Beginner's Guide To Web Data Analysis: Ten Steps To Love & Success
The goal of my recent post on the Yahoo! Web Analytics blog was to pull us up 10,000 feet to do something we do less than 1% of the time in the web analytics world – look at the bigger business picture.
It was called: Secret To Winning With Web Analytics? Starting Right!
While that was a very strategic post, it got me thinking at a tactical level.
What if I was given the login and password to someone's web analytics data and asked to "find something interesting?" How would I start the process of web data analysis right? Even without any knowledge of the company's goals or help from a stubborn HiPPO or clients who just want data pukes? Can I add any business value?
A real challenge!
It turns out, astonishingly, that even with all those barriers (no objectives or goals or cooperation or business guidance), you can spend a couple hours and do decent enough analysis, sourced from your experience, to deliver some minor data-gasms of insights.
Not quite the real intense ones that you might experience if all the foreplay had been done correctly (see the YWA post above), but still never say no to even minor orgasms right?
Setting The Right Expectations
It is nearly impossible to find earth shattering insights that you can action from your web analytics data in just a couple hours. And yet finding some delightful starting points might be less hard than you might imagine.
Starting points to start valuable analysis from. (What data should I look at first?) Starting points for a customer centric strategy. (What are my customers telling me?) Starting points for gaps in your online marketing efforts. (Where am I wasting money?)
Secret To Winning Web Analytics: 10 Starting Points For A Fabulous Start!
This blog post is a starter guide that outlines the steps I personally undertake most commonly when handed the keys to the data for a website.
I want to share where in your web analytics data you can find valuable starting points, even without any context about the site / business / priorities. Reports to look at, KPIs to evaluate, inferences to make.
I hope you'll benefit from my humble experience. Let's go!
Step #1: Visit the website. Note objectives, customer experience, suckiness.
My biggest beef with web analysts and consultants is how quick they are to jump into Google Analytics or Omniture or WebTrends. It's as if they have never seen a report with Visits & Conversions before. Jeez!
The very first thing I do, and I recommend you do, is visit the website whose data you are analyzing. See how it looks. Go to the product pages. Go to the donation pages. Go to the B2B dancing monkey video (what!). Go to the add to cart page. Go to the RSS / Email sign up page and sign up. Go read some customer reviews (if a ecommerce site) or visitor comments (if a blog). Go download the white papers. Go use site search.
Get a feel for the company's vibe. Get a feel for the information architecture and cross sells and font size and buttons and tab structure and user experience etc. What's hideous? What's awesome?
Bonus points for visiting one competitor's website. Do all of the above.
Take out a note pad and write down your thoughts. What did you like? What did you hate? What frustrated you? What was obviously broken? What's the site trying to do?
At the very minimum your notepad should contain answers to these two questions: What is the macro-conversion? What are two or three micro-conversions? Remember those terms apply to ecommerce and non-ecommerce websites.
The site owner / client did not help you, but you've not got super valuable context. You're ready for data!
.
Step #2: How good is the acquisition strategy? Traffic Sources Report.
This is the very first place I end up because the first thing I want to know is how savvy the company is about online marketing. All other site data comes second because if you stink at online marketing then there is not much of a victory to be had by torturing website data.
No company in the Milky Way has succeeded without having a balanced portfolio of acquisition channels (fancy word for source of traffic). How's yours?
What to look for:
I am really looking for a balanced portfolio of traffic sources. Search, Referring Sites, Direct, Campaigns. Which one is strong? Which one is missing?
Based on my own humble experience the site on the left is what approximates the kind of "best practice" (note the quotes) you are looking for.
Around 40% to 50% Search is normal. If the number is too big (site on the right) it indicates an overexposure to search rankings and algorithm changes (not good at all). If it is too low you are simply leaving money on the table. And of the search traffic, you want a big portion to be Organic so you are not just "renting" traffic or suck at SEO.
20% or so Direct Traffic. If the web analytics tool is implemented right these are all your existing customers or people from offline campaigns. You want a healthy amount of both. If direct traffic is low, I worry if you are any good at customer service / retention (the latter is so often just an afterthought).
20% to 30% Referring Sites. You can't just rely on search engines or spending money on campaigns. A healthy web strategy includes a robust amount of traffic from other sites that link to your products and services, and praise (or slam!) you, or promote you on Twitter and Facebook and forums and otherwise link to you. Free traffic (usually) and you do want that (for many reasons).
10% Campaigns. Google Analytics (sub optimally) calls this Other. It is email campaigns, display / banner ad campaigns, Facebook display campaigns, social media campaigns etc. You want at least 10% of the traffic to be the ones you invite to your site deliberately, after solid analysis and great targeting. Outside of Paid Search. It's a sign of a healthy business that has a diversified customer acquisition strategy.
Consider the above as broad guidelines, again based on my online marketing experience. YMMV, certainly for esoteric types of businesses.
What to do next:
I'll make note of where the company is overleveraged and make a note to dig deeper with the client / HiPPO. Expose the dangers to them, brainstorm how to diversify.
For each bucket I'll look at least the top ten rows. Additionally, for at least one of the four buckets I'll dig deeper by looking at the standard report for that segment to identify some strengths or weaknesses. Surprising keywords, missing sources of traffic, trends in campaign vs. direct visits etc.
At the end of this you'll understand how sophisticated the client is, where you'll attack acquisition first (if you get the time and money to do more analysis).
Step #3: How strongly do Visitors orbit the website? Visitor Loyalty & Recency.
I have a sense for the site and I have a sense for the client's acquisition savvy. Time to focus on the Visitors!!
Most people create sites just for themselves and with no obvious purpose in mind. Furthermore the content publishing schedules, perceptions of "engagement" are all out of whack.
So what I (and you, dear blog reader) want to do is get a sense for how strongly attached are the Visitors to the site. This is of course crucial for any type of content site, but you'll be surprised at how important it is even for an ecommerce website (retention, support, repeat purchases, yada yada yada).
The report I'll look at is the standard Visitor Loyalty report. It would show how many times in a given time period the same person (persistent cookie actually) visits the website. Or, how tightly the Visitor orbits the site!
All tools have this report, Here is how it looks like in Google Analytics:
What to look for:
For site number one it is clear there are a lot of one night stands (47%). But notice that bottom, an astonishing 40% of the people visit the site more than 9 times a month! You get a sense for content consumption patterns, you get a sense for your next task (segment the 40, learn what's working there, apply to the 47!), you get a sense for whether the site's delivering on its business objectives.
If the data looks more like site two, cry. Ok, most of the time cry. This site simply engages in one night stands, and while I can think of some sites where that can still be the basis of a long term sustainable business model. . . I can't think of a lot of them.
Not a tight orbit. So what? Remember the notes you took in step one? What's the impact of the loyalty pattern on the objectives you noted? With some solid data you are ready to have a discussion with the client / site owner / HiPPO. Take a tissue.
While I love Loyalty the most, I also take a quick peek at Visitor Recency. This is specific to content sites (newspapers, yellow pages, hospital, "I am the next facebook-killer" sites, etc).
Visitor Recency measures the gap between two visits of the same visitor. Or, When was the last time you saw the same person (cookie really). Here's the report:
How amazing is it that 37% of the traffic on the site last visited less than 24 hours ago! Talk about orbit!
Segmenting this data is best (by content, source, campaign, outcomes, etc), but even a cursory review will help you understand how people behave.
What to do next:
I always review two more reports that give me a sense for content consumed. No, no, not the silly reports that show mostly useless metrics like Average Time on Site and Average Pages Per Visitors (averages stink!).
I am talking about Length of Visit and Depth of Visit:
With Loyalty and Recency we measured visitors visiting, but once they are here what are they doing? That's what you are trying to get a sense for with these two reports. [Remember visits with just one page view, bounces, will be in the first bucket 0-10. For why, see: How time on site & time on page are computed]
If you have some time segment out the bigger buckets (beyond 0-10) and analyze the data. If you don't have time just knowing Loyalty, Recency, Length, Depth tells you a lot about how tightly Visitors orbit this site, and understanding customers is precious.
Step #4: What can I find that is broken and quickly fixable? Top Landing Pages.
Understand site? Check! Understand traffic sources? Check! Visitors? Check!
Time to take off the gloves and some clothes and get dirty.
Companies spend lots of money acquiring traffic, often badly, so why not find top places where that money is being wasted and which home pages might possibly be stinky? Visitors refuse to give you a single click? That might be a useful signal. : )
In your web analytics tool this is a standard report. It shows bounce rates, sweetly indexed against site average, for the top entry points to the website:
What to look for:
The red parts! See why I like "indexed against site average" feature? It is easy to know what smells bad.
Three landing pages (entry points to your site) are performing really well. Seven seem to be bouncing at a much higher rate than normal, some spectacularly so. At this point you don't know why.
When you see that a page has a high bounce rate it could mean one of two things:
1. Wrong people are coming to your site (highlighting problems with campaigns, SEO, etc) or
2. The page itself is poorly constructed (missing calls to action etc) or otherwise broken.
At this point you don't know which of the two (or both) is the problem. Since you don't have a lot of time pick two of the biggest losers above. Click on the arrow thingy next to their name in the above report and visit them. Sometimes the problem is obvious. Next click on the link itself in the above report and visit the page level report. There in the drop-down pick Entrance Sources and Entrance Keywords. That segmented view will quickly tell you which sources and / or keywords are contributing huge bounces.
Now, at least for two or three pages of the site you are analyzing you know that they stink and you have decent clues of what the cause(s) might be. Give yourself a pat on the back. Great job!
[An exception: Analyzing bounce rates for a blog, or "bloggish" site? Segment and look at landing pages for New Visitors; for all other sites the method is as above.]
What to do next:
In this case you are in a position to recommend specific fixes. You have looked at the pages and sources of traffic (proxy for customer intent). You can use a heuristic evaluation process to tell the site owner what fixes will help reduce bounce rates.
A clean and handy checklist is here: Qualitative Analytics: Heuristic Evaluations Rock!
I am telling you people are going to love you for being this awesome.
Step #5: What content makes us most money? $Index Value Metric.
Most effort on any given website is spent on content creation, and hence for step five I encourage you to stick with page reports, but flip the funnel from an "input metric," bounce rate, to an "output metric," $Index value.
For an ecommerce (with revenue) or a non-ecommerce website (where goal values have been defined) $ Index value essentially computes "how much revenue" has been earned by a page (more like contributed by a page). It is a great way to gauge the value of a page.
Go to any content report, in this case Content By Title and you'll see the last column:
What to look for:
Quite simply the types of content that add most value to an ultimate outcome for your website. In the above screenshot it is pretty clear that even in the top ten the range of value added is from $8.79 to $0.10.
Would your boss / client not die and go to heaven if you told them exactly what types of content they should be writing / pimping more and what less? How about pages of which product / services generate most value?
Or my favorite report to look at: Content by Drilldown.
That report is particularly apt for sites that are organized in clean directory structures (like /products, /videos, /demos, /blog, /whatever else). Now you are able to discern which groups of content is most valuable to the company. Are videos really valuable? How about really heavy painful Silverlight demos? Wish lists work? The answer awaits!
If you don't have clean directory structure you can still use segmentation to group content and do the above.
What to do next:
This takes very little time. Use the Analytics Weighted Sort feature.
What I am doing above is answering this question: "Forgetting about top $index value pages and the bottom ones, which currently high value $index pages should I focus on to have the highest impact on my bottom-line?"
That is a very hard question to answer unless your web analytics tool has algorithmic intelligence built in. You click that check box and boom! There's the answer that will endear you to your boss / client for a long time. Hugs might even come into play.
Focus not just on what's causing good stuff today, focus on hidden areas where good stuff might happen in future.
Step #6: How Sophisticated Is Their Search Strategy? Keyword Tag Clouds.
Pages, and valuable bits of content, done. Time to refocus on high value acquisition.
Search.
It is really hard to get a "big impact" understanding of search strategy sophistication just from tables of top ten rows in Google Analytics or Omniture or CoreMetrics. Mostly because I don't want to look at the same lame obvious things.
So I like to yank the data out and create a tag cloud of all 40, 50, 100 thousand rows of data. Export as CSV. All Rows. Paste into www.wordle.net Magic:
What to look for:
[All data in examples below was taken from competitive intelligence tools as I don't have access to data for sites used here.]
A tag cloud very quickly shows the story in hundreds of thousands of keywords. In the case above, The Church of Jesus Christ of Latter-day Saints, it becomes quickly apparent that the Mormon Church has done a near magnificent job with search.
The brand dominates (as it should), but what is truly impressive is how many other words are also prominent (the valuable non-brand ones, even the long tail ones). If you know even a little bit about the Church you'll also be impressed that this tag cloud is a validation that the words the LDS church would like to be associated with it are prominent, and ones it does not are not as prominent. An amazing job by them. Pretty easy to see the themes (music, scriptures, family, etc) that you can then take back to your client/boss and validate whether they line up with business goals.
You can easily create these views for Paid Search keywords or Organic and find even more valuable insights.
Tag clouds are great at understanding the big strategic picture and understanding the sophistication, or lack there of, for any brand. Compare the LDS church with. . . something completely different. . . Chase Bank:
For Chase, walk through the analysis I was doing above for the Church. What do you think?
In Wordle even after I remove the big words (representing massive traffic) from the cloud (say the words Chase and Bank) the story does not get much better.
How about this one:
See what I mean? Simple data presentation technique, some pretty big insights.
Tag clouds have limits. You don't know what the problem is. Is it people? Is it a lack of sophistication? Is it using too much cruise control? Is it bad SEO? Is it. . . you'll have to dig. But you have 1. A great understanding of the site's search data and 2. Something of incredible value to present to your boss / client.
What to do next:
I am a big fan of internal site search analysis. Few other sources contain as much direct customer intent as this. Visitors to your site are directly telling you what they are looking for. The challenge, as always, is gathering up all that intent into something understandable.
How about downloading all that and creating a quickie tag cloud?
I can look over my pages viewed and time spent and Google'd keywords and all that. Or I could analyze the story above and at a glance understand what people are seeking.
Then, based on time available, we could analyze where people start searching, how many of them bounce off the internal search results page, what is the conversion rate / goal values for at least the top x of the above searches, etc etc.
I could even mine data above to see what other topics I could write more about (or in your case. . . what new products you could sell / stock / invent!).
Step #7: Are they making money or making noise? Goals & Goal Values.
With a tiny detour into search (always one of the biggest components of most people's acquisition strategies) we will go back to my first love: Outcomes! Ok, ok, ok it's customers, but outcomes are close.
I can tell the sophistication of any business (and the HiPPOs) by what I see in this report, Goal Conversions & Goal Values:
What to look for:
The first thing to check is if you see anything here.
If you don't see anything here, and the company has been around for some time, then you know you are going to struggle, in case this is a consulting gig, to make any decent money off them or, in case this is your first day in your job, you are going to not get a lot of love in this company as an Analyst. I am not saying quit, I am just saying dig in 'cos it is going to be a soul-searing struggle if this report is empty.
On the other hand if macro and micro conversions are present then get down on your knees and say a silent prayer because this is going to be fun.
Check if the actual goals & conversions are what you had noted in Step 1. If they are not then what are the visitors to the site doing of business value? Anything you noted in Step 1 that is not here (new goals to create?). What do the trends over the last 12 to 16 weeks suggest? What Goals are contributing the most amount of value?
At the end of this little exercise you should be able to confidently speak to your client / boss about how the website is meeting business objectives, and possibly where it is failing. If you did Steps 2 through 6 well then you might even have other actionable recommendations to make.
What to do next:
If there were some goals you had identified, or your client/boss was expecting, then take a moment to configure those in the web analytics tool.
If they do not have any behavioral goals created (99% of the people don't), then create those, takes just a moment. Refer to your insights from Step 3 to set the threshold values.
If this was an ecommerce website I would typically create one segment as a little bonus for the client. Orders where the total value was 50% higher than the average order value. Essentially the "whales" – people who order way more than normal. My hope is to get particularly valuable insights about where these people come from (geographies, campaigns, keywords, etc), what they do on the site (content consumed etc), and what they buy (shopping cart / basket analysis etc).
You want a lot more of these people. It is good to understand them really well.
Step #8: Can the Marketing Budget be optimized? Campaign Conversions/Outcomes.
Remember the only three outcomes that are important in web analytics? More Revenue. Reduced Cost. Increased Customer Satisfaction.
In this step I focus on the second item, reducing cost. Perhaps it is surprising as this is our very first foray into the web analytics data, we have received limited love from the client / HiPPO, and we don't have all the company business specific knowledge that might be necessary. Yet we can help reduce cost of marketing / customer acquisition.
My favorite report? Goals / Conversions by Campaign:
What to look for:
Campaigns as in Paid Search and Display and Email and Social Media and really anything of value you'd discovered in Step 2.
Start by looking at the horribly named "Other" report in Google Analytics (or perhaps the appropriately named Campaigns report in your web analytics tool). Initially allow yourself to be guided by that column at the end Per Visit Goal Value. It is a measure of efficiency. Notice above you go from $1.02 to $63, it is not hard to guess one campaign is working better than the other.
Then work backwards and see what conversion numbers look like. Then work further back and see which individual goals might be causing that high value to be created.
At the end of this exercise you should have some preliminary recommendations for at least one or two places money is potentially being wasted, or at least inefficiently spent. Killing opportunities (example: More better email campaign, less crappy Facebook pages!). You should also have some sense for where improvement opportunities might exist (I had a bunch above where PVGV was 40 cents).
What to do next:
Pick one or two major campaign strategies the company is executing and dive deep into ecommerce analysis (if the client is ecommerce). The two screenshots of Paid Search and Yahoo! Display campaigns gives you just a small hint about how much opportunity exists below the surface to dig and understand the deltas between conversion rates and the average order values and the trends for those key metrics.
Far too often in the web analysis world our obsession is with analyzing behavior (visits and time on site, etc.) or with focusing on how to get more visits (spend more money!). If you want serious attention (and love) from your client / boss then you'll focus initially on cost reduction. You've seen that obsession of mine in almost every step here. I have learned this lesson from a lot of painful personal experience. I encourage you to embrace it as well.
Step #9: Are we helping the already convinced buyers? Funnel Visualization.
Ending on cost reduction was a good point. In this step we are going to do one awesomely sweet thing: focus on perhaps the fastest way to increase outcomes for the business!
The poor funnel report is so underappreciated. While unstructured path analysis is the biggest waste of time you could engage in, structured path analysis is literally manna from heaven.
You want people to go through a series of steps (one after another) to meet a goal (for them and you). Credit card applications and ecommerce orders and donations to non-profits and leads to potential email spammers (I kid, I kid, I kid the spammers!).
