metrics, but now it's

comparing performance of every page in Timeite site to average amount of time people spend on your website. And this is very cool. Because now I don't even have to think.

I can say, oh, cool, cool, cool, sock, sock, sock. Very, very easy. You can do this anywhere. I'm not getting impressive with animations, right? But you can do this anywhere you want.

It's very easy for you to find a best friend for your metric that allows you to define performance. And then it's very easy for you to identify where you need to be paying attention.

All in two clicks. Exactly the same thing, right? Everybody here loves search. So you can look at all your search data for goals, outcomes, exactly the same thing.

Which search engine is more important? In this case, if you want drive conversion, actually Yahoo seems to be more important than Google. Although Google brings a stunning more amount of traffic. So you can balance your strategy. It gives you insight.

Exact same view. But now, comparing to site average, drives action. The wonderful thing about the difference between this graph and this is the same one that I had mentioned before.

In this case, people try to understand data. And some analysts amongst you want to try and spend time understanding data. This is what you send to your senior management.
They should not spend time understanding data. They should be making decisions. And it's easy, if you look at this view, for it to be accretive to driving decisions quickly.

So one of the more popular posts I had written on my blog recently was that on non-e-commerce. I've become a big fan of not making money.

And I'm sure there are many of you here today who might not be actually have e-commerce websites. And the nice thing is that there is a really wonderful

area in Google Analytics that you can use to track effectiveness

I contribute $20,000 of that, and he contributes $100 billion a year. And that's the important thing to remember. Averages often hide the truth.

So for example, rather than measuring the number of visitors and the unique visitors that come to a content site, in this case, you would assume that approximately each person comes twice.

Half a million, a million, that's kind of what it looks like. But these are all real numbers, by the way, for a site, the reality actually looks like this.

The averages hide the truth effectively. When you do distribution, you get a much better understanding of what is actually going on in the website.

So notice, most people actually just come once to this hardcore content site that's trying to spam people with banner ads. God bless them.

The distribution paints a radically different story. You have a big head, then it shrinks really quickly, and look what happens here. Whoa.

There are these loyal, crazy people like me who are actually coming to this site 100, 200 times. If you actually understood that you have this swath of

traffic, about 30-odd percent of the traffic right here, that is spending between $14,000 to $200,000. You have to understand that this is a very, very they're coming to your website, wouldn't you design the site differently? Wouldn't you react to them differently?

Wouldn't your content strategy be different? This is a very, very powerful way of looking at data by simply using distribution rather than using an average. So this is a great metric.

If you are a content site, you want to know if people are loyal to your website. Right? At the same time.

slightly better in July, got 63% over there. Then it's recency. Again, purely for non-e-commerce websites, it's important to know if that's your business.

Do people come again and again? Or when was the last time I actually saw Jim? It gives me an indicator of success. If I have good content about providing value to you, you will probably come more frequently.

On this non-e-commerce website, 66% of the traffic was brand new. That's, by the way, a stunning number for them to realize. Because they had designed all of their content navigation

structure based on the fact that you knew them really, really well. They had no pages defining the value that their site was bringing to them. And yet there was this massive amount of traffic that was brand new to the franchise.

So then you can go and figure out, hey, what's going on with recency? So this goes back, literally, this website has had Google Analytics for more than a year.

So you will go back, take the people who you have seen in this time period that you're interested in, and will go back all the way in history and say, hey, when was the last time you saw this person? Very powerful report.

If you're doing a content site, you probably want people to be over there, not over here. Another average one is people use time on site.

11 minutes over there, and look at the distribution. So you have lots of people who stay a very little amount of time, and then you have a chunk of people here who stay a really, really long amount of time.

This is very effective in you making decisions and measuring success of your website. Same thing with depth of visit. How much content do people consume?

The wonderful thing about these distributions is that it truly helps you understand the value of a non-e-commerce website from your web analytics data. Are people loyal to you? Do they come back again?

How long are they spending, and how much content they consume? And if you're cnn.com or you're sap.com, you're any website that does not do e-commerce, if you want to

measure success of your website, it's a powerful way for you to get started. The next thing you would probably do is you'll say, OK, let me segment this data. How do the numbers look from Google? How do the numbers look for my marketing campaigns?

