Cool custom report: Bounce
Bounce rate is kinda hard to interpret. Sometimes it’s good that your visitors find everything they need on the landing page. Often it’s a bad sign – they didn’t see the relevance of the landing page to their search, or they maybe they found all they were after but you’d rather they also looked at what else you offer.
You, as analyst, get to do the interpretation. We, as Outsiders, want to help you obain the data to interpret.
Why Bounce Rate?
Avinash Kaushik and Rich Page, among others, have often talked about bounce rate. You should be looking at bounce rates for individual pages, pay-per-click keywords, campaigns, campaign messages, and so on.
Examples:
- Site A example: We saw a very high bounce rate for certain paid keyword ads that landed on a product category page, when the keywords were clearly about products, not categories. People weren’t seeing the link from the category to the product so they left in large numbers without getting below the category level. We had thought we were being smart by exposing people to several related products through the category page, but it apparently wasn’t working out that way. So we changed the landing page to the product page and bounce rate went down a lot — because after getting the information they were looking for, people still took a look at other products.
- Site B example: Similar situation, but different. We saw a very high bounce rate for paid keyword ads that landed on an individual product page, when the keywords were, as above, about products. People were apparently getting all the information they wanted and leaving without realizing the site offered a number of related products. It was a flipped-over version of the previous example, but people arriving on the product page were, similarly, not looking at the category’s other products. We used the high bounce rates to get funding for adding “related products” functionality to the product pages.
WebTrends and Bounce Rate
First off, we don’t have a way to get calculated bounce percentage statistics (67%, 23%) out of WebTrends. Sorry.
But we can show how to get the next best thing – total entry page visits and single page visits in the same report. When these two numbers are next to each other, it’s not hard to see problems by eye and if you export to Excel, it’s easy to calculate exact percentages.
The reports will look something like this. Note the center comment bubble – it’s the point of this report, the calculation needed to get bounce rate for a page.
Making the Custom Report
The Primary Dimension
The primary dimension is whatever you want a bounce rate for. It must be a visit-type dimension having to do with entry pages, meaning something that applies to the whole visit. Entry page, search term, marketing campaign, referrer all qualify, and so does a query parameter on the entry page. See our previous post on visit- versus hit-based custom dimensions for more information.
The Secondary Dimension
The secondary dimension is a new custom dimension, let’s call it “Page View Bucket,” which shows “One” if the visit was one page long, “Two or More” otherwise. A visit of 1 page is, of course, a bounce.
You have to create this dimension and it is based on the Page Views dimension. But, for this dimension, go further than usual and use a lookup table to collapse the long list of results you get from the Page Views dimension (visits of length 1, 2, 3, … 100 or more each in their own row) into just two rows, i.e. single page visits and longer visits.
The trick of using a lookup file (a.k.a. translation file) that turns a long list into a short list is cool. Here is how.
The Lookup File
- Create the lookup file and put it on the server (see “The Details” below). If you use WebTrends On Demand, you will have to submit it to them and have them install it. Lookup Table provisioning may or may not already be allowed in your account, talk to your account manager.
- Tell the WebTrends program about the name and location of the lookup file so it will appear in the list of lookup file choices. You do this by defining it as a Custom Lookup Table. You’ll find ”Lookup Tables” as the next choice below “Filters” in the custom-report building menu. (WebTrends On Demand tech support will do this for you, see above.)
- Go into the edit screen for the dimension, click on “Advanced” (in the Based-On tab) and check the box for “Translate Substring Retrieved Above.” When you check the box, WebTrends will show you the list of available lookup files; choose yours. You’ll have to tell it what the Key and Value columns are; they are A and B for the file we supply.