The funnel report shows where in your three or four stepped process people leave.
What to look for:
The red bars. The bigger red bars.
I wish I could write lots and lots about this, but that is all there is to it (if the funnel was correctly created).
Look for where the highest exits exit in the funnel. Go look at the page with your eyes, get your mom and your BFF to look at it as well, identify improvements (heuristic is ok), submit them to your client / boss for fixes. The utterly lame ones you can just kill, the moderately lame to "don't know if this might be an issue" things go into an A/B testing bucket.
Either way, here is the analogy. Someone walked into your supermarket. They filled their cart full of stuff. They line up at a cashier and take out their wallet. They notice the loooong line. They move the cart aside and leave. You don't want them to leave! It was so hard to get them to come and add to cart and line up! Fix anything that stands in the way of the open wallet and you.
What to do next:
In many cases the above funnel process happens over multiple visits (sessions). In that case your normal Google Analytics (or Adobe Site Catalyst or WebTrends or NetInsight) funnel won't work. Well, it will "work" but show imprecise data.
Switch to something like PadiTrack. It measures pan-session funnel conversion performance. The same Visitor can enter and exit and finally convert across sessions and you'll be able to see that behavior.
Another compelling thing about PadiTrack is that you can view, praise the lord, segmented funnels! Search and Display and Email Visitors convert via different behavior, and finally you'll be able to see this.
PadiTrack is free, works using the free Google Analytics API, and works on historical data! Most web analytics tools, including paid tools, can't do that!
Step #10: What are the unknown unknowns I am blind to? Analytics Intelligence.
Without any help from my boss or marketer or mom I have got through nine steps of web data analysis and found a few concrete and meaningful bits of actionable insights.
But the danger of doing this with no tribal knowledge, or intelligent party at the other end, is that I might miss something I simply don't know because I don't know it.
The unknown unknowns!
So before I close any analysis for a website I go look at the Google Analytics Intelligence reports. There I can count on the fact that the unique intelligent algorithm in GA has done forecasting and applied control limits and statistical significance and much more math to help identify anomalies in the data. I see its soothing embrace:
What to look for:
Initially I set the Alert Sensitivity to Low (multiple standard deviations away from the mean) and see what automatic alerts show up. These are big events, so most important. Then I move the slider slowly towards the right see what other alerts pop up (see the Traffic Source part in the above image).
I am looking for events and activity, on the site or caused by others externally, that I (and usually even the client / boss) would not be aware of. The unknown unknowns.
Your discoveries here are great way for you to check your own work in the above steps (perhaps some of what you thought did not make sense does make sense now). They are also a great way to impress your client / boss that you somehow, let's just say magically, discovered things that even they, the most data driven of data driven companies, might be unaware of in their own data.
What to do next:
The hard part with Intelligence (custom or automatic alerts) is to isolate the root cause. Look at the newly released GA Intelligence Major Contributors section. That has the clues about root cause. Leverage the advanced segmentation feature to isolate the activity causing source / behavior / outcome and dig deeper.
[For more checkout the videos on Google Analytics Intelligence.]
And you are done! Does that not feel awesome? And more importantly, doable?
Summary: The Beginner's Guide / Tips / Best Practices For Web Data Analysis
In case you needed a handy checklist, here's what we've learned today:
Step #1: Visit the website. Note objectives, customer experience, suckiness.
Step #2: How good is the acquisition strategy? Traffic Sources Report.
Step #3: How strongly do Visitors orbit the website? Visitor Loyalty & Recency.
Step #4: What can I find that is broken and quickly fixable? Top Landing Pages.
Step #5: What content makes us most money? $Index Value Metric.
Step #6: How Sophisticated Is Their Search Strategy? Keyword Tag Clouds.
Step #7: Are they making money or making noise? Goals & Goal Values.
Step #8: Can the Marketing Budget be optimized? Campaign Conversions/Outcomes.
Step #9: Are we helping the already convinced buyers? Funnel Visualization.
Step #10: What are the unknown unknowns I am blind to? Analytics Intelligence.
The first time you go through the steps outlined in this guide it might take you more than 120 minutes. But I promise you that with time and experience you'll get better.
I wish just reading this blog post (it probably took you 120 minutes just to read it!) would be enough. It is not. You'll have to go practice it on many many clients. The more you do it the better you'll get as your sense of direction, data, discovery and deduction get better and better and better.
It is optimal to start any web analysis with a clearly defined web analytics measurement model. But if you don't have one then you no longer have an excuse not to provide something small that is incredible and of value from any web analytics tool you have access to, for any website in the world. And know that I am rooting for you!
Ok, your turn now.
When you are thrown into a website's data blind what are the first few things you do? What reports and metrics do you attack first? Over time have your discovered any strategies that work across multiple clients? Do you agree with the order of the steps above? Would you do something differently?
Please share your thoughts / critique / best practices / tips via comments.
Thanks.
November 15th, 2010 at 02:47
What can I say to such a post ?
Great! Instructive and gives a real insight for web analytics beginners !
November 15th, 2010 at 04:37
Nice, covers the whole analytical chain for web marketers.
Great post,
Niklas
November 15th, 2010 at 05:17
When I perform analysis on client website, I try to go to the website first and get a fill for what they are trying to convey to their audience.
That helps me understand what they are doing and how I can help them even more.
TrafficColeman "Signing Off"
November 15th, 2010 at 06:49
Thanks for bringing up the Wordle example again. I have yet to use that tool. I will now though!
I do as you suggest in this post and get myself oriented with the clients website. One mini step I'd like to add is doing a bounce rate against site average during step #2. Or, add that to step #4. In some cases, I'll find that google / cpc is performing really poorly and I'll be able to dig deeper by individual campaigns / groups / ads that may be mistakenly triggered by bad "broad keywords".
Insights aside, is it our responsibility as the analytics dude to make sure their Analytics software is installed properly? And if so, how far do we go with that analysis? For example, do we recommend Event tracking / Custom Variables at this stage of the relationship?
November 15th, 2010 at 07:28
Brian: It is likely that some of the cooler features (even tracking, custom variables etc) are not implemented, and also likely that at least some of the analytics code is not complete (some pages might be missing tags).
As Web Analysts I have noticed that we consistently tend to make our first engagement with the client to be about impressing them with our technical competence and hence tags and code and other such things. I wanted to make it about finding insights, even if they are not supremely awesome or even 5% incomplete because of implementation issues.
