your opinion matters!
Web Analytics
Diverted Visits Are Not Bounced Visits OR Why it is important to measure all your outlinks and downloads
Mar 29th
Flawed? The accuracy of your website’s visitor behavior measurement is more important than ever before. But sometimes you are not sure if your website measurements are correct or if the numbers are true reflection of your visitor’s behavior. Since, it is the data that guides you on your decisions for spending and optimization, it should not be flawed.
Here, I would like to examine one of the important measures on your website. This measure says more about your content and marketing relevance than any other measure. It is also an important measure of your site’s user experience. Bounce Rate is widely known to be one the most telling metrics you can monitor on your website. The short definition for it is “single-page visits”. It is generally measured by taking the number of visits that included only one page view and dividing it by the website’s total visit count. Since the single-page visits are considered “bounced” visits, this calculation will give you the website’s bounce rate.
Single-page visits / total website visits = web site bounce rate
Like any good analyst, you should also measure bounce rate at the page level. The method for measuring page bounce rate is same as your website.
Single-page visits to page / total visits to page = page bounce rate
To further refine this metric you must determine the weighted bounce rate but that’s the subject for another blog post.
A bounced visit is understood to be a wasted visit. Irrelevant. And the business objectives set out to reduce the bounce rates on websites. However, in reality, not all bounce visits are wasted visits, depending on how your site measurement is configured (as we will examine bellow). A visitor that comes to your site from any outside channel, lands on a page, and then clicks on your Facebook link, may appear to be a bounce visit, but in fact is an engaged visit, not a wasted visit. These visits that only view one page on your site and, because of your well designed content and relevant messaging, do not go to another page on your site, are not bounced visits. They may end up clicking on a link that takes them to a sister site, branded site, micro site, competitor’s site (for service or price comparison), social page, … and therefore are diverted not bounced.
Diverted Visits – I am certain, regardless of your website’s objective, a portion of your bounced visits are in fact diverted visits. Diverted visits are those visits that view only one page on your site, during their visit, but still perform an action you had intended but not captured as part of that visit (eg. clicking on an outlink on that page).
Viewer Visits – viewer visits are also those that only have one page view in their visits, but while they are on that page, they perform other actions that are not captured as part of that visit (eg. downloading a document, viewing videos, clicking on an email link, interacting with Flash content).
How to fix bounce rates – to effectively fix your bounce rate measure you will need to remove the Diverted and Viewer visits from your single-page (bounce) visit counts. The best way to do that is to tag your outlinks, downloads, Flash, Videos, and email links. Some of the web analytic tool vendors, by default, include the logic in their tag to automatically capture clicks on outlinks and downloads. But whether this tag feature is configured or working properly on your site is a different matter. To check if your tag is capturing these link clicks is fairly easy to determine by using a tool that shows you the tag’s operation (HTTP header) as you browse your site. There are many of these tools out there. One of them is called HTTPFox, a browser plug-in. I know this is easier said than done. But the consequences of not checking your tag for these links means inaccurate bounce rate measures. Once you have checked your tag’s operation or have tagged these links manually, the clicks on outlinks, downloads, … are captured and counted as part of the visit. These visits are no longer single-page (bounce) visits since the subsequent clicks on the page are also counted as “page views”.
Other option for fixing the bounce rate is segmenting the single-page visits by those that have clicked on an outlink or download link and subtracting this segment from your total single-page visits before calculating your bounce rate. This tends to be a custom filter or segmentation that is usually not included in your reporting tool and may be more trouble than the previous option.
It is hoped, by many analysts and practitioners, that one day we will all operate from a set of standards for data collection, measurement, and analysis. But until then, the practices we follow today will need to continue to be fluid and improve as the online business practices change and improve.
Steve Bashiri
Web Analytics Governance and Metrics Standard
Oct 1st
An effective web analytics practice needs proper staffing, management, planning, and diligence. To be effective, a governance committee must be formed, consisting of individuals with specific roles and responsibilities (see table below). With the right resources and clear expectations, the online business can put valuable data into action to achieve higher performance. But, how do you obtain this valuable data?
Data must traverse several stages before it is ready to be acted upon. From visitor generated raw format to the actionable metrics, data must go through processes similar to a typical factory before it’s consumable. Once raw data is sessionized and aggregated, it must go through segmentation and calculation. An analyst then slices and dices the data to find correlations and nuggets of information that is valuable to the business. Data analysis generally produces many findings. With data sharing and collaboration within the governance committee, findings can be pared down to a few high value, high confidence metrics. These will then be communicated as recommendations to the business users.
An online data driven organization usually has many contributors to the decision making process. These data consumers have many sources from which to extract data from. Even though the nature of data will be different for each of these sources (ie. marketing, finance, or technology) there need to be a baseline. A standard for data means having a common denominator, a similar yard stick by which everyone gathers and analyzes data. Some examples include how online visitor interactions are measured, eg. organization wide acceptance of measuring visitors (by cookie, login ID, or session parameter), or what constitutes a page view and does it accurately represent visitor’s interaction with business content, or what method should be used to represent bounced visits. Is it all visits with single page views, or only single page views from a filtered list of IP addresses, or those from “known” marketing programs (SEO, SEM, eMail, banner, and affiliates). The final decision on the organization’s metrics standard will depend as much on the nature of the business, as it does on the individuals in the governance committee. However, once established, everyone’s analysis and findings will be based on these standards. And only then the collective wisdom of the organization wide data becomes far more valuable than any of its single contributor.
