A common question in web analytics, since most tools use a cookie to identify returning visitors. Some tools fall back to IP-address and browser agent, but that will not help either when multiple computers are used.

In some cases website visitors have to login which is a great alternative to increase the recognition of returning visitors. But what to do when you do not require users to login and you feel that they might use multiple computers more than average?

A real-world example from an online insurance website shows you one of the solutions. Based on the conversion per marketing channel the people responsible for online advertising were sure that a lot of visitors hit one of their banners or links before converting. During the order process people have to fill in a specific registration number for a car insurance (legal requirement, many people don’t know this number), which increases the chance that they abort the order process. By proving an email link the visitor can choose to receive an email with a link to resume the order process. Unfortunately it is quite likely that this will be done on a computer at home, while the order process started on a computer at work. No cookie, no similar IP and thus no way to match a conversion with a previous marketing campaign response. As a result the reported ROI on campaigns is too low, because many orders cannot be matched with a campaign.

The solution is to use an identifier which is available in both parts of the order process. For this car insurance it is the license plate number. Before visitors have to fill in the code (which they can only find on their car registration papers) they always fill in their license plate number which shows the premium for their car. The license plate number is matched with all marketing campaigns they have responsed to (Google Adwords, MSN banner, etc.). When they complete the order process the license plate will be stored in the web analytics tool on the order confirmation page. By combining this information with the list of license plate numbers and corresponding marketing campaigns almost 50% of the orders without any campaign information could be allocated to one or more campaigns.

Disclaimer:

Sometimes a little out-of-the-box thinking is required to get valuable insights out of your web analytics tool. Adversitement is a premier European partner of Omniture and is specialized in advanced web analytics projects.

Met vriendelijke groet,

Mirte Romanillos

10-Jun-08 8:00 AM
View post: 
How can I recognize visitors if they use multiple computers?

Popularity: 23% [?]

Many times, statistics from web analysis can be misleading. It is all too easy to end up doing the wrong thing based on analysis of website statistic. Here are some pitfalls for the following metrics:

1. Number of Leads by Keyword

   It is very important to measure conversions by keyword. This lets you know where to focus your marketing efforts. However, many B2B sites and especially those with high cost items have a relatively small number of conversions. In addition, the conversion may occur:

   • During the 2nd, or later visits when the original keyword used is now lost (due to cookie erasing, subsequent search using product name, etc)
   • By a coworker of the original searcher who lands directly on site since the URL is now known

   Since the numbers are small and may not be traceable to the original search, in many cases it is statistically invalid to decide on actionable items based on the number of leads by keyword. In those cases it is best to find proxies for conversions. Possible proxy candidates are time-on-site or engaging actions.

2. Percentage of Leads

   The percentage of leads from total visitors or any other segment is not always relevant. In many cases the profit from a B2B sale is big enough so that a good lead can justify the cost of a campaign even though the percentage of leads is small. The absolute number of leads is more important in this case. If you see that the percentage of leads from traffic is going down but the absolute number is going up–you can should go out and celebrate your success.

3. Absolute Number of Leads

   Measuring the absolute number of leads can mislead you. Yes, I know I just contradicted the previous point. Unfortunately you may sometimes notice the number of leads is decreasing. Before you panic, it is useful to measure the percentage of conversions. This is because there are many times when seasonal or other time-based factors impact total traffic.

   If we just measure absolute numbers and we see a 50% drop in leads we start to panic. But if you see that the percentage of leads is consistent over the last few months you know that the reduction in number of leads is due to a drop in traffic.

   You should then analyze to see if the reason is seasonal or other factors we have no control over. If it is seasonal, we can relax–although we should still try and improve the percentages. If it is a factor we do have control over, we can then start to panic and work to rectify the situation.

   It is useful to measure percentage of conversions to use as an early warning sign, however our main goal should be to increase absolute conversions until the expense of increasing them outweighs the profit.

4. ROI

   This important metric has great PR but it is undeserved. As long as you are profiting from a campaign, the return on investment should not be use to eliminate ad campaigns. You can use it:

   • If you need to reduce your advertising budget this metric then becomes necessary in order to guide you to which areas it is best to reduce the budget.
   • To measure your optimization efforts

   If you are selling on your web site, make sure to measure the revenue for each conversion and not just the number of leads. Not all conversions are created equal.