How do the numbers look for my key more important keywords? You will segment this data, and you'll begin to understand a lot more effectively what's actually going on the website and what action you should be taking.

And my recommendation for non-e-commerce sites always is throw a survey up. Throw a simple little survey, three or four questions, and ask people what they thought of the content.

This piece of data still does not, it's kind of, sort of, the value of content and other fuzzy metrics. So if you want to do fuzzy metrics, throw up a quick survey, and I can say, hey, as a result of my experience, is my

brand getting better? Is the likelihood that I'm driving people to offline sales better? Or nine months from you, are you going to buy a big caterpillar system from me? Because you're not ready to buy it now.

So in combining a quick little survey methodology with what you saw in GA with these metrics, you can truly begin to understand if your content website, if your non-e-commerce website is performing as well as it should.

And stay away from averages. At least in this particular case. So I am a huge fan of the metric bounce rate.

As I'm putting this deck together, I did not want to recommend a metric. Because I fundamentally believe that there's no such thing as, oh, here are God's KPIs, and if you measure them, you're fine.

That's like 10-year-old methodology. KPI, KPI, KPI. Every business is different. So my mindset is, here's the data, here's an effective way for you to navigate the data. You figure out what's best for you.

Rather than me pontificating on it. This is the one exception I will make. Only exception today is a metric called bounce rate. I love it. I call it the sexiest metric ever.

And the reason I like bounce rate is because it is so powerful and so, I call it brilliantly dumb. Because it's really easy to understand what a bounce rate is.

There's no need for definition. It, bounce rate, defines this experience for your customer. I came, I puked, I left. Literally, that's the definition.
The other thing is you all spend tons of money on AdWords and direct marketing and affiliate marketing and all of those things.

And what this metric can help you understand is where are you acquiring crappy traffic from very quickly. So it's easy for you to go in and say, oh, what's the bounce rate on my site?

70% in this case. So a lot of people wonder why the conversion rate is 1%. Actually, for most of you, if you do e-commerce, I bet you your conversion rate is 2%.

That's the average conversion rate in the United States of America, whether you're selling elephants or iPods. 2%. If you're really good at direct marketing, your conversion is probably close to 30%, 35%.

2%. This is why your conversion rate is 2%. Because most traffic comes and leaves instantly. And yet, when I do seminars or go talk to companies, I'm

stunned at how few people actually know what the bounce rate of their website is. So I encourage you to do this. Now, it's possible that your bounce rate is more like this, in which case it needs a love sign.

This is a great bounce rate. On average, you'll find the bounce rate for your site will fall between 40% to 60%. It should probably be closer to this number. It's a great bounce rate.

You cannot convince all the traffic on your website to stay. You can't. Everybody has ADD. You don't have what they want, they leave instantly. So you can't win them all, right? But that's a great bounce rate.

It's a fabulous bounce rate. I was stunned. Actually, that's low. The next thing you want to do is you say, ah, I have all these sites. They're sending me traffic. And I think if you look at the graph, you'll notice it actually is trending up and to the right.

Slowly, but it is trending up. This is good to know. This is nice. But then you can say, where is all my traffic coming from? OK, sweet. Loveindeed.com and mail.google. Working great.

I wonder what the bounce rate performance is. And they all are sending quickly unqualified traffic. So it could be I'm running the wrong campaign. Maybe this is not a great place to go.

Well, bounce rate is a great qualifying metric. It's a great metric that helps you ask the right questions.

Exact same thing here for search. My keywords, what's my bounce rate? I can actually look at my landing pages and my bounce rate. I love this one, by the way.

Definitely look at the top 20, 30 entry pages to your website, where people are. Entering and look at the bounce rate for those pages.

Because remember, your home page is not the home page of your website. This is where people are entering. Are these pages doing as good a job on your website as your

home page might be, where you spend a lot of love and attention, making it the perfect golden home page in the world? But what about these pages, where all these people are entering?

Measure bounce rate at a site level. Measure bounce rate for your core acquisition. Measure bounce rate for your top entry pages. If you're spending money on AdWords, measure bounce rate for your landing pages.

It will not give you all of the answers, but it will help you quickly distill down where things are not going right. Then you can say, is it that I'm bidding on the wrong keyword?

Is it that the creative on my landing pages is wrong? What's going on here? It's a quick qualifying metric. You're going to hear a lot of people say, well, I'm not going to sell. So you're going to hear more today about topics that I am going to skip.

You're going to have Stephanie and Alex talk about segmentation. I'm a huge fan of segmentation. At an aggregate level, it's hard to find tons of insights quickly. Segment, segment, segment, segment, that's the religion. Ten times a day, say, segmentation, segmentation,

segmentation. And they're going to talk about it. Stephanie's also going to talk a lot about search. And at the end of the day, Tom's going to talk about customer experience optimization and another thing I'm very passionate about. out.

So this is the last part of my presentation, rules for revolutionaries. And these are sort of my key learnings from having done this for a long time in companies big and small.

So the first one is a rule I had created three years ago. Until recently, I was the director for research and analytics at Intuit.

And so I'd done a presentation at E-Metrics Summit three years ago, and I'd created this rule, the 1090 rule. And the goal for everyone, every single one of you in the room, if you're here and paying

attention to me and not sleeping, the goal is you actually want to get lots of value from the implementation of analytics.

And the rule that I had was if you had $100, spend $10 on the cost of the tool and professional services for the tool, and spend $90 of that on the people.

The web is an extremely complex beast. There are more people on your website trying to do more weird things that will cause your data to look funny.

It's really, really, really important that if you want to get serious about analytics, that you don't simply throw a tag on your website. Most tools use tags now. You throw a tag and forget about it.

You will need somebody intelligent who is going to analyze the data, who's going to understand your business, your goals. So that's the first one. You can actually extract value.

Because at the end of the day, tools are only a way of getting data. And I find that a lot of people underestimate the value of business acumen and common sense

that you need to attack the data with. So no matter what web analytics tool you use, the differentiator for you between success

and failure will actually be, well, maybe all of you. I'm sorry. It's the investment that a company makes. I'm stunned that companies invest a million dollars in a tool and they give it to the admin.

It's not going to work. And it's not the fault of the tool. It's a damn good million dollar tool. I would love to have it. It's the fault of the company for not investing in the people.

The web is very complex and it will remain so for now. It's very important that you actually invest in people. I understand your businesses and can analyze the data.

The next rule I have is that reporting is not analysis. A lot of people confuse these two things.

And the interesting thing I find in my experience is that both of these activities take exactly the same amount of time. You could spend all your life doing reporting.

You could spend all your life doing analysis. And you have to make a conscious choice about what you want to do. These are dead trees. All right. Dead trees are arriving at your desk.

I love this animation. These are dead trees arriving at your desk, right? This is an actual Excel report that I was getting at that time, a couple of years ago.

This is not even the whole of it. There is more. These are prettier dead trees. Hold on. Sorry, Brad.

If you actually want to get value from your data. And you have somebody in your team whose title says analyst, this is what they should be doing.

Only 20% of the time spent doing reporting because you can't get away from it. But you saw so many different ways today early in my presentation in which you can do data discovery. You can get context. You can segment.

80% of somebody's time. Well, 70% because I'm a very generous person. I give 10% of the time for bathroom breaks.

But the rest of the time for somebody whose title says analyst, this is what they should do. Not all these days analyst should actually be spent in doing analysis, unstructured analysis.

You cannot expect to get massive insights from any analytics tool simply by implementing the tool. That is the start of your pain, not the end of it.

The other one is to avoid the data quality trap. And the original title of this slide was Data Quality Sucks, Get Over It! But I thought it would be rude.

One of the more interesting things about the web is it changes massively. The web is changing at the speed of light or even faster than if you can imagine it.

And it is probably the most perfect medium to collect imperfect data. It has been deliberately built to screw with you.

And the interesting things are there are many, many reasons why your data is not perfect and it's not going to tie. And if you spend time trying to figure that out, you will be a very unhappy person.

But the interesting thing is the web still provides probably more data and more valuable data than any other medium on the face of this planet. Now I am biased.

A, I'm a web person. B, I'm standing at Google. But I can't believe that people take out ads in magazines. Again. Remember, I am biased.

Because ads in magazines to me is a faith-based initiative. Thanks to President Bush for that term.

Because you say a Fortune magazine gets a million subscriptions a year and I will put my ad in it because I have a faith-based hunch that it will drive sales.

Or I will do a quick market survey and ask my mom if she saw the ad. That's a faith-based initiative. On the web, you can do better.

I'm stunned that people hold the web to a standard so much more higher than they hold the magazine ad. It's just simply insane. So my recommendation to you, remember, I want you to take action.

My recommendation to you is when you look at the data, assume a level of comfort with it.

When I analyze data and I know the web decently well, so I'll say, you know what, I have approximately a million subscriptions. I have approximately 80% confidence in data. And I can go present to John.

And John is a new person. He's always printed catalogs and he knows nothing about the web. I'm not saying that.

But John could look at the same piece of data and he'd say, I have 40% confidence in the data. It's okay. Don't argue with him. That's the first thing.

No matter what your level of comfort with the data, the nice thing is you can take a decision. Make a decision. Don't argue about confidence.

I was giving a speech at a very big company last month and I said, look, if I have 80% of the confidence in data and I come to speak to Rob, Rob is the CEO of this Fortune 10

company, Rob would say, whoa, let's go buy Google. 80% confidence in the data. But if he only had 40% of the confidence in data, he might say, let's go buy Yahoo.

Let's play with this for a little while. Let's make a less risky decision. The important thing is to make a decision. You might not want to send a rocket to the moon right away, but start assembling the engine.

What I find most people do is they bicker and argue about the quality of the data. Remember, even with 10% confidence in a piece of data, you can make a decision. Trust me, you can.

Then what you want to do over time is you want to take micro-segments in your data. So you can say, I want everybody who comes to my website to buy Google. Right, on the keyword Avenash sucks.

I will take that keyword traffic, just that, and over time I will try to understand why quality might be differing. What will happen is over time, your confidence in the data will go up.

You will move from 30% to 40%, 50%, 60%, 70%, 80%, never going to get to 100% quality because it can't. That's the problem. And that's OK.

At some point, we will all have RFID chips in our brains that will communicate with our website. web servers and give us great quality data. Until that day, make sure that you make decisions rather than

worry about quality too much. Because a good friend of mine, he gave me this wonderful quote that I love. An educated mistake is better than no action at all.

I am amazed at how often I forget this and amazed how often other people forget it. The next one that's very important, especially I

understand that a lot of you come from large companies. It's very important, if you want to create data-driven organizations, that you actually have a process and a

structure around data, around analysis, around taking action. Because most people say, and this actually happened to me when I first got to my prior company, people said, we have a report publishing schedule.

And that's a process. And that's not a process. This is actually a graphic, and it represents the Six Sigma Demake process. Very simple.

Define, measure, analyze, improve, control. Simple process. It's very important that for your web businesses, you actually figure out what a process is.

In this case, an example is I want to improve the merchandising capabilities of my website. A lot of people think when you talk about process, it's flow charts and complex things that take a billion years. It takes 50 years to do that. It takes 15 minutes to do the slide.

It is a PowerPoint slide. It tells the entire organization exactly how we behave in each step. Who is responsible for doing what in every single step?

It levels the playing field. It creates a repeatable process. As you go back and look at your analysis, you look at your reports, you look at your dashboards, you look at what

needs to happen in order to take a decision, to create a test. Make sure that you put some level of process in place because process scales. Ad hoc does not.

This is the last rule I think I have, which is experiment or go home. This is sort of a painful lesson I have learned over the last few years.

The lesson is that most web experiences are created by HIPPOs. HIPPO is an acronym that stands for the highest paid person's opinion.

Obviously, I have worked in a lot of companies. You take all this data you have, all these brilliant insights and all this great understanding, and you walk

into a room and the person who gets the highest salary, no, no, no, no, I want a woman running in slow motion against a blue background.

It doesn't matter what your data says. Since in my case, my HIPPO is worth $2 billion. $2 billion and pays my salary, the woman goes on the site.

And it has been really hard, at least until the last sort of 18 months to two years, to actually figure out how to get around the HIPPO problem.