Business User
Role
- Responsible for budget
- Interface with business analyst
- Acts upon analysis and recommendations
Tools
- High level knowledge
Analysis
- Light analysis
Technology
- High level knowledge
Management
- Budgets and other dept/Org resources
Web Analyst
Role
- Interpreting web data
- finds nuggets of high value information
- Focus on performance and optimization of the online properties
- makes recommendations
Tools
- power user
- can interface and extract data from all tools
- configures tools
Analysis
- Deep analysis
- slice and dices data
- interprets quantitative and qualitative data
Technology
- Good understanding of designs and methods
- Assists on evaluation and recommendations
Management
- Little to no duties
Business Analyst
Role
- Interfaces with business users
- Deep understanding of the websites
- Gathers business requirements
Tools
- Interfaces with tools to extract data
- Designs reporting solutions
Analysis
- Analyzes online and offline data
- Documents requirements
Technology
- Conceptual knowledge
Management
- Some budget and human resource
Developer
Role
- Develops tagging and programming to capture business data
- Interfaces with web analyst
Tools
- Deep knowledge of web analytics tools
- Designs and develops best practices
- Documents technical requirements
Analysis
- Light analysis
- Reviews technical data for optimum online system performance and availability
Technology
- Deep knowledge of the methods and practices
- Makes recommendation on and evaluates new technology
Management
- Little to no duties
Project Management
Role
- Major project owner
- Facilitator and responsible for achieving deadlines
Tools
- High level knowledge
Analysis
- No analysis
Technology
- Understands concepts and main drivers
Management
- Manages major projects
- Overseas all resources contributing to projects
Robbed in the Daylight
Nov 8th
Online businesses are losing billions of dollars every year to fraud. Businesses in finance, ecommerce, social networking, and other verticals are being targeted everyday by individuals that have made it their job to steal from these businesses. These are individuals (fraudsters) with skills ranging from novice to highly skilled programmers, systems and networking experts that can hack into businesses and steal. The damages they cause range from stealing identity, steal products, to stealing the customers. With sophisticated programs they appear as normal customers as they engage in attackes on a business. Most businesses are ill-prepared and have little or no plan of action to combat these types of challenges. The ones that do have process in place to deal with this type of attacks deal with them in a reactionary way. They plug the hole after it’s discovered. After it’s done some damage. By monitoring and analyzing data, fraudsters can be detected before they can inflict significant damage.
Methods
A fraudster may create multiple accounts on a social network site. By multiple I mean hundreds and sometimes thousands of phony accounts. This obviously would be difficult for a human being to do, manually. So the fraudster will develop code to do the work automatically. The code would automatically signup accounts on the web site with bogus information and create user profiles. These profiles will attract legitimate users that will offer their information in hope of connecting with others. This information can sometimes be very personal. But mainly, the fraudster is interested in collecting account information to sell to other businesses or trick users to join other online services. This scam could be collecting the customer’s email addresses which they can use in Phishing scams. A Phishing scam is when a fraudulent web site poses as a legitimate web site to collect information from victims, information like their credit card number or online credentials to other web sites.
Detection
An online business will need to collect and analyze data to detect fraudulent activities. The sophistication level of fraudster’s method can make it difficult to detect their behavior. However, by informing themselves, businesses can review their online data for unusual visitor activity and investigate the cause. This is done by learning the patterns in their data. Every business will have its own unique patterns of visitor interaction. One of the common methods of detecting fraudulent online activity is to look at large amount of activity by a single IP address, in a short amount of time. This could mean a fraudster is running a robot program from a single computer to perform attacks. One has to make sure the IP address is not a proxy IP address which at times represents many different individuals, possibly legitimate ones. The activities to look for can be further segmented to focus more on the high valued ones. These can be account signup, login, and sending emails to other users.
To look more like a legitimate visitor, a fraudster’s attack can come from multiple computers spread across a large geographic location, or at least across multiple IP addresses, making it harder to detect their attack. Fraudsters with more resource at their disposal can hop from location to location and have banks of computers and modems to avoid detection. Like a business, they optimize their code and methods to be more effective.
Prevention
A common method of prevention for online fraud is using CAPTCHA. Wikipedia’s definition of CAPTCHA is: “type of challenge-response test used in computing to ensure that the response is not generated by a computer”.
These are sometimes images of single words or phrases that are morphed and distorted so a human can read them but it would be difficult for computers to decipher. These images are placed on web sites where visitors signup or log into their account, or perform some high value actions. The visitor is asked to type in what they see before proceeding with the action. Bellow is a sample of CAPTCHA image you might see on a web site.
There is no substitute for prevention like educating your visitors/ customers on fraudulent activities that they might get subjected to. Businesses need to regularly communicate with their visitor community on what to watch out for, in relation to their site, and not become a victim of online fraud.
As the Internet has evolved and continues to evolve, so has the online fraudulent and criminal activities. Businesses that have not been paying any attention to this area will need to start engaging now. Chances are they are being subjected to fraud in one form or another. To start with, the online business will need to stay vigilant on educating themselves on different forms of fraud. They would then need to develop process for collecting and analyzing data that would provide insight into possible fraud activity. Next, there needs to be tool(s) put in place to support the designed processes. These can be software available online or developed by business’s own resources. There are also companies that provide services to combat fraud.
Fraud management should be treated like most other business processes management. It needs to have its own life-cycle (Education, data gathering and analysis, detection, and prevention). First, learning what are the different fraud methods being used, then through data analysis fraudulent activities are detected, then prevention measures are put in place (plugging the hole), and then the process begins all over again. This represents a continuous cycle. Those involved in managing fraud need to understand that, just like pests in the house, once you see one or two cockroaches, there are probably hundreds or thousands lurking in places that you can’t see them.
Steve Bashiri
Recent Comments