A common theme I have noticed across many Web Analytics practitioners is that the moment they get access to the data they dive, and they dive deep.
You open the tool and they are instantly on: “how is that page doing” or “can I report on all the campaigns that are driving traffic to these six pages and measure retention” or “ I want to pull out these six pages and measure exit rates” or “can the tool email me a daily summary of all the visitors who visit my page and don’t submit” or “I have my key performance indicators (KPI’s) and I want a detailed report across all top pages on the website.
The theme, almost 100% of the time, is that the users go into micro-reporting, but they have often spent very little time either understanding business website goals or understanding what is happening at a macro level. Rarely do they bother doing what's perhaps key: spend time understanding the bigger patterns in the data.
If you are into Web Analytics almost always start with a rock solid understanding of understanding data at a macro level and resist the urge start understanding data at a micro level. You might not see the forest for all the trees.
The amazing thing is that this the “simplest” thing you can do and it is is surprising how many people don’t spend time on it (it is perhaps hard to resist the allure of having every piece of data you could possibly want on your finger tips, for every page of your site and for every person who comes to your site).
This is not very complicated. Here are four simple macro-questions that you should answer before you go anywhere deep:
# 1: How many visitors are coming to your website?
This is the simplest first question you can and should answer. Measure Visits to your website (sum of sessions) and measure Unique Visitors (sum of unique persistent cookie_id’s).
These metrics look deceptively simple, they are not. For more on what you should look for and things to be careful about please read this post: Standard Metrics Revisited: #1: Visitors.
For both of these metrics see how you are doing, is it what you had expected, are they being measured correctly (they might not be, refer to the aforementioned post). Get a comfort level that you understand these metrics and understand them well.
Quick Note: It is optimal to avoid diving into number of repeat visitors, and this rate or that view and all that, for now. Just for now.
# 2: Where are they coming from?
The second question is a even better one, so full of promise and hope and goodness. Where do all these visitors come from. Look at two reports: Referring URL’s and Search Keywords.
You will almost always be surprised at how people find you (and your guess as to how they might be finding you are probably misplaced).
Referring URL’s helps you understand which websites are sending you traffic and which are not. It is a great way to begin to understand both what you are doing that is causing traffic to come (relationships, direct marketing, other campaigns, affiliates etc) and what you have not done that might be causing traffic.
Look for surprises, you will find them.
With search dominating the landscape look for how much traffic you are getting for search engines (in your referring url’s report) and then dive deeper into what key words and key phrases are sending traffic from each search engine. This is a gold mine of actionability, specifically for Search Engine Optimization (SEO) and, if you are big enough, Pay Per Click (PPC) / Search Engine Marketing (SEM).
Look for non-branded keywords (they will indicate you are getting prospects – people early in the consideration cycle) and secondarily if you are getting traffic at the right level for your branded keywords.
Quick Note: Notice we are not doing which countries or states or zip codes and all that. Unless you are in a deeply geo specific business (say in Europe) this can be a distraction at this early stage.
# 3: What do you want them to do on the website?
The problem with web analytics data is that once you get access it can be such a huge time sink (especially for geeks like me). Every place you turn there is a new piece of data, a new rat hole you can go down on, another mirage you can follow. And it can be kind of fun.
Don’t do that! :) Step away from your website and take a long cold hard look at yourself and your business.
Then answer these simple questions:
A) Why does your website exist?
B) What are your top three web strategies that you are working on (could be paid campaigns, could be affiliates, could be updating content on the site, could be you are trying to get digg’ed, could be effective merchandising, etc.)
C) What do you think should be happening on your website?
Write down the answers and publish them far and wide in your company and your local newspapers.
The output (your answers) could be metrics or KPI’s that you think measure success (though this might be unlikely), it could be simply a list of acquisition strategies for your website (SEO, PPC, DM etc) or it could be a mission statement that somehow ties to your company bottom line. The altitude you can calibrate later, it is important that you have some precision in what you want your customers to do on your website.
Quick Note: Marketers, Analysts, Website Owners: Notice this is the third question and not the first one. That is because I think you should have some semblance of context from your web data to even think about this clearly. Often sans the web data you really don’t have the low amount of basic web reality understanding that would help you answer these questions correctly.
# 4 What are they actually doing?
Your first shallow dive into the data. Look at these four things in your reports:
A) Top Entry Pages: Learn how people are getting into your website.
B) Top Viewed Pages: This is a great way to know what content is being consumed, it will probably be different from what you think should be consumed so it is a great way to get educated. It can also help you, in conjunction with top entry pages, why people end up looking at what they do.
C) Site Overlay (Click Density) Analysis: For your top viewed pages look at the site overlay report and analyze the click patterns (only on the top ten most viewed pages on your site, to keep it simple). It will help you understand navigational challenges on your website, it will help you understand visitor intent, it will suggest optimization actions you can take.
The goal is for you to simply get acclimated with content consumption and navigation behavior on your website. This will give you so much more context and a richer understanding of customer behavior. That in turn will be critical as you dive into measure obvious famous metrics like Conversion Rate.