How do you actually make decisions based on what your data is telling you, what your customers are telling you? And the interesting thing is HIPPOs come under all ways, shapes and forms. Some HIPPOs are like that.

Other HIPPOs are cuter. It is very, very common for me to interact with people in black turtlenecks with torn jeans. They are called creative directors.

Are there any here? I hope not. But they puke over all your ideas. The nice thing about testing and experimentation, this is why I'm a huge fan of it, is it creates a democracy.

You can now get a multivariate tool, an A-B testing tool, and you can say, if we have eight great ideas about what this page should look like, or what kinds of promotions we should run, or what is the best creative for an AdWords ad?

You get four lines and that's all you get. Test it. Why should you let your HIPPO decide what the customer experience, the promotion, the content, creative, bullets, font, what should it be?

Why don't you simply test it? You can test it. Testing is faster, it's cheaper, and it helps your customer tell you what you should be doing.

The more important thing I have learned about testing is that in most companies, 80% of the time, you are wrong about what your customers want. And the reason is really simple.

Just because I work at Google and I use the Google search engine does not mean that I'm a customer of Google. I am too close to the company. You're too close to your company.

Even if you use the software services of your company, even if you work at American Express and you use the American Express Bank. It does not make you a customer representative of the company. You're too close to the company.

The nice thing is you're going to be wrong. The wonderful thing about testing and experimentation is you can figure out how to be wrong quickly and efficiently. And that is an awesome thing.

And you're going to hear later at the end of the day from Tom about all of the wonderful things you have at your disposal to do this, but when you go back to your offices, you should...

Every single person, every one of you, no matter what tool you use, should figure out what your testing strategy is. Because it will be the difference from you swimming in data and taking no action to simply

throwing up experimentations and hypotheses on the website and let the action be taken for you by your customers. That's it. Thank you.

I'll be happy to answer a few questions and then we have a break at 10.30. And then at 11 we'll have the wonderful Stephanie. Yes.

About time on page. Yes. How does Google measure that with regard to tabs, people being on your page when you browse over the page? How is that measured and how does that impact that measure?

Paul can correct me if I'm wrong, but Google measures it exactly the way every other tool does. And the way all tools will measure time on site is most tools now use cookies and hence

when you first come to a website, they will start a session for you. And if it detects 29 minutes a day, it's going to start a session for you. And if it detects 30 minutes of inactivity, you will automatically terminate your session.

So it assumes that you have gone now to sleep or to meet with your boss. The time on site for the last page is the tricky problem for every single web analytics

tool because the way a web analytics tool will measure how long you have spent on this page is you look for the time stamp when you saw this page, you will look at the time stamp on the next page you saw.

And it will compute the delta. And it will say, you were here in minute one, you were here in minute two, so you spent one minute on this page. This is literally how most analytics tool compute time on a page. Then time on a page translates into time on a site, plus, plus, plus. Now the challenge with time on page, this especially affects blogs for example, is if

you only see one page on a website, the tool actually has no idea how long you have spent on this page. Because you load a page and you leave. So a great example of this is my blog.

On my blog, all my latest entries are on the home page of the blog. So almost everybody who comes to the blog sees one page, gets every single thing they

want and then they leave because essentially it's a one page website. In that case, it's much harder for me to compute what time on page is, how long people are spending. Now there are hacks and tricks. And DOM on logs.

There's a lot of code and pinging things you can do that will allow you to compute how long people spend on every single page including the last one. But most tools by default don't do that.

So the answer is, most tools use cookies. They create a session when customers come to the website. They use the delta to figure out how long you spend on each page.

And the time spent on the last page is a tricky proposition for now. Make sense? Paul, did I do it okay? Thanks. Okay. Yes?

So where? Where? On your AdWords campaigns or on your site? Yes.

AdWords. Just Google Analytics. Oh, in Google Analytics? Sure. Most non-humans, our beloved friend robots and crawlers, actually mercifully

do not execute JavaScript tags. So if you use Google Analytics or many other tools that are now standardized on using JavaScript tags to collect data.

The robots, so the non-humans, are actually not executing them, so they're automatically excluded. Non-human traffic was more of a problem during the weblog days because when crawlers crawl

your website, they do leave all the entries behind in your weblogs. So if you use weblog files as the source of your data, you have to manually filter them out.

