Understanding Analytics

Presented by Matt Bailey

Analytics is fun and awesome, and by the end of this presentation, you're going to love it too!

Standard stats dashboards are worthless. No analytics program is going to tell you "there's a problem in your shopping cart, and people can't go to the next step."

Where do analytics come from?

Log files: a big text file that records all the requests to the web server. Who requested a page, what page they requested, etc.

Page tagging: things like Omniture, Webtrends, Google Analytics -- a bit of JavaScript on your page that sends information to a third party server every time somebody requests that page.

Either way, they're just recording information about accesses to your pages and files.

Concern with page tagging is that the data is on a third-party server and isn't portable when you want to change providers. Matt favors log files because the data is his, and he can maintain historical data when he switches to a new log file analyzer program.

Log files are full of information about your visitors and what pages they requested. Another problem with page tagging is that it doesn't track search engine spider activity (they don't do JavaScript), which is another reason he prefers log files.

No analytics program will tell you how to improve your website. You have to ask questions of the data in order to get this information out.

Keep in mind: there is no such thing as "accuracy." This is why analytics programs show different numbers. It's a matter of interpretation, so programs don't agree about how to count visitors. Also, some analytics programs can't tell the difference between humans and scraper bots.

Unique visitors can only be counted in two ways:

  • Cookies -- a text file dropped on browser that identifies if you're a new or returning visitor. Confusion between first-party and third-party cookies. First party cookies (those from the site you're visiting) are OK. Third-party cookies (those from other websites, such as ad-tracking cookies) are evil, and blocked by browsers.
  • Client-side Javascript -- even some of these use cookies as a backup

Basically, without cookies, you're not tracking unique visitors, you're just tracking visitor sessions. Matt doesn't think people need to worry much about users deleting cookies, because most web users don't even know what cookies are.

Visitor sessions: a measure of the number of times visitors view your software (without regard to whether those visitor sessions come from new or returning visitors).

Additional definitions

Search Engine referrals: measures how many people came to you through search engines.

Bounce rate or exit rate: the number of people who come to your site, view one page, and leave. Matt loves to keep an eye on this one, because it lets him know if he's reaching the right people with the right message.

Conversion rate: the number of people who do what you want them to do. It's about taking some kind of action (not necessarily a sale). Average website conversion rate is 1.8%. Matt says that's horrible; we can do much better on the web.

We tend to look at our site as having a single function, an over-arching goal. Matt calls this a "macro action." It's important to look at the "micro-actions" -- all the little decision points people go through on our website on their way to the final goal.

Analytics is not about numbers, it's about the customer experience. What obstacles are visitors facing when going through the process on your website?

Recent survey of online advertisers: 72% of advertisers only track the click-through. Only 11% are doing a detailed ROI analysis.

We need to equate money to these micro-actions. For instance: what is the value of a lead? You need to know this to determine if your website is making a profit for you.

We need to get beyond reporting and move to analysis instead.

"Web analytics works best when measurement expectations are clearly defined in advance, not after the fact or on an ad-hoc basis." (Eric Peters quote)

If you haven't defined goals, you're not analyzing, you're just reporting.

As yourself:

  • What is the one thing you want people to do?
  • If they don't do that, then what?
  • Do you have data integrity? (Use one program for your reporting data)

What most analytics programs give you is data, not information. You have no context, so you can't draw conclusions or learn anything from it. Doesn't tell you how to improve your website.

Data + Context = Information

Information + (more) Context = Knowledge

You want information about a specific group of people, what they were looking for, and whether they found it.

Analytics stops at KNOWLEDGE. To get to UNDERSTANDING requires YOU. The secret is the analyst -- the person. Matt recommends the book Web Analytics: An Hour A Day by Avinash Kaushik. Avinash says if you're budgeting for analytics, budget 90% for the person and 10% for the software.

Start by building context. People aren't cattle. They don't move as a herd from one place to another. People come to your site for many different reasons, while your aggregate analytics data only gives you "cattle data."

Segmentation is the realization that not everyone is the same.

The (In)Famous Star Trek Analogy

The Red Shirt Phenomenon: the assumption that if you wear a red shirt, you'll die. But is it true?