If we can show we are not just techno geeks and there to deliver solid business value, even after the first two hours, I think the rest can follow. They'll be happier paying us our going rate and happy to put the time and energy required to fix things and implementing advanced features.
It will be our responsibility to ensure the technical implementation is complete and accurate, but after:
1. We have delivered some business value and
2. Impressed them enough with the first part that they'll work with us to create a web analytics measurement model which will guide advanced features roll out (so that we are only doing things the client will value and not what we feel are important) and then finally
3. Hello code here I come!
-Avinash.
November 15th, 2010 at 07:36
This is the best post from you so far.
Great summary to love & success. Thx!
November 15th, 2010 at 10:44
Great post! I am so glad that you included Step #6: Keyword Tag Clouds in this. It is something that I have never really thought of doing. I can really see how doing this could bring some great insights and be a good way to demonstrate to the higher-ups what is really happening (or not happening).
Thanks!
November 15th, 2010 at 11:03
Forget beginner's – I think that everyone in the WA industry should read this post. It's just a solid outline for success that even the most advanced / senior WA people in the industry should read to shake the rust and dust off – and receive a clean, fresh perspective.
And – how can one not like a post that includes Weighted Sort in Google Analytics – easily one of the most unheralded new features out there?
Have you ever thought about becoming an author? :)
November 15th, 2010 at 11:06
Hi Avinash,
It is always great to read your post and explore more about Web Analytics. You have covered almost everything within this post related to Web Data Analysis.
Quick Question : In Visitor section, I think we need to look out for Member and Non-member visits (Authenticated Users). This data can provide us how many authenticated visitors we have and what they do after login in the website.
What are your thoughts on it?
November 15th, 2010 at 14:03
I have one problem with your blog articles Avinash, for example this one took me about 2 hours to read it (as im not english native), catch "big picture of article" and write notes in Evernote for later use. After this my head is going to misfire, its painful, pffff! :) But from regular sport activities i know this kind of "pain" is welcomed, so thank you very much! :-)
November 15th, 2010 at 16:03
Thank you for all this information, this will help me to improve my knowledge and make real my dreams.
Greetings from Perú.
November 15th, 2010 at 17:31
Excellent process!. The initial web assessment that I have been doing for many years I call the First Impression. Using the approach of a first time visitor to a website one can quickly identify the objectives of the website, the targeted users, establish credibility, brand consistency, calls to action, create emotion – e.g. is the site including you vs pushing how good they are, depth and breadth of content among other factors.
It is surprising how different some website owners perceives things and they even try to argue that they are right and the visitor just doesn't get it – which is why I love the use of analytics to prove the point.
With a lot of my clients I usually avoid discussing the organizations objectives until I did my initial assessment then the discussion becomes interesting when one does a gap analysis between what a visitors sees and what the organization expects them to see.
As a process person I like this 10 step approach as a guide, each client is different, but many of the principles are similar.
Thanks.
November 15th, 2010 at 17:37
G'day Avinash,
I was a little surprised you lumped email in with campaigns in Step 2, and that traffic from total campaigns was so low at 10%.
Personally I treat email as a separate traffic source. Usually it accounts for at least a third if not more of a site's traffic and revenue.
Like you the Traffic Sources report has been my goto report for many years.
What I've found from doing long term yearly analysis is that email, search and direct traffic all even out over the course of a year and bring in similar amounts of revenue.
So if I saw a potential client's report that showed only 10% from campaigns including email I'd know that their email marketing was either non existent or totally sucked.
Of course that would be an immediate low hanging fruit area where we could make significant improvement.
November 15th, 2010 at 21:31
Hi Avinash,
I think this is a great start for someone who wants to get their feet wet in the analytics world. Although one thing I would like to add is that without understanding the business model it would be harder to understand the data. Your suggestion to get to know the website is really invaluable.
One thing that I find it disturbing is that the folks using the tools keep piling on to it without real understanding of the value of the reports that they create. I would definitely recommend them to follow some of the steps outlined above.
November 15th, 2010 at 23:31
A great post — Avinash you have an admirable way of cutting through all the jargon and showing in clear terms what is really important and why.
I've read whole treatises about metrics and KPIs while a week later I realized I've retained only a fraction of the information. This, on the other hand, is a repeatable process that a novice could use to get a real start analyzing a site. And they could remember it better each time — because the connection to the real world is clear.
So I appreciate the expertise but also the learning model being used.
Thank you!
November 16th, 2010 at 00:17
Hello Avinash,
A brilliant post and a nice recap of many of the things you have talked about in the past.
To me the fudamental value-gem from this post is the underlying message – "Everybody only hears what he understands" (Goethe). So start delivering value from the get-go (guided by the 5 steps from your Yahoo post and this post) and then slowly build towards a robust measurement model and analytical complexity.
Also, success in web analytics is not about a perfectly tagged site or if you have all the support. At the end of the day, it is about how enterprising the analyst is and how creative he/she can get to identify the rose among thorns.
Enjoyed the read.
Regards,
Ned
November 16th, 2010 at 06:46
Avinash,
I've been planning on hosting a free web analytics clinic in my area but have been struggling with the curriculum. How do you start to explain the power of a tool like GA in only two hours, especially to a room where 1/3 haven't touched web analytics?!
This blog post has given me some pretty outstanding ideas on how to structure my curriculum.
And by the way, I'm sure the percentage of your same day visits is quite high if your other readers are like me. I find myself coming back several times throughout the day when you have new posts, squeezing in paragraph here and there as I work on tasks throughout the day. :)
November 16th, 2010 at 08:32
Great post, Avinash. There is a lot of excellent advice here, and I agree with Joe that this is hardly just advice for beginners.
I particularly like that you started with the suggestion to actually use the site (and competitor sites, too). I find it's critically important to try to understand the customer experience before diving into behavioral metrics. To that end, (and this will sound biased from me now, but I swear I did this even before I worked for ForeSee) I always like to start with customer experience/perception analytics and general attitudinal feedback directly from customers. I wrote a post a while back called "Is elistism the source of poor usability?" (http://www.retailshakennotstirred.com/retail-shaken-not-stirred/2009/07/seeing-with-someone-elses-eyes.html) that addresses the processes I like to use to get to the heart of customer perceptions about their experiences with a particular site. Combining analytics that come directly from customer feedback with analytics that reflect customer behavior on the site can really help us get a much better understanding of exactly where we should focus our efforts.