And I think there is a website, robots.org or something, do a quick search in Google, and you'll find a list of all the robot user IDs. So if you want to filter out, you can do it, but with most tag-based analytics solutions,

this is not that much of an issue. There are a few robots, some obscure ones that are smart and execute JavaScript tags, but it's very, very rare that that happens.

Most robots, 99.99% will not execute JavaScript. But this brings up a great question. I am a huge fan of SEO. So if your website... On your homepage?

On your homepage, all the links, or most important links, are wrapped in JavaScript tags because you click on the link and it pops up a window. You click on a link and it does some magic with JavaScript.

Remember, the crawler from Google is not executing those JavaScript calls and it will not index that content. So it's very important to know robots don't execute JavaScript tags or JavaScript.

Hmm. Paul? Paul is the director for engineering for Google Analytics. And he's available. He's a very smart person who saves me all the time.

Any other questions?

Or your campaigns are coming from other searches? As long as you have campaign tags on your landing pages, and we will pick those up and we'll be able to determine that that's paid traffic.

If we don't see any marketing campaign information associated with the inbound traffic that's coming to your site, then we will then assume that this is an organic search or natural

search and we will put it in that category. Do you mean by help? Yep. Yes. Yes. Correct. Yeah.

And later today, I know Alex is going to show you how you can split organic traffic and paid, let's say, for Yahoo or for other search engines because you can filter out.

Every paid click comes with a piece of data that says it's a paid click. Now the interesting thing is most web analytics tools do tell you that we will automatically,

brilliantly, geniusly figure out what the difference is, but most of the time, no matter what tool you use, ask that question. And make sure, because if you're using some agency in New York who is playing with things

and redirecting traffic through their servers and sending it over here, there's a bunch of stripping going on and adding, in that process, it actually gets pretty hard for any analytics tool to actually figure out the split between organic and paid.

So most tools, by default, will do their best job to figure out what the split is, but for your tools, ask the question. Make sure it's working right.

So, for example, if I want to look at 30% where the curve got a little bit bad or...

Yeah. You can. You can. So, Alex, dude, you have a hard job today. He's going to show you how to segment and filter your data, but this is exactly what you would use. And I encourage you to do that, remember.

Data in aggregate is really hard for you to understand what's going on, so, especially wherever you're spending money, segment that data. So you do affiliate, AdWords, Overture, whatever, it's called something else now, YSM.

Whatever you spend money, make sure that you... You segment the data out so you can truly understand what populations make up the kind of traffic that you want. So you should filter. You should use profiles and filters.

And Alex is a great example later at 1 o'clock. So attend his session. Yes. For example.
That's not bad. Borderline bad. Yes. So the interesting thing is... I know. I know the answer. Yes. I want to show it to someone. Yes. It's really good. It's really bad.

Yes. Yes. Yes.

I believe that according to terms and conditions for Google Analytics, it does not mix and merge data across sites and things like that. So that is not a option right now. But what I advise people is every website is unique.

Your business of selling Sony bios is very unique. So index yourself. Index over time. Rather than looking at bounce rate for this month, why don't you compare it over the last 12 months and see if the line you're getting better or worse.

Because under visitors, I think there's a button. You press on it. It will give you that exactly for bounce rate. So index against yourself to see, hey, over time, is it going up or down? Because there is no ambiguity. It has to go down.

That's the clean thing. The other thing I would do is I would benchmark against other things that we're doing. So what is... I can say, oh, the site average is 70% bounce rate. But for AdWords, it's 12. Good. Right?

Because it's good to know what is good and evil. So use your own data for now. Also sometimes bounce rate is a tricky one. I don't know of a source where somebody's publishing bounce rate.

But if you sign up for shop.org, it publishes study every year, couple times a year, that creates and provides standards for conversion and things like that. I'm not aware for bounce rate though. I'll tell you that you would understand that.

Oh, yeah, yeah, yeah.

So at the moment, that data is not available. Yeah. Yes. There are lawyers at the back of the room watching me.

I'm kidding. I'm kidding. Once a month. I'm kidding. If you could fancy up one of every enterprise solution that's out there, uses live reports,

uses Omniture website, and give me a picture as to where does Google Web Analytics fit in that picture?

