CHAPTER 15 Web file log analysis tips

Log file analysis

Some Web sites may be attracting hoards of visitors, but does a long list of impressive statistics really add up to success?

Ever since the dotcom boom captured the attention of the world's media, an inordinate amount of attention has been given to how many people are using the Internet, how many are visiting a particular Web site, and - perhaps most importantly - how much money these people are actually spending online. With many Internet businesses still struggling to rebuild dotcom dreams that had been built on sand, it's not surprising that industry attention has now turned to finding reliable metrics to use for predicting online success.

What is surprising, though, is that this commercial view - the idea of using metrics to determine how much revenue a Web-based business will make - has only come into the equation recently. Previously, a high number of hits to a Web site were often enough to pave the way for the over-valued stock floatation's that have now become part of Internet folklore. For obvious reasons, this mindset is changing as the drawbacks of this method of measurement become more apparent.

For a start, hits and page impressions do not equate to real visitor numbers. Page impression logs consist of every single request for a page or graphic that the server receives. Today, Internet service providers like AOL or Freeserve save their own versions of popular pages. When their users ask to see these pages, the ISP shows them the copy they have saved on their own server, rather than accessing the server hosting the site, which can of course play havoc with page impression tools. The page may be requested from the site's server just once by the ISP, and then shown to thousands of users without registering on the server log.

Plus, since many people connect to sites from work-based computers that lie behind a protective firewall system, the numbers can be skewed even further, since everyone behind a firewall will appear to have the same IP address. To the page impression analysis tools, these multiple users will look like one single user.

Drowning in Data

But the use of page impressions as a key measure of online success has another, bigger problem: page impressions don't pay the bills. In the early days of the Internet many sites used page impressions to gauge a site's effectiveness, but today the mere novelty of a Web presence is not enough to merit success. The sites of today have to earn their keep, and people have learned very quickly that an impression doesn't always translate into a sale.

As a result, more sophisticated forms of measurement - solutions that look at a much wider range of online activity - are being introduced all the time. It's not difficult to see the value that can be gained: Web-based business should be completely transparent and, because it's interactive, customer activity can be measured and recorded easily. In theory, this model should sound like a marketing manager's dream, but in reality, once they move beyond measuring page impressions, many companies often end up feeling like they are drowning in data they can't interpret effectively. So how do you find the balance?

When trying to determine whether or not your Web site is effective, first you have to define 'effective'. What are you trying to achieve with the site? Perhaps you want to generate high traffic volumes via search engines to generate leads, or maybe work towards building brand loyalty? Perhaps you simply want to sell digital goods online. The key metrics you should be looking at will vary according to the overall objectives of the site.

Obviously, the metrics you need to look at will depend on what you want your site to do. If you have an e-commerce-enabled site then revenue will probably be your key metric; you may have reducing unique user visits but more spend per user and so be happy with site performance. If you have a 'brochure' site that has been designed to make people aware of your products, then the new users, combined with length of time each user spends in the products section (as opposed to the company information or complaints section, for instance) will be your key metrics. If your site is intended to be a key resource for your supply chain, giving them specific information on what to deliver in each product to make sure it fits specification, then the site will be effective if it prompts lower product recall problems and lower costs for you and your suppliers.

One Step at a Time

So, in effect, measuring online activity should come as a second step; the first step is deciding exactly what you want from your Web site, which can actually be more difficult in some cases than the measurement. In other words, before you get bogged down in too much data, you first must ask yourself some key questions: 1. What is the purpose of my Web site? 2. When visitors come to my Web site, what is the main action I want them to take? 3. Bells and whistles aside, what is the easiest, most effective way to track this?

The answer to question three is where technology, and software developers in particular, enter the picture. The key here is to invest in software that accurately tracks user activity. The cause of inaccurate data is frequently down to the misuse of server logs, which were never designed to inform business strategy.

In many cases you can use a basic analysis tool to give you an idea of where you stand. Once you calculate a set of baseline statistics for your site, track them religiously and make changes accordingly. The key here is not to get overwhelmed with millions of different data points. Relying on two or three metrics can make a world of difference.

Indeed, there is little value in measuring accurate data unless you analyse it for maximum benefit to business performance. Yet recent research from The Survey and Statistical Research Centre, Sheffield Hallam University, has produced surprising statistics about the way companies use the information they collate on how their customers are using and interacting with their Web sites. The report, commissioned by software giant Oracle, revealed that although 91 per cent of respondents from more than 500 companies surveyed said their Web site was their most common channel to market, half said that only 10 per cent or fewer of their visitors were converted into customers, only 28 per cent could tell who their best customers were from usage of their Web site, and only 34 per cent could measure visitor loyalty by looking at the number of repeat visitors/customers.

The first thing you should do, before anything else, is sit down and think about what the aim of your site is. Is it to showcase products? Is it to collect product feedback? Is it to promote a brand? Is it to inform a few key customers? Is it to save my workers from having to answer basic questions about my company? Is it to attract new customers? Or to entice old customers to return? Or to keep current customers happy? Or is it simply because everybody else did it, in which case you should start rethinking whether you need a site at all. Then put the aims in priority order. This information should be the rock solid basis for any metrics work you do.