The Details of the Lookup File
Copy and drop the following into a text file. You can call it something like “bucketize_1-99.txt” or something. If you use the software (and aren’t uploading it to On Demand tech support) we suggest putting the file here:
/WebTrends/storage/config/wtm_wtx/datfiles/datasources/
1,One
2,Two or More
3,Two or More
4,Two or More
5,Two or More
6,Two or More
7,Two or More
8,Two or More
9,Two or More
10,Two or More
11,Two or More
12,Two or More
13,Two or More
14,Two or More
15,Two or More
16,Two or More
17,Two or More
18,Two or More
19,Two or More
20,Two or More
21,Two or More
22,Two or More
23,Two or More
24,Two or More
25,Two or More
26,Two or More
27,Two or More
28,Two or More
29,Two or More
30,Two or More
31,Two or More
32,Two or More
33,Two or More
34,Two or More
35,Two or More
36,Two or More
37,Two or More
38,Two or More
39,Two or More
40,Two or More
41,Two or More
42,Two or More
43,Two or More
44,Two or More
45,Two or More
46,Two or More
47,Two or More
48,Two or More
49,Two or More
50,Two or More
51,Two or More
52,Two or More
53,Two or More
54,Two or More
55,Two or More
56,Two or More
57,Two or More
58,Two or More
59,Two or More
60,Two or More
61,Two or More
62,Two or More
63,Two or More
64,Two or More
65,Two or More
66,Two or More
67,Two or More
68,Two or More
69,Two or More
70,Two or More
71,Two or More
72,Two or More
73,Two or More
74,Two or More
75,Two or More
76,Two or More
77,Two or More
78,Two or More
79,Two or More
80,Two or More
81,Two or More
82,Two or More
83,Two or More
84,Two or More
85,Two or More
86,Two or More
87,Two or More
88,Two or More
89,Two or More
90,Two or More
91,Two or More
92,Two or More
93,Two or More
94,Two or More
95,Two or More
96,Two or More
97,Two or More
98,Two or More
99,Two or More
> 99,Two or More
Note that the last line is “> 99″. It’s the last line WebTrends will show for reports with this dimension; everything above 99 goes into this bucket with the WebTrends label “> 99″. Because WebTrends won’t show granularity above 99, this is as far as you can go with your buckets.
Postscript
If you’ve gotten this far, you probably realize that this lookup table business is very versatile. You can modify it to produce any kinds of buckets. For example, we could have made the above to produce “One or Two” as one bucket, “Three to Ten” as another bucket, and so on.
25 comments
OK Crew, I am now a WebTrends Outsider fanboy! This is the close to cracking the Da Vinci Code. While this produces a nice report in WebTrends Reporting, it simplifies creating ODBC reports by several steps!
Thanks for posting the trick …. as well as explaining the flexibilities of the solution.
And, if you have a really good (and probably long, and probably not all that easy to do) lookup file that takes your most common search phrases and assigns each one to a group or type, the first dimension can be search phrase type.
Example types: phrases that contain the brand name (i.e. the individual knows the brand), phrases that are various kinds of generic (problem-oriented phrases, generic product descriptors not related to a problem, competitor-related, ridiculously general phrases)
Exactly, bosilytics. ODBC only has to deal with one report. However, the Excel stuff involves concatenation, a sort, and a two-row formula. Or maybe you have a really great way to do it? Please step forward — I know you are a master of this stuff!
You need to add .txt file into default.ini to show it in WebTrends Analytics.
[datasourceprofile6]
dbvendor = sql
delimiter = ,
dsn =
password =
path = .\wtm_wtx\datfiles\datasources\HourOfTheDay.csv.txt
profileid = ijOYF8yFTF5
profilename = Hour of the Day
profileversion = 1
readonly = 1
type = file
username =
profile id ?
Ibrahim, if you’re having to go into an ini file, then I’m guessing your WebTrends version doesn’t have Custom Reporting, because the data sources can be set up through the GUI with Custom Reporting. Or are you missing it? It’s in Web Analysis >> Report Configuration >> Custom Reports >> Lookup Tables. That’s for WebTrends 8.5. For WebTrends 8.0, it’s slightly different — Web Analysis >> Report Configuration >> Custom Reports >> Data Sources.
The “profile ID” refers to the unique identifier character string assigned to that particular data source. It’s created by WebTrends, just for that data source. The word “profile” is confusing because we’re accustomed to having only one kind of “profile” in WebTrends, the report profile. But it’s a common word for any unique identifier in the raw config files like the one you are showing.
“First off, we don’t have a way to get calculated bounce percentage statistics (67%, 23%) out of WebTrends. Sorry.”
Just noting that this comes out of the free box with Google Analytics. Very cool report idea, but I hope the CMO has someone in product development working on this basic element.
Heh, you and about a thousand other people. Any chance you would be willing to submit a polite but firm feature request? It seems silly to have to do so, but it does matter despite it being, for most of us, pretty obvious. Customer Center >> Contact Us >> Send Feedback >> Feature Request.
I looked all over the WebTrends forums for this and could not find…
I love this solution – it’s simple and flexible.
Thank you, thank you !
Hey rocky,
I am confused by your lookup file. To be more precise the last line confuses me “>99″.
Can WT evaluate this? If not why did you include this line? If WT can evaluate this expression why did you state all the lines 2-99 and not just use two line:
1,One
>1,Two or More
Would be great if you could clarify this.
Sebastian
Sebastian, you’ve pointed out something I absolutely should have put into the text, thanks!
The > 99 is actually a text entry. It matches the text “> 99″ (note the space) that WT puts as the last line when the dimension “Page Views” is used. It’s WebTrends itself that sets the limit as 99 for that dimension.