It actually fits perfectly into that picture. Because it's actually not uncommon. I was talking to someone earlier today. And he said, I am sometimes surprised that people keep switching between the top tools.

It's kind of the same. There's no point in you switching from one to the other. to the other. If you don't like it, just stick with the enterprise class tool that you have

because it's as good as the other one that somebody's anointed it as an enterprise class tool. I think every tool has its unique value proposition. Omniture has its own proposition,

HBX its own, Webtrends its own, GA its own. One of the things that is really hard to do but I encourage you to do is figure out what is the core strength of the tool that you

have, and does it match with what you actually need? In my mind, the core strengths of Google Analytics are some of the things I've covered today.

It is extremely easy to use, it's extremely efficient at helping you discover data and trends and insight, and it has many, many things built in to make sure that even the most

expert person or the most simple person can understand the data that's there and take action from it. It's very, very good at it. I especially love the AdWords section of the Google Analytics tool.

So I wish I was doing Stephanie's section, because I would like to be doing my own. I've got a bunch of exams, all over the place. It's just amazing, I'm surprised. I should not have been obviously, but I was surprised at how good the AdWords reporting is.

It's a core strength. The loyalty metrics you saw today is a really good strength. Each tool has a strength. I wrote a really long post last month, I think, on my blog and

it says, what is an enterprise class tool? Who defines it? Who gets to define it? And I think the days of somebody, some analyst or author or random guru through telling you what tool you need.

I think the nice thing is that you have choices, and you should figure out what you need, because every tool has a strength. The nice thing about GA is I would say, you know, you can get it for free, and you can install it on your website.

You can try it, and you can actually learn by doing. Because the one core difference is that if you use any other tool, you will learn what it does by a bunch of marketing slides.

Right? That's exactly how you do in RFP. And you can actually use it. Put it in your site, learn, use the data. Because after you implement the tag, two and a half hours

later, you have data. All data, all standard reports, any tool will provide. Put it up there. Try it. And through the learning experience, you might figure out that Google is the right tool for you.

Or you'll figure out it's not the right tool for you. But the interesting thing is, if you do arrive at a conclusion that it's not the right tool for you, when you go to the next vendor, you will ask them intelligent questions.

Because you would have figured out what you need. Best case scenario, you use it. Worst case scenario, you get smart. Win-win.

You cannot import data into Google Analytics at the moment. Nick's here. Nick, where are you? I saw him before. Nick will be here later today. And you can talk to him about tracking online, offline.

There are many different things you can actually do to track online and offline. There are lots of case studies and real examples that we'll make sure you get to talk to Nick today.

So like click to call, or ads in a magazine, or flyers on your Office Depot store. All of these things actually can be tracked by doing a few different things. It won't give you perfect, perfect data.

But it will give you a very good, solid indication. Better than faith-based initiatives. It is totally possible. I think Jeff, at the end of the day, is going to give you a So it is totally possible. I think Jeff, at the end of the day, is going to give you a might be touching on audio ads, for example. That to me is a great example.

How the hell do you track that audio ad? So you can do online, offline tracking with web analytics tools now. We'll make sure that there's Nick. That's the guy standing over there. Go find him.

And he has some real examples and case studies where he can show you exactly how this tracking can be done. Even though you can't import the data, you can track and measure success. Yes?

You're cutting into your own eating time. How do you get past the whole data quality argument? Because that's probably one of the largest ones we face, where someone has an Omniture. Yes. Yes.

Well, Google Analytics shows this. Yes. Mark Tool shows this. Tell us why Google is wrong or Google is right. Yes. The interesting thing is neither one of them is right or wrong.

The beautiful thing is web analytics is everybody can create their own version of history. No, but the interesting thing is every single tool actually collects data differently.

And every single tool, well, slightly differently at least, even if they use the same mechanisms, they will collect the data slightly differently. So the interesting thing is the data is not going to tie. It simply won't.

And there is, so the core, usually, usually, the differences fall into these buckets. Either they're using first or third party cookies. That explains a bunch of stuff.

The way that they actually start and end sessions, most tools differ in when they terminate a session. Like I mentioned, 29 minutes, the 29-minute session timeout.