November 16th, 2010 at 17:48
For ecommerce websites I would personally start with funnel visualization then with the rest.
The lowest hanging fruits are the things on your site, not externalities like campaigns, keywords, email… excellent article, as usual :)
November 16th, 2010 at 18:01
Prashant: My overall focus for this post was the first few hours of "blind analysis". In that time period it might be prudent to leave the Member and Non-Member visits for later.
But you are absolutely right in the value of that analysis. For example for my internet marketing certification start up http://www.marketmotive.com this analysis is the bedrock of all analysis. And the wonderful thing is that with release of event tracking and custom variables it is ever more easy for us to track this (in fact with the latter we can even track the Member in subsequent visits even if they did not log in, how amazing).
Hope this helps.
Eelke: I am glad that you found the blog post to be of value. I empathize with the effort required. I am afraid there are no short-cuts to glory!
But I do hope that the blog post will help channel the effort more efficient.
Mark: For different businesses the actual balance in the portfolio might be different. In your case, clearly, Email is a much bigger part of the portfolio. Understanding the business context would help decide what is too much or too little.
Peter: I try to get the business objectives from the business leaders first but I like your idea of not talking to them until you have first made your observations. I am going to use that strategy more in the future. :)
Kevin: No disclaimer required! I know that your feedback is unbiased and always welcome here.
In this post the situation was "day one going in" and it is the sad fact of life that most of the time clients / companies rarely have voice of customer data available. Like you I passionately advocate jumping on the feedback bandwagon as soon as the client / boss / opportunity will allow.
The amazing thing is that now there are so many free and affordable options to get started very quickly, be it surveys or crowd-sourced feedback or input from experienced designers, usability experts and marketers. [Specific tools in the VOC section here: http://zqi.me/wamodel ]
-Avinash.
November 17th, 2010 at 10:56
This is a great post Avinash and is going to be really useful to give me a framework to work to when pitching for ecommerce business. We always ask to get into their analytics to have a look beforehand and now our thoughts will be all the more comprehensive. Thanks!
November 18th, 2010 at 01:56
As a newcomer to web analytics, this post has been really helpful in navigating through all the reports available.
I particularly like the way you use wordle, I would never have thought of that.
Thanks.
November 18th, 2010 at 04:33
As always great article!
Working in an agency I often encounter this kind of project start "have a look at the numbers, maybe you find something". And this is how to do it *right* :)
Many website owners (still) don't have a clue what their targets are. Web Analysts have to guide them through the process of setting goals.
November 18th, 2010 at 08:15
Outstanding post, Avinash! Absolutely stunning.
I think it's a very special post, because somehow you always manage to break it down to the most important questions and issues.
Like your writing a lot. Direct approach, crystal clear and always spot on.
This post was quite helpful for me. One can get lost in the flood of information and in the number of reports provided by a tool very easily. So stressing the point of following a certain path is an excellent point.
I agree with most of the other comments here: It's certainly not just for beginners in webanalytics. – I actually think it's not for beginners at all. :)
November 18th, 2010 at 09:36
I just used the wordle search phrase mapping with a client to great success – since the company name is the clients name – seeing their name in large bold figures made a strong impression. I would like to think my analytical skills contributed to the contract extension but … Thanks for the many interesting snippets of advice, practical and easy to integrate into one's bag of tools.
November 21st, 2010 at 09:58
Wow very detailed I am going to send this to my webmaster.
Thanks!
November 22nd, 2010 at 07:04
Thanks for your great post again. Can I use it for input on my dutch online marketing weblog? Of course, I'll refer to you.
November 22nd, 2010 at 07:27
Dude – Top post. I usually feel more like Fire-fighter than a Ninja so this post is excellent, quick ways to useful data.
It would be better to have a 'web analytics measurement model' but sometime that's out of our control… Post like this can give some control back.
Cheers,
November 23rd, 2010 at 08:41
Great post, as usual. I've come back to read this 3 or 4 times now since you've posted it. Your top landing pages report in particular is very digestible and actionable.
Much appreciated.
November 23rd, 2010 at 13:32
I love the fresh look on this topic. Very in-depth and I'll have to go through it over the next few days to get the most out of it. Very good resource for bookmarking.
November 23rd, 2010 at 14:19
I second Markus' sentiments (comment #23 in thread). Requests for blind analyses like these are extremely common. And I've found that launching a strategic discussion with site owners is often MUCH more productive if you bring something to the table to discuss.
For example, I often face blank stares when I simply ask, "What's the purpose of your site?" If, instead, I come armed with my best shot at an initial analysis, blind as it may be, people are very quick to jump in and say, "No no, this isn't what's important to me!.. the purpose of this site is really to…"
So, my key piece of advice is: Don't be discouraged by the inevitable "negative" feedback you'll receive on this kind of blind analysis! That's EXACTLY the kind of information you need to produce meaningful insights in the long run.
When you present this kind of report, just call it what it is: an objective site review from an uneducated user–basically, the perspective of one of their site's actual users! Explain that the purpose of this initial analysis is to generate discussion, and humbly and eagerly accept the resulting feedback.
As a result, you'll probably get more useful tribal knowledge in one meeting than you'd get in 10 generic "stakeholder 'Discovery' interviews"!
November 23rd, 2010 at 22:05
Some really great points here.
I particularly appreciate point 4. I am going to spend some more time looking at bounces on top landing pages.
November 24th, 2010 at 06:27
Emily #31 has articulated my experience with many customers, and my solution is a "First Impression" analysis taken from a new visitors perspective which I bring to the first real meeting. See #12. It is even better if one can look at the stats also before the first meeting. Thanks Emily for your view on this tricky issue of convincing a prospective customer to commit.
November 24th, 2010 at 08:57
Emily (/Peter): I have been guilty of advocating a deeper and more persistent engagement with clients/bosses (though not: "10 one hour stakeholder "discovery" interviews"!!). We do far too little of tying our work to business value in our web analytics world.
But I have never approached anyone with a six week discovery process. I have always gone with some data. Even when I don't have access to their data I'll take some along from Compete or Insights for Search or Trends for Websites. As you point out it makes for a much better conversation.
Thank you for adding your valuable perspective.
Avinash.