Quick Note: The ordering here is important. We tend to dive directly into measuring conversion rate and it will turn out to be rather pathetic and then we work our way backwards (with our eyes closed) and inevitably it is frustrating. A better ordering is to understand customer experience to the extent you can with these simple reports and then work forwards. Also notice we have not done path analysis.
I am sure you will agree with me that there is nothing particularly genius about the questions above or the suggested reports. They are simple and straight forward. My goal is to simply encourage those who have just started with web analytics, or those who are currently frustrated, to focus on macro analysis and not step into the quicksand of micro analysis. Far too often we all go micro, and sadly we are never able to go back.
Would you agree that these are good questions for macro analysis? Do you have a good understanding of macro behavior for your website? How can this be improved? Please share your feedback via comments.
[Like this post? For more posts like this please click here.]
Hi Avinash,
Enjoyed reading your post.
Indeed, analysis of referring hosts is full of surprises. You notice traffic stemming from all over the planet (mostly via G) and realize the power & potential of the medium (web).
The q parameter gives a clear idea on hot keywords and keyphrases that generate traffic. Very actionable info as you suggest.
‘Often sans the web data…’ ‘basic web reality understanding ‘ Some marketeers think of the web channel as just another customer touch point, like the call center.
“understand visitor intent”: as we all know this is key to becoming effective. Research on that is currently underway by the major SE players.
You define an initial analysis level (‘shallow dive’) which allows to later proceed to a higher analytics maturity level in an informed way.
K
Good post Avinash. I have recently started reading your posts and I think you bring up valid points.
Having said that, I believe there is value to micro-analysis to support or find more details that emerged from the macro-analysis.
Example, you find that the top 10 pages includes certain pages that you never expected people to visit. In this case, a detailed analysis might be necessary.
To summarize, micro-analysis should always lead from a macro-analysis and never be a project unto itself.
– S
Would you elaborate on one point? You wrote:
What does it mean to get traffic "at the right level for your branded keywords"?
Thanks!
Great post Avinash. I like your blog a lot.
Just FYI – The image you put in #3 has Korean words within the image and they are not matching with the context. (I guess you don't know how to read Korean) It bothers me little bit since I am a native Korean speaker :-)
Jonghee: You are right I don't speak or read Korean (not yet!). I am sorry you were upset about the image. In my searching for the right it seemed to fit (four people seated on chairs and some kind of process / step description boxes moving towards right).
I have replace that image with one of my own creation. I hope you like this one.
Thanks,
Avinash.
Thanks for the quick action! This new image looks better…
Greg: Ahh good question. Tough for me to answer but good question! :)
Most websites today will get a large chunk of their traffic from search engines. There isn't a "standard" number but it is common to see 30% – 35% traffic from search engines (most of it Google). That's true for e-commerce websites or non e-commerce or even this little blog.
Of that traffic a large percent, (again YMMV) 70 – 80%, will come from the top 15 or so keywords).
Frequently most of these top keywords will be associated with your most obvious identity phrases (brand). This is because it is the easiest ranking to get in terms of SEO (after all that is who you are) and it is common for most companies to spend money on PPC / SEM on their brand key words / phrases.
With that context the answer to your question…..
The right level depends on your business and your web strategy. From a marketing and sales perspective the right level of traffic on branded key words should probably not be 80% because that would mean people find you after they know about you already.
You want to capture people earlier in the "consideration process". For example it is easy for people to type in "acteva" and find your site. But by then they have already decided they want to use your service.
You want to capture people who are typing in "online registration" or "event planning" or "event payment" etc because these are the "undecideds" and more valuable to convert.
In that case maybe, I am hypothesizing, that the "right level" for traffic from brand key words is 60% and from non-branded key words is 40%. Or to have five of your top 15 key phrases to be non-branded.
By saying "at the right level for your branded keywords" I am trying to, perhaps poorly, emphasize the need to focus on non-branded key words and key phrases as a effective macro web acquisition strategy.
This can be a competitive advantage because it is hard to do and most people obsess about brand keywords almost all of the time.
Hope I have atleast somewhat answered your question.
-Avinash.
PS: Sorry for the late reply, your comment for some reason was stuck in the blog spam folder.
Since we met last week, I have had several philosophical conversations with customers. They are so concerned with finding the ‘diamond in the rough’, they are completely overlooking the basics. You did say that success with Analytics starts with a strategy of leveraging the web. If that is not in place, then everything else is set for failure.
I'd like to add another question to your list:
WHY are they here?
Whether we're talking about a retailer's site or a financial services site, or even a B2B firm's site, it's important for marketers to understand why customers and prospects are visiting the site AT THAT PARTICULAR POINT IN TIME.
What can analytics say about the INTENT of the site visitor based on that visitor's behavior?
What can analytics say about the group of site visitors over the past month and what the top 3 reasons were for why they came to the site?
And how have these reasons changed (if at all) over the past year?
This is what I'm thinking of when I say that analytics needs to broaden its horizons from the micro to the macro in this post:
http://marketingroi.wordpress.com/2007/02/02/response-to-marketingnpvs-predictive-analytics-predictions/
Ron: I agree with the question you proposed 100%. But I would put it after the "Getting Started" phase. Mostly to keep things simple initially.