It would be wonderful if WT’s translation files could do numeric types of recognition or wild cards. They can’t.
I’ll go back and change the main text.
Perfect, thanks for the clarification. Would have been nice
I have found it! (after much frustration).
Create a page views dimension- as per the post above. Then, create a custom bounce rate report and include the ‘Percent of All Visits’ measure. This will give you a percent of single page visits (bounce rate) under the One Pages row in the report.
Whoa. I can’t wait to test it out.
Marco, that’s not working out for me, and maybe I’m not doing it the way you are. In the “Percent of All Visits” column, my numbers are much smaller than bounce rate – they are the percentage of all visits to the site where page ABC was a single page visit. Am I missing something?
Hi Rocky, we maybe calculating bouce rate differently…?
I’m calculating bounce rate: (single page visits/total visits).
This report gives you the total single page visits under the ‘One Pages’ row divided by total visits. I can e-mail you a screen cap. if you like. Does that help?
Hi Marco,
I’m calculating it as
Bounce rate for a given page: (single page visits to that page / total times that page is an entry page*)
Bounce rate for a site: (total single page visits / total visits to the site)
What you’re getting, I *think*, is (single page visits to that page / total visits to the site)
With that clarification, are we on the same page? (I must admit that my information about what your modification produces is obtained from blog visitor Mary, who sent me a screen shot of what she got.)
Yes, do send me a screen cap at the direct contact email below!
* thank you to Mateo for pointing out that the original version of my text wrongly had total page visits as the denominator (5/19/2009).
Hi rocky,
here comes another definition of bounce rate (from the Web Analytics Association): single page visits/entry page visits. I am using entry pages as the first dimension. In general the bounce rate report works fine with this. But interestingly I get a slightly higher visit rate for entry pages there than in the standard entry page report.
Any ideas why?
Aaargh. First of all, we really blew it on our definition didn’t we. The one you quote is correct. Our previous response will be edited and the posting text clarified.
Regarding your real question — it’s not unusual to have very slight differences from one report to another, especially for anything based on visits. In some reports, WebTrends will add to its tally at the beginning of a visit. In others, the visit must be completed and closed before WT changes its stats. This makes visits at the very end of the logs problematic, because some of those visits aren’t finished (closed) and won’t be closed out until a few more hours of data gets analyzed. If this is the cause, then you’ll usually see the discrepancy go away if you choose a time period in the middle of the analyzed data.
Mateo, try that out.
There are probably other explanations too.
Hi Rocky,
To clarify, I am using the WAA definition for a ‘A site-wide bounce rate’ (percentage of total visits that were single page view visits.)
Follow the following steps for ’site-wide bounce rate’ in WebTrends:
1. Create a dimension as per the blog above
2. Create a custom report using a single dimension only (Page View Bucket)
3.Use the % of all visits measure.
This will give you a ’site-wide’ bounce rate.
I’ll send you the screen cap shortly.
Marco
Marco, Okay then, we are on the same page! Now I see what you are talking about. I’m going to build it as you say and take a look.
I suggest this – stop trying to be nice, stop asking nicely – we in the WA world have been putting up with this vendor malaise for too long now. Lets keep in mind that WE are paying THEM.
to the Moderators: sorry about that mini-rant yesterday, it was not intended against anyone despite the similarity in name to ‘rocky’ – in my hurry I did not make the best choice of names. clearly I was in the middle of a bad moment with (the software version) of WT. It was a shortsighted comment in that 1) I got a lot from the discussion and 2) Your website is extremely helpful and a ‘gift’ to WT and all its users.
I have been watching web analysts deal with these vendor issues for 10 years now – it’s always been a love/hate relationship with them. There is no need to treat these vendors with kid gloves. The cart is ahead of the horse.
Thank You, WebTrends Outsider, again for all that you do here.
Hi RnB – We didn’t take it personally (well, Bullwinkle has been grumpy, so maybe he did). As for “nice,” the very existence of The WebTrends Outsider is fundamentally not nice. We expose a lot of its shortcomings as well as its strengths. I know for a fact its competitors are all over this site and I can just imagine how Outsider content is exploited in sell-against activity. And, I guarantee that the people at WebTrends who know who The Outsiders are would never use the word “nice” to describe our other interactions. But our statements about the product and the company are still true – after ten years in the industry, this is the tool we are staying with. As a constructively critical customer.
Stacey, which part can’t you find?
If you don’t have custom reporting in your version of WebTrends, the “edit screen for the dimension” won’t be there. If you do have custom reporting, it’s in Report Configuration >> Custom Reports >> Dimensions.
Your site is the best
Leave a Comment