Matt decided to study it to see. Apparently, out of 54 total crew members who died in the course of the show, 79% of them wore red shirts. Let's call that a 79% "conversion rate."

So it seems clear wearing a red shirt seems to be hazardous to one's health.

Now we know what happened, but we don't know why. Let's dig in further. Turns out the primary situation in which the red-shirts die is when they beam down to the planet's surface with Captain Kirk. BUT, if Captain Kirk meets an alien woman, the redshirt survival rate increases to 84%, and there were no deaths at all when the alien woman was green.

As a red-shirt, your survival rate increases if you stop beaming down to planets, encourage Kirk to hook up with alien women, visit peaceful planets and stop fighting. (And maybe change your shirt color?)

So now we understand what factors increase the "conversion rate" for that specific "customer segment."

And now if we think about it, we know people use our website differently depending on what they're looking for, and where they come from. We need to measure the statistics for each different segment.

Once again, people aren't cattle. Don't rely on aggregate information. Segment all this stuff out.

Matt tells the story of a company that had a home page bounce rate of 50%. Digging in, he discovered they had worked to get their page ranked high for "UPS" because they sold an Uninterruptable Power Supply. But they didn't have info on the home page about their UPS, because it wasn't a major product for them. (The ranking pursuit was apparently simply for vanity.) The bounce rate for people who came to home page via search on "UPS" was over 80%. So the bounce rate for other segments was much lower, but that one segment was skewing aggregate numbers. This is why aggregate numbers can be misleading.

Need to slice the data to figure who's there, why they're there and what's the context of their visit.

Matt analyzed traffic, time on site and conversions for a site based on where visitors came from:

High context, low competition for attention: Blogs and articles. Highest engagement and highest conversion rate.

Moderate context, moderate competition for attention: topical search. Moderate engagement with an average conversion rate.

Low context, high competition for attention: social news. Stayed on site for less than 10 seconds on average and a conversion rate of zero.

Different people come to your site from different places with different goals in mind.

When you develop a report, is it understandable, clear and supportive of organizational goals? Credibility requires reporting bad news as well as good news.

Path analysis is useless because there are so many paths. Pie charts are terrible unless they're very, very simple. Emphasis should be on the information, not the picture.

Seven reports for measuring success

  1. Keyword segments (or "buckets") -- people who come to your site after searching for one term, aren't the same as people who come after searching for another term. May also need to further slice even more. Segmenting your search data is a great source of information.
  2. Segment referring URLs. Where people come from and what the referring page says about you makes a big difference in how people behave once they arrive at your site.
  3. Segment your bounce rates.
  4. Segment your content. Which pages hold attention and drive conversions?
  5. Segment visitor behavior. Get away from the technical and find simple ways of expressing that behavior.
  6. Compare and contrast segments -- apply knowledge to improve site.
  7. Tell the story.

The most important thing to do is to ask questions. "Question asking is the most significant tool human beings have." (Neil Postman quote)

  • Ask Questions
  • Take Action (if you don't do anything, you're losing money)

It's good to know which links are good for branding, which links are good for ranking, and which links are good for business. Focus your energies on those that are do for you what you need.

If your host won't give you access to your log files, get a new host. That's your information, your data and you need it to maximize your website profits.

September 22, 2008

Learn more about the ways Diane can help improve the performance and profitability of your business web site, or request a no-obligation personal consultation, by visiting www.NineYards.com.


Is your hosting company your only access to web logs or can they be accessed through Google Analytics?

@Sayyebo, Google Analytics uses JavaScript tagging, so it doesn't make use of your server logs at all. If you want to analyze your log files, you generally have to log on to your hosting account server to access them.

I actually think that analytics are "fun and awesome" too! A new perspective Ive gained from this article is the idea that your website exists to fulfill the "customer experience." They need to have a good time at your site so that they will stay on your site longer, return to your site more often, and refer your site to others. By looking deeper into your analytics, you should be able to determine if you are doing a good job of this or if you need to put a little more work into your website design to enhance that customer experience.

Comments closed after 30 days to combat spam.

Search Engine Guide > Diane Aull > Understanding Analytics - SBM Unleashed