November 24th, 2010 at 09:32
As always, excellent post. Avinash you rock! This post is awesome. A perfect guide for anyone taking a look to define the initial web strategy.
It seems more dedicated towards web analytic generic user rather for beginner. But for those who are familiar with web analytic and this could be used as a first step towards building things for their target audience.
Nevertheless I loved it thumbs up!
November 24th, 2010 at 12:21
Thanks for the post. Looking at lots of numbers and trying to make sense of them can make me a little dizzy sometimes, and guidance as to what I'm looking for is extremely helpful.
One question – you mentioned setting up a segment that will enable you to see people who order 50% higher than averae order value. Where do you do that? I looked in advanced segments and goals but couldn't seem to find anything of the sort.
Thanks!
November 24th, 2010 at 12:24
Great article, Avinash! Just what I was looking for.
Margo
November 24th, 2010 at 13:41
Aviva: I am sorry I should have been more clear on that recommendation.
In Google Analytics you can set custom alerts for Average Order Value (so you'll get alerted if people start buying lots of stuff) but you can't segment using Average Order Value in advanced segmentation. Super annoying.
What I have done, if I am limited to using GA, is to create a segment using Quantity and set the condition to be "greater than or equal to" and value to something greater than the average quantity per order. It is imperfect, but is a close enough proxy.
Of course if I really want to do this I can also use one of the business intelligence tools in the Google Analytics application gallery.
But the context of this post is not to go that deep into the engagement so might not be the best recommendation in this context. For normal engagements it might work.
Avinash.
November 25th, 2010 at 01:59
Once again, a post worth bookmarking!
However, I'd tend to disagree on the traffic sources percentages. The proportion depends on many factors: age of the site, effectiveness of different traffic sources, marketing strategy etc.
For eg, except organic search (and may be referring traffic) traffic from every other traffic source can altered based on our marketing budget allocation. If I get 1000 visits from an affiliate, out of which only 1 converts – I don't mind it as long as I pay only for the conversion.
Lets say that I pay $25 for the conversion. And if SEM gets me traffic at $1 per click, and 1 conversion per 50 clicks: Thats $50 per conversion. I'd prefer to put most of my budget on affiliates.
So, obviously the traffic is likely to be skewed, and I don't see any harm in that. Am I missing something?
November 25th, 2010 at 17:56
Linus: My recommendations on "best practice benchmarks" (note the quotes :)) are not a reflection on what you and I can convert better or worse for segments of our traffic. I was emphasizing the value of having a robust portfolio strategy when it comes to acquisition.
For example what if some search engine tomorrow changes something overnight in how ranks your site? Your over-reliance on SEO could be deadly. What if your competitors, insanely, decide to start paying a lot more than you for ad position? What about others you compete with in the affiliate business?
Absolutely positively maximize the traffic and conversions from the source that works best for you. But also ask yourself what else you could do for your business / your client to diversify and attract new customers from new sources (not least increasing the portion of direct traffic that is existing customers who buy more / engage more frequently!).
-Avinash.
November 27th, 2010 at 19:14
Hi Avinash,
Greetings from a Brazilian guy living in Thailand. :-). Wonderful post!
I believe that is a good practice to keep track on GA configurations, like a change log, using a spreadsheet. Am I correct? Do you have any example to share? Please advice me.
Thanks!
November 27th, 2010 at 23:42
Daniel: For any mature web analytics practice, in any company, the process of keeping tracking of Google Analytics / Omniture / WebTrends configuration updates is extremely important. As is using the annotation feature in tools like Google Analytics and Yahoo! Web Analytics to keep track of important business / marketing changes so that everyone can see them as they use the tools / data.
But in context of this discussion (approaching someone's website for the first time to provide some valuable insights) it won't be within scope of the work.
With regards to examples… most that I have seen are simple excel spreadsheet with columns for date, change, reason for making the change, name of the person who made the change and name of the person who authorized it. Fairly generic stuff, but often a common sense approach works best! :)
Avinash.
November 28th, 2010 at 06:26
Thanks a lot.
Well explained.
Cheers,
Daniel
November 28th, 2010 at 22:32
You have provided a complete guide if one thinks about it, is beneficial for both the users and the site owners, because user can enhance the way of browsing and experience,and owners can formulate strategies about how to go along with the user preferences…
December 2nd, 2010 at 14:30
Amazing post.
Will always try follow the steps you mentioned.
Thankx.
December 2nd, 2010 at 15:42
Avinash,
Nice post, especially the ideal breakdown of Search/Direct/Referral. I have never had numbers, but I have expressed to clients to think about a balanced approach to generating their traffic. Do you know if there has been any work on a benchmark based on type of site?
Thanks for a wonderful post. With your permission and reference to your blog, this may be an inspiration for posts I am working on. Thank you again!
December 13th, 2010 at 08:54
I'm not quite understanding how to export the CSV for keywords for wordle.net word clouds. My biggest words end up being 'Visitors, Total, Returning and Visits' after exporting a CSV from Google Analytics and pasting it in. What am I doing wrong? Do I need to delete rows? Help.
December 13th, 2010 at 09:26
Terri: Here are detailed instructions on how to download data from Google Analytics:
~ Back to Basics: Tip for exporting data
It also shows how to download more than 500 rows of data at one time.
After you download the data… open the file in excel, just copy all the search keywords from that file, paste them into a open text file (using a program like notepad), save this, now copy all the rows there (just hit Ctrl + A), paste that into http://www.wordle.net/create , hit the button Go and you should see the tag cloud you were expecting.
-Avinash.
December 14th, 2010 at 16:38
Fantastic post, Avinash. This might very well be your most actionable post in my eyes.
For our turn, you asked about our experience when first tackling a site so here's mine – and it also encompasses one of my biggest struggles.
Rarely do I jump into a web analytics account that is properly configured enough to do this much up front analysis. The most common scenarios I experience for small to medium companies are:
1) They just copied/pasted the code onto every page then walked away
2) they use some terrible, clunky, counter intuitive WA system that makes you want to give up being a web analyst and go start a farm raising sheep or
3) they don't use anything at all. So here we are – hired to tell them why their site stinks but first we need to install, fix, de-suck their web analytics just to start what we were originally hired to do.
All your points are still incredibly valid, but I have yet to dive right into a beautifully functioning and configured web analytics program that lets me immediately break out my ninja mask and sword.
Still, once we get past the point of de-sucking the poorly installed/configured web analytics we can jump right into your points, sword in hand!