5. Number of Downloads

   Not all downloads are created equal. Whitepapers are typically downloaded earlier in the sales cycle. In addition, you may get many non-qualified people downloading the white paper who are interested in the subject.

   Data sheets, on the other hand are usually downloaded by people later in the sales process and who want to see detailed specifications of your product. By measuring downloads you are lumping these and other different segments together. Best to measure whitepaper downloads separately from data sheet downloads.

   An upsurge in white paper downloads in October may be because students are studying the subject described in your white paper. On the other hand, an upsurge in data sheet downloads is usually great news–unless you find out that it is your competitors doing all the downloading.

6. Using Numbers that are Statistically Valid

   In many cases metrics do not have enough information to be statistically valid. Unfortunately there is a tendency to want to come to conclusions fast. This could be because:

   • You want to prove something and are over eager to bring the testing (with the results you wanted) to a conclusion
   • There is pressure to present actionable items to others

   Avoid the pressure. I have seen many tests where the results flip flop once or twice before the numbers are valid.

7. Experience and Knowledge.

   Numbers are great and no one loves them more than me. However, they are just numbers and have many disadvantages:

   • There is still a lot of information they don’t include. For example they don’t explain why people do things
   • There may be mistakes in the data
   • The conclusions may not make sense and by being stubborn and digging deeper you can usually find the reason
   • In addition, they can be manipulated to prove preconceived ideas-sometimes uncounsiously. As my seventh grade math teacher said: Figures don’t lie, but liars figure

These are some of the web analytics pitfalls and mistakes we have come across. I am sure there are many more. If you have any, I would love to hear from you.

1-May-08 5:00 AM
Read the original post: 
7 Web Analytics B2B Metrics Mistakes

Popularity: 20% [?]

I remember what it was like to walk through the door at my brand new job, my very first job as a web analyst, wondering what I’d gotten myself into. In retrospect, what did I wind up learning the hard way? What would been helpful to know up front? What should I have been prepared to expect? With that in mind, here are 10 things I wish I knew when I started in web analytics:

1. You will sit between the techies and the marketers. Figuratively, and maybe literally. Make friends on both sides of the fence.
2. You will learn all about your business. Not just the stats part. Not just the web part. The work you do in web analytics will only make sense once you’ve put it in the general context of your business.
3. Ahem, what is this thing you call a “Visit”? Know your standard web metric definitions by heart, and be able to recite them concisely for people who ask. They will ask.
4. Dirty, dirty, dirty. Numbers won’t match, they won’t add up, they won’t make sense, sometimes they won’t even exist. Know how much dirt you’re willing to live with, then accept it and move on.
5. You will learn to love the query string. You will come to see it as a beautiful haiku. You will know it backwards and forwards. You will repeatedly explain its usage to people who need to append campaign codes to URLs.
6. CSV stands for “comma-separated value” … it’s a file format, every data analyst’s friend, and - inexplicably - it doesn’t even have to be comma-separated. Huh.
7. Operators are standing by. Know the support hotline number for your commercial web analytics vendor of choice, and don’t be afraid to call. If you have one sticky note on your monitor it should be that number. Actually two sticky notes. The other one should say, “Patience is a Virtue.”
8. Don’t fall into the “report monkey” trap. Manually-repetitious activities are not a good use of your time, so automate wherever possible. Strive to spend your cycles doing thinking fellers work, and leave robot work to the robots.
9. You are not alone. Right now there are other web analysts sitting at their own desks, somewhere between the techies and the marketers, and they’re facing exactly the same issues that you are. You will meet them at Web Analytics Wednesday.
10. Think long-term. From the very beginning, think about where you want your career to go and make every effort to develop in that direction. Your entry-level position in web analytics can/should/will lead to other things, so know what you’re targeting and go for it.

About June Dershewitz

June Dershewitz is currently Vice President of Analytics at Semphonic, a leading web analytics consultancy with headquarters in San Francisco and offices in Boston and Washington, DC. Read June’s blog at http://june.typepad.com/.

11-Feb-08 10:30 AM
Original post:
Top 10 Things I Wish I Knew When I Started in Web Analytics

Popularity: 15% [?]