If there are any universal questions, they include: why are users dropping out during registration or online purchase? Why are users dropping out prior to registration or online purchase? What is special about users who do register or purchase online? Metrics that relate to common error messages, or help establish which page of an online form results in the highest dropout, allow sharp-end issues to be identified and tackled quickly. Monitoring, aggregating and comparing the paths of users who register/ purchase and those who don't can reveal patterns that explain the differences in behaviour. Sometimes the analysis might link behaviours across many pages to reveal that certain content or features have a strong correlation to a positive or negative outcome.

If you receive a large number of complaints about a particular section and Web logs show sharply declining visits to that section, you've got strong evidence that a change is needed somewhere. Customer feedback is particularly useful as a complement to user statistics - for instance, you might see that fewer people are visiting your fun and games section. You might provide a quick poll asking if users want to see new games, more screensavers, e-cards, and so on. It is expensive and time consuming to get customer feedback from primary research, but it can fill in gaps in the 'hard' metrics. Our recommended approach is to use the two together in a phased approach. First, analyse the site's performance using hard metrics and identify problem areas, then get soft feedback from users focusing on the problem areas identified in step one, and then fix the problems and use hard metrics to confirm that they have been fixed.

Back to Basics

Yet, so far, all the metrics mentioned make one very large assumption - that the site is online and available to customers. Isn't Web site performance the one universally important metric, relevant to all sites regardless of their specific business objectives? "Absolutely"

You should expect nothing less, and in some cases even more in regard to performance and customer satisfaction, from an e-commerce site than you would from a bricks and mortar retail store. Downtime is not an option: 'Open 24 hours a day' is one of the unique selling propositions of the Internet. If your site can't be accessed, a competitor is only one click away, so finding a reliable host is important. Ask your service provider for a year report. Beware of repair/cleaning downtime, in this day and age this should be done while the network is still up. Hosting companies should also be able to guarantee 99.999 per cent uptime, which means only 288 seconds lost per month, compared to 3.65 days a year with 99 per cent. Also, reports should be available, free and always assessable. Any reliable company should be monitoring performance on a continual basis and reports should be available on the Web 24 hours a day.

If you're beginning to think the whole issue of Web metrics sounds a daunting task, don't worry. In terms of a call to action for SMEs I suggest that, third-party service providers should be invited to pitch for business. There is no cost involved and it is up to the service providers to demonstrate the value of their approach. Essentially, the important issues and the benefits and suitability of metrics will become much clearer, allowing SME managers to make more informed decisions about the value of investing in reporting software or commissioning the work to specialists. In the mean time, user feedback should be encouraged, as this is far more likely to uncover key issues and requirements than doing nothing at all.

ROI

Return on investment is one of the keys to internet marketing success. Imagine that you can pay 10p (a) for every user you receive from another website. And you know that your conversion rate is 5% (b).

If your average order value is 100 pounds (c) and the average profit is 40 pounds(d). This means if you receive 100 visitors to your website it will cost 10 pounds.
(100 * (a) = 10 pounds).

From this 100 visitors you receive 5 sales (100 * (b)).

These 5 sales are worth 200 pounds profit (5 * (d))

This means for every 10 pounds (e) you spend on advertising you receive 200 pounds profit (f).

You can now work out your ROI (f/e *100) in this example it would be 2000%.

How many companies can say they receive more than 100% ROI on each 1 pound spent on advertising?

Using the internet and the above equation you can guarantee to achieve this. If you are not receiving 100% you need to either increase your conversion rate or decrease your cost of acquiring a user.

How Do You Measure Success?

You'd think that success would be easy to measure, but advertising has never been a simple art. Ad agencies have their unique self-serving spin, advertisers set their own objectives, and banner ad designers see something else again. Here are some of the factors involved:

Click through Rate (CTR)

This is a basic measure of how effective an ad is. CTRs range from the industry average of about 0.39% to 10%. As a general rule, the more targeted the site, the higher the CTR. For example, you'd expect an ad for Wilson Tennis Racquets to get a higher CTR on a tennis site than on a general sports site. A run of site on a general site such as MSNBC would get an even lower CTR. (Disclosure: I hold no financial interest in Wilson Sporting Goods, but wish I did.) Directories and search engines also sell banners ads that pop up when a particular keyword is entered. Thus your banner could show only when someone entered a search word that included the word "tennis." However, the more targeted the banner exposure, the higher the CPM (cost per thousand banner views).

Cost Per Sale

A much more important figure is the actual cost of making the sale of a tennis racquet. In the final analysis, you don't care how high the CTR is if it doesn't result in a proportionate number of sales. What complicates this is the fact that your banner ads on the World Tennis Ratings site may actually sell fewer tennis racquets than those on NCAAChampionships.com. You can only make this determination when you use sophisticated tracking methods using cookies to separate the lookers from the buyers, and determine which sites and which banner ads had the best result. This kind of precision is enabled by using the DART ad server system from Double-Click, http://www.doubleclick.com/publishers/service/ as well as the sort of tracking used in affiliate software programs.

Branding

While CTR and cost per sale relate to direct marketing objectives, another way of looking at banner ads is as "branding" tools. They create brand awareness, and a brand image in the viewer's mind, whether or not the viewer clicks on the ad. But hopefully, when the viewer gets ready to make a purchase, those "impressions" (a wonderful ad agency buzz word!) will cause you to select Coca Cola over Pepsi, or Barnes and Noble over Amazon, or JCrew over Lands' End. Branding is very difficult to measure, but can be very powerful. Typically, only the larger and better-established companies have the budget to pursue branding consistently. Brand awareness is sometimes measured in surveys with questions such as: "What brand names can you recall in the field of tennis?"

CHAPