The world of web analytics is blessed with many interesting, intelligent, innovative people. But if I had to pick just one person to spend an hour with from that entire community it would be Dr. Stephen Turner.
Dr. Turner is the author of one of the earliest “web analytics” application Analog, which is still the most popular logfile analyzer in the world. He is also the Chief Technology Officer at ClickTracks. His is an academic by background (his research was on Stochastic Networks, just writing that makes me sound smart! : ), for his research publications click here.
I have had the pleasure of meeting and collaborating with Dr. Turner on many occasions. Every single one of those sessions has been insightful and I have come away with tremendous respect for Dr. Turner’s thinking and intellect (or as I like to call it “raw compute power” : ).
This is a nice long interview but I think if you read through the whole thing I think you’ll end up agreeing with my assessment.
1. How do you make yourself indispensable to a company?
Find your niche, whatever it is, and excel in it. And be the sort of person that people want to spend time with.
2. What are some examples of activities and surroundings that motivate you?
Anything that challenges me intellectually. I love to figure out a neat solution to a difficult problem, either on my own, or even better by arguing it out with smart friends and colleagues.
3. (From our regular reader Jen) What did you want to be when you were 10 years old?
I don't think I had a specific job in mind, but maths was always my favorite subject, so it would definitely have been something to do with maths.
4. What did you really love about your last job?
My last job was very different from what I'm doing now, because I was an academic, doing mathematical research. I loved the academic lifestyle — being able to work on whatever interests you without commercial pressures, the relaxed attitude, and spending time with a lot of exceptionally smart but fun people. But I think I'm a lot better at what I'm doing now, so I wouldn't go back.
5. What is ClickTracks’s biggest strategic strength, and its biggest challenge looking into the future?
I think our biggest strength is our design process. We never add a report just because the data is available and easy to churn out, or because other analytics companies do it. Instead we think hard about what the analyst really wants to find out, and design an innovative report that gets to the heart of the question.
Anyone who's struggled to get anything useful out of the usual Funnel Report, and then looked at our Funnel Report, will know what I mean. Most Funnel Reports are designed around the assumption that people move around your website in a specific order. Ours is designed starting from the assumption that they don't.
Our other strength is our relentless emphasis on segmentation. Every one of our reports is required to show statistics for any visitor segment the user wants. That's because we think that segmentation (looking at the same statistics for different groups of visitors) is the key to really understanding your website, and using data to improve it.
You can produce reams of paper every day listing all your data, but segmentation is what gets you inside the head of your visitors, and turns you into a true web analyst.
I should try and justify that. You see, segmentation is what enables you to ask and answer questions about your site. It's one thing to notice that your average revenue per customer is too low, for example, but what are you going to do with that information? That average revenue conceals a lot of different types of customers, some of whom may be doing well for you and some badly, so you need to follow up by asking questions like "which landing pages improve conversion and which put people off?", "which referrers sends quality leads and which cost us money without generating sales?", "do people prefer the red or the blue home page?" and so on.
All these questions are questions of segmentation. And when you've answered them, you'll come to a point where you can understand what your visitors were looking for and why they did or didn't buy, and at that point you can use the information to improve your website, which after all is the goal of the whole process.
You asked about challenges too. One thing which I think nobody understands well yet is how to do web analytics on sites built with AJAX and similar technologies, particularly sites where a visit is no longer as simple as a sequence of page views.
How would you do analytics for Google Maps for example — what would you report? And I think it's not just a matter for the analytics companies — the website designers will need to think about analytics at a far earlier stage of the design process, defining what is meant by an "event" on a site, and what they want their analytics program to report.
We may see more customized solutions, as it becomes less useful to report the same statistics for each site.
6. I would only ask this question to someone who knows what an ErdÃ¶s number is: What is the "coolest" piece of mathematics in ClickTracks?
That would have to be our What's Changed Report (which shows you which pages, referrers and search terms have risen dramatically this week, even if they're not yet in the top items) and our Click Fraud Report (which shows you which campaigns are performing badly or have suspicious activity).
They both work in the same way internally, alerting you to unusual patterns in the data that you would never notice if you just looked at the "top 10" type of report.
7. What is the one thing you feel Analog still does better than ClickTracks? :)
The data the system administrator is interested in, such as total bytes transferred, list of top redirects, and that sort of thing. ClickTracks has never tried to do that.
We made a decision right at the beginning to focus on the needs of sales and marketing and web designers — web analytics instead of web statistics, if you like — and we think we can serve them better by not cluttering up the program with data they're not interested in.
8. On my blog I talk about the challenge with the mass of metrics available in our tools, what is the analysis that you most wish ClickTracks’s customers did but something most don't?
I wish people used A/B splits, and I think most don't. Doing an A/B split on your website is a great way to find out whether a change you're thinking of making is actually a good idea or not. And you don't usually need complicated mathematics to interpret the results, because a lot of the time the winner will be obvious.
Besides, I've got an ulterior motive for recommending A/B splits, because we've got a really cute feature in ClickTracks that shows you the two versions of your site in side-by-side browsers with the link overlays drawn on each, and I'm concerned that not enough people see it!
9. There are a lot of metrics ("KPI's") that are great, lots that are sub optimal. From your experience which are the most hyped metrics that a analyst should stay away from, or atleast be very careful about? Why?
I think page views is a very popular metric, but not actually all that useful. I make an exception for sites that generate revenue on each page view by showing CPM banner ads. But for most sites, it depends too much on the structure of your site, and it counts all pages equally. You're far better concentrating on the number of visitors who reached certain important pages.
10. Independent of vendors and software what is the biggest hurdle to an awesomely fantastic web analytics success in a company?
I think the most important thing is what you've been talking about on your blog — the quality of the analysts. Some programs are better than others, but a good analyst can probably get usable data out of any program, and a bad analyst won't get usable data out of the best program in the world.
You need someone who is curious about data and really wants to get to the bottom of what's happening and why. And then the analyst also has to be a good communicator to explain the data to those who can act upon it, and management have to be willing to trust the analyst and act on those recommendations.
11. Why don't all the vendors get together in a consortium and decide for the sake of world peace, ok for our peace of mind, that they will all use the same ways to count sessions and visitors? Is there really such a magnificent advantage to each vendor inventing their own methods?
Well, even if you lay down all the rules in great detail, it's almost impossible to make two different programs agree completely. But let's suppose you could, or at least get very close. You'd still have the problem of what to do when someone comes up with a better rule.
Let's take an example. ClickTracks discards a session if the first request is for the file /robots.txt, because it's probably a robot we don't know about; but many other programs don't do that. Of course we can explain why we think our rule is a good idea, but we can't force every other vendor to change, and in the end we have to decide what we think is the right thing to do and do it.
And actually, this may be controversial, but I also don't think it's that important. It is important when switching vendors, but that can be solved by running the programs side-by-side for a while, which you're probably doing anyway. But otherwise, the exact numbers are far less important than most people think: it's much more important to look at the differences between different visitor segments, and the changes over time.
If you're still not convinced, think about the differences between logfile analysis and page tagging. Those differences are much greater than the differences between different vendors, and yet you can still get excellent actionable data from either technology.
12. What is the one thing about the future of web analytics that none of us know, but you do? :)
However much the web changes, people who can understand both business and data will still make the best analysts. Or maybe you did know that already.
My favorite answers were to questions 5 and 10 (not surprising to you I am sure). Did you have a favorite answer? Is there a question I should have asked but you think I did not? How can I improve future interviews? Please share your feedback via comments.
[Like this post? For more posts like this please click here.]