In my humble experience it can be, surprisingly, hard to be able to measure it with any degree of accuracy with clickstream data (which is what most web analytics applications have access to). No matter how hard you try, all you are doing is interpreting clicks (and sadly they don't "talk" very well).
To answer exactly the question you have mentioned I have proposed measuring Primary Purpose ("Why are you here today?") and measuring the Task Completion rate for each of those Primary Purposes (notice this is the customer's rating so clearly the emphasis is on how they feel – how's that for customer centricity!).
More on Primary Purpose on any one of these older posts:
Stop Obsessing About Conversion Rate
Measure the Real Conversion Rate & “Opportunity Pie”
Standard Metrics Revisited: #2: Top Exit Pages
Thanks so much for your comment and feedback, always appreciated.
-Avinash.
Avinash —
I don't care where you put it (I'm not trying to be flip).
But putting the question I proposed on the table does a couple of things:
1) [Hopefully] forces the business to think about what they know (and don't know) about the customer's decision making cycle, and
2) Makes Analytics think about what data they have (and don't have) to help answer #1.
I strongly believe that at the end of the day, there isn't going to be a neat, mathematical equation that answers the question I proposed. But part of developing strategies (marketing strategy, business strategy, whatever) is developing THEORIES about the way the world works.
Analytics can bring data to the table to help shape those theories. It's OK if it doesn't all fit neatly into an equation. Sr. execs are looking for ideas and theories about what will succeed — and if they can be backed up w/ data, all the better.
Excellent tips on how to get started. I talk with quite a few people who when they are just getting started want to get involved in click paths and more complicated ways of tracking visitors instead of establishing the basics. As it was just a few weeks ago I would remind them that they have to play the regular season before they get to the Super Bowl. Because of this I often stress for them to look at the summary reports first as they do encompass quite a few of your recommendations at the macro level.
Ron, Avinash:
After reading the notes, I feel like this issue could be "solved" by a combination of clickstream data analysis and survey.
Clickstream analysis will help you segment the different customers based on their behavior. You can then select samples from each set and present them with a survey to "just ask" – nicely, ofcourse – how they currently use the site, and what would they like to use the site for, going forward.
To drive the point home (and that I like to harp on sometimes), you could survey the customers directly (read, blindly). But, when done after proper segmentation, it increases confidence in survey results – that they accurately represent the various segments of customers you have.
Makes sense?
S: Great suggestion. Surveys can be a great source of understanding the Why. They are not God's answer to every problem but applied correctly they are very good at continual listening (almost as insightful as doing clickstream analysis).
Here is a post I had written a while back:
Got Surveys? Recommendations from the Trenches
Thanks so much for the feedback.
-Avinash.
Nice article.
The problem that I saw the most is that people have the stats but don't use them.
They have modern analytic software but fail to judge the results und take the right steps from there.
So often the problem is not located within the computer, but in front of it.
Keep up the nice work.
very nice articles and a contribution the web analyse statictics theory
thanks
best regards
Frank
I do get about 500 visits per day to my site, but was never really sure about how to use the information captured in GA. With this post, I will start creating some metrics and track these weekly. I will keep you posted on my findings. Thanks a bunch for this Post.
Hi Avinash,
I am enjoying your book thoroughly! Great topics and excellent writing.
Thanks,
Sonia
Read most of your articles related to site metrics. It really enriched my knowledge. Thanks again.
This is a really great post for people new to analytics. It can be sooo easy to get lost in data or to end up reporting metrics that you thought were useful but actually aren't (at least from the perspective of your site's goals, that is).
Avinash, I'm going to be taking the UBC analytics course. Are you still teaching?
Hi
I am a great fan!
I am reading your book at the moment and WOW it is hleping me alot!
I have been doing Analytics for the past couple of months and I do my reports in the office for my clients sites.
I have a question for you…
Which is the best way to set up goals???
Do we write out the entire url for example http://www.netage.co.za/pages/online-quote/complete
or
/pages/online-quote/complete
Which is the correct way???
Quotable quote: "The problem with web analytics data is that once you get access it can be such a huge time sink (especially for geeks like me). Every place you turn there is a new piece of data, a new rat hole you can go down on, another mirage you can follow. And it can be kind of fun." :)
Thank you for this outstanding article.
The problem is to draw the right conclusions from the software!
I am reading your book at the moment and WOW it is hleping me alot!
Hi Avinash,
Thanks for the post. These pointers are a good starting point for analyzing data. Just one question: How do you measure Most Viewed Pages?
I first started the micro way as well, but you are totally right, the wrong path to go.
Change the first instance of the word "rarely" to "often". No need to publish this comment.
I am using Clint Ivy's spreadsheet and it is working fine except for a couple of things:
1) How does he get just the outliers to show up?
2) How does he get the range between UCL and LCL to be shaded, as shown in his graph example?
I have looked at the formula in his example but it does not seem to make sense.
Anybody have suggestions?
Thanks in advance!
Indu
Thanks for the nice post in this blog. ;-)
Thank you for this outstanding article.