December 20th, 2010 at 02:43
Hi Avinash,
This is my first comment on your super blog. Sorry i got late for the party.
The kind of analytical insight i get from your posts in unparallel. And i would really appreciate if you write shorter posts and post more frequently. Your posts are so big that i can consume them only on weekends. Shorter easy to consume posts are more viral and result in much more organic traffic (as you target many more keywords via title tag, dilute keywords cannibalization and frequently produce fresh contents) and also because people don't read contents on the web they skim them. This is tried and tested stuff so you can count on me. Otherwise you can always test yourself :)
Regards
Himanshu
December 30th, 2010 at 01:27
A quick question:
Which technical skills a good analyst should have in your option?
Do I need to be a JavaScript programmer?
Thanks,
December 30th, 2010 at 03:24
Daniel,
None,
if you're focusing on technical skills you're focusing on the wrong area.
A good analyst doesn't need technical skills. What they need is an ability to understand and interpret the data, and how that relates to the business objectives.
December 30th, 2010 at 07:20
Daniel,
In my opinion, it really doesn't hurt to have some technical know how, at least related to Google Analytics and how it works. I agree with Mark that you don't need to be a jscript programmer, but I think it's important for anyone looking at the data to understand how the data got there.
In my experience, there are many many scenarios that can cause the GA data to get distorted (e.g. improper configuration on site, wonky filters, careless web developer that did something to the GA code, etc). Knowing what these are and recognizing them through the reported data can help you spot issues before you start making decisions. For example, I was looking at a site where the bounce rate dropped to single digits. I knew this didn't seem right and with some quick investigating saw that the GA code was copied twice onto the same page – making one pageview two pageviews in the eye on GA, thus plummeting our bounce. Had I not understood how the code works I may not have spotted that. Instead, I may have thought I had a super campaign that delivered the most relevant traffic EVER!
For the record, I have little to no programming experience, but I find learning more and more about the GA code makes me a more rounded analyst and user of GA.
December 30th, 2010 at 08:45
Thanks guys for your reply.
I know a bit of programming, enough to understand mostly of time what is going on. But after read Google Analytics by Justin Cutroni (I already read yours Avinash :-) ) and saw his JavaScript code, I felt that I have a long way to go.
Anyway, I agree with @Chris that technical skills are a very helpful add-on to your analyst career.
Cheers!
December 31st, 2010 at 17:11
Daniel: I'll add just a smidgen to Mark and Chris's valuable comments.
In many small companies you or I have to be jack of all trades and in as much acquiring technical skills can be super valuable in adding complete value. It is important to realize though that an Analyst (if indeed that word is in your title) will mostly be called upon to do analysis of the data and less display proficiency javascript / html5 / flash skills.
That is not to say being a Implementation Specialist (not an Analyst) can't be a fruitful career choice.
I encourage you to look through my descriptions of the four job roles in Web Analytics and make the best choice that fits your skills.
~ Analytics Career Advice: Job Titles, Salaries, Technical & Business Roles
The post includes descriptions, career prospects, salary prospects & long term growth of each job.
All the best.
Avinash.
December 31st, 2010 at 18:53
Great article! love it.
I need directions on step 4 and how to access that? It's 'top landing pages' –> then what do I click to see bounce rates indexed against site average?
January 1st, 2011 at 08:39
Jeremy: On every single GA report with a table you'll notice on top right of the table (to the very right of the tabs) there are a set of five icons next to the word Views.
Click on the fourth button that shows bars on two sides of a line, it is called Comparison if you hover your mouse on it.
You'll see the indexed view.
You'll see a drop down for the metric on top of the bars. Click on the drop down and choose Bounce Rate (or any other metric) and you are in business.
Good luck!
Avinash.
January 1st, 2011 at 12:59
Thanking you Sir! that's excellent :-)
January 8th, 2011 at 13:08
I agree it is very good article. However, I would like to point that some examples shold not be taken quite literaly, even when the conclusion is right. E.g:
True. However, what is balance portfolio depends on your business and your goals. If you have e-shop with HD player, you can have great success with minimal direct traffic. If there are great reviews on referal sites and you rank well on all important key phrases, you are in great shape.
January 8th, 2011 at 21:14
Mirko: There are always exceptions to the rule. But most businesses need to have a portfolio strategy if you want to win in the long run.
Even the scenario you outline in your comment you have 1. minimal direct traffic 2. great reviews (and links) on referral sites and 3. rank well on search engines.
That's a portfolio strategy. :)
Avinash.
January 9th, 2011 at 00:11
Avinash, I agree. Just wanted to add my 2 cents, that "balanced portfolios" are different for different business. So balanced portfolio should not be thought as "at least 20% of each traffic source".
In other word, before starting investigation it is necessary to know what is a good strategy for given business and judge the traffic against it. When you find it differs significantly, you have a starting point, but it does not automatically mean that there was no good reason for the actual strategy applied.
*
Not sure it is related, but an interesting event on the czech market: Google started Chrome promotion on huge banners near roads at least in Prague. My guess it is attempt to beat local search engine (Seznam) that dominates Czech Market.
January 24th, 2011 at 01:53
Dear Avinash,
I have difficulty in calculated goal value. Ideally, based on what goal value measured?
Thanks a lot,
Mega
January 24th, 2011 at 10:20
Mega: I am so glad that you asked as computing Goal Values / Economic values is the subject of this detailed post:
~ Excellent Analytics Tips #19: Identify Website Goal Values & Win!
Please checkout the five specific strategies with examples from B2B, B2C, b2p and b2anything companies!
Avinash.
February 3rd, 2011 at 07:55
Avinash, this blogpost is so good (as is OR overall) that I'm adapting it for financial advisors ( http://www.triplestopllc.com/2011/01/06/advisor-site-analytics-1/ ), breaking it into 10 posts (steps) and making it more industry-specific.
Of course, I'm crediting you all the way, and if we succeed you will get more than a few additional readers–oops, micro-conversions–with time. These advisors are an analytical bunch.
Thank you for the fantastic content you're putting out there.
February 10th, 2011 at 08:33
This is a great introduction to Webanalytics and websites analysis… I shared it with my collegues (most of them don't know webanalytics).
Thanks Avinash !
May 21st, 2011 at 06:25
Fantastic post, really useful guide to making an efficient analytics process.
Thanks!