Skip to main content

Tracking traffic

Over the past few months I have been giving this topic a lot of thought. We all use information every day, and without it we would have a terrible time trying to run our businesses. Traffic is just one such measure, but as a consumer-oriented industry it is used to drive just about all our metrics. Traffic can be defined in many different ways, but for the sake of this article, let's define it as some kind of usage or throughput measurement. There are so many different ways to track traffic, each one playing an important, and sometimes crucial, role in our business.

Every day I spend the first part of my morning looking at traffic. It has become routine for me to wake up around 6:00 a.m. and immediately turn on my computer. My compatriots on the East Coast know I am alert (well, maybe) and starting to track my day. They have become so accustomed to this routine that they often schedule calls at this time. In fact, most of my traffic reports are generated overnight and arrive by e-mail in time for my first cup of java. Let me briefly describe some of them.

The first is an administrative report that tracks registrations on specific AOL services that I am interested in. This report is gathered and manipulated by various information sources within the company. Some of the data comes from our network switches, telco switches, routers and servers and is processed by one of our most valuable resources — our reporting group. It tells me how many people have signed up for a service on a daily basis, which promotion they used, how many canceled, 30-, 60-, 90-day churn rates, service-usage statistics and a whole wealth of other information.

This type of tracking has been invaluable to our business leaders. It is immediate (I only get a daily summary but finer reports are available if needed) so, as an example, we can tell if a new promotion is working or not. More importantly, as we roll out new features we can judge their effectiveness. Usually after looking at the raw numbers my eyes immediately migrate towards the trend graphs. A trend is a set of statistics plotted over time. If we are trending in the right direction (usage is up, registrations are up and churn is down) we get to call the team to congratulate them. If we start trending the wrong way, we discuss adjustments to current market plans or features.

The next type of report is a little more technical and gives us hours, minutes and seconds of use for the entire service. We then break this information down by region, function, feature and overall usage. This allows us to do network trend analysis but, more importantly, we can derive the health of our network/service. If we see a feature that all of a sudden varies from its normal trend, a red flag is raised. Most likely that feature is having technical difficulty and our customers are not able to access it properly. Certainly, if we had an update on the service we would know if something is broken. These trends allow us to make sure we are predicting our growth so we can order more bandwidth, increase staff and beef up our server complex.

The last report I get in the morning is the roadway traffic report. Are the bridges packed? Is BART on time? Is the 680 or the 880 flowing? Are there any accidents? This is the type of flow report that helps set up my day. So how does all this affect the way you should look at tracking your traffic?

The interesting thing about any report or statistics is that there are so many different ways to integrate the information into your work. A bandwidth usage report means totally different things to a network engineer than it does to a marketing VP.

The first thing you need to do is determine what information you have available. Then decide how it is going to be used, by whom, how often, and lastly, how to parse all the raw data into reports with meaningful statistics, ratios and comparisons. Most tracking reports should allow for trend analysis, as this is usually the most telling information.

If you are using an outside hosting company, most of them will provide you with various reports that flow from the raw data logs of their servers and switches. You need to be able to specify what data is important to you and how often you need a snapshot.

Two good reasons for studying growth is to plan capacity and track business models. This can be done by interpolating the possible range of growth. This is derived from the “best-fit” regression curve based on your trend analysis. A regression analysis allows you to look at growth within a confidence interval or a range of possible values.

Using this kind of analysis will help you know how valuable your data is. At a 95 percent confidence interval, the true value will fall outside the estimated range five percent of the time. A small variance within this range gives you “confidence” that the data is accurate. You should discuss these types of analysis with your reporting groups so you can understand how they are derived and how relevant they will be to your needs.

Recently, a product announcement came across my e-mail and I was intrigued so I checked it out. Visualware recently announced a new release of VisualRoute, which is a graphical trace-route tool that provides Internet connectivity information for network troubleshooting and IP geographical locations for security purposes. VisualRoute 6.0 is available on the Windows, Linux, Solaris, FreeBSD and Mac OSX platforms.

VisualRoute has the ability to identify the geographical locations of routers and servers. It can provide you with an effective means to identify the location of different groups of your users. The release includes new features, such as integration with Internet Explorer, automatic protocol recognition and an improved interface. VisualRoute automatically analyzes Internet connectivity and performance problems, displaying the results in an easy to understand table and on a world map. A trial version is available from Visualware's website at

Steven M. Blumenfeld is currently the vice president of advanced services for America Online.