Some people actually have, other tools have other intelligence, intelligence that terminates a session. So normally that's another one that's the difference.

And another one, sadly, but very common, is that the two tools won't be all sitting on the same pages. So the tagging is not right.

Usually, usually, most of the things will fall into these three categories, the differences, why the numbers are different. But they will be different. So if you do end up in this situation.

The question to ask not, is not what your country, no I'm just kidding, but the question is not which one is better or worse. I have seven tools on my site, seven. I need more.
I simply compare the trends. And what you will notice is, if you have a couple different tools, and you train the data for a few months, you will notice that the delta stays the same. The delta actually stays the same.

It is very difficult for you to say, my numbers from tool A, they're not the same. So if you have a couple different tools and you do try to match them, you will notice that the numbers from tool B will match tool B. Because they fundamentally are doing different things.

If you do have a paid vendor, stress them and say, why are your numbers different to GA? That is a cute tech question to ask. But mostly they won't tie. Trend it over time. Make sure that's good.

If you notice something funny like this month, the delta is 20%. And next month, the delta is two, something funny is going on. You want to investigate that, but you'll notice the trend is the same.

Yes. Last question. Yes. Yes.

You can't. You can't. But most, and that's true for most analytics tools. So the question was, if I, you can have four goals in GA, what if you run, create a site,

put the tag, and you're running it for a few days, and then you realize, oh, I forgot goal four. Four. You cannot go back in history and retroactively reprocess data. And this is true for most analytics tools.

There are solutions now where you can download and do things like that, or some solutions of data warehouses, things like that. But your core analytics data, once collected for every vendor, is fairly well set.

It's kind of harder to go back in history. There are a couple tools that do it, but mostly you can't. It's just the way the data is collected, stored, and processed. Yes. What's the privacy policy for Google Analytics?

Google.com, and then at the bottom, it says privacy. There are terms and conditions on every single Google page. You will actually see a privacy policy. It's very long and expansive.

So please, every single person should read it. Every single, no, I'm not kidding. Like if you're worried about privacy, you should read it. Privacy is a question that comes up often about Google. You should make sure you're comfortable with it.

Is there a specific aspect of privacy you wanted me to touch? Because it's like 12, 15, I don't know. 900 pages. I personally don't have a law degree, so I don't think I would be able to read the implications. What's your concern?

Well, the concern is you guys having all that data on a website, the idea of how to do that,

who you'd share that with, and stuff like that. My recommendation is read the policy. The thing is, if you're running a business, you're not being fair to your business if you don't read the policy.

Please read it, because you're right, it is an important thing to read. The policy states out very clearly what Google will and will not do with the data. So for example, in what you said, there are things that Google will not do with your data.

They're very explicitly stated out and they touch exactly what you said. Are you going to share? Are you going to monetize? Are you going to spy on me? Rather than me saying, trust me, I'm going to say, go read the policy.

Actually it's, and there is actually a two paragraph section that touches on that. It's very short. Find it, read it, and from your privacy is in the eye of the beholder.

The best thing you should do is read the policy, make sure you're comfortable with it, because there are things that Google can do with your data, and there are things that it will not do with your data, and those are laid out fairly clearly.

But it is open to interpretation, which is, I find, different for each person. That's okay. But it is everywhere. It's very, very... The easiest thing to find in Google Analytics is not a report. It's the privacy policy.

Are there plenty of examples of companies that have policies and change them? Yes. Yes. If Google... It was the state of policy and then a year later, change it, and then use the data.

So most reasonable companies, I would put Google in the reasonable bucket.

Most reasonable companies, by law, if they change terms and conditions on you, will actually make sure you explicitly are aware of it. Yeah, who did it? Yeah, I'm involved... Yahoo.

Yahoo would take the policy without telling anybody. Right. So you look back, and you can find out . And see, the wonderful thing is if you change your policy without telling your customers, that customer will show up somewhere else and tell everyone else about it. Right?

And... It's a really, really big thing. Yeah. No, it is. It is. And you should... Most reasonable companies, I find, that will not change terms and conditions without informing you about it. It's the right thing to do. Right.

It's a very googly thing to do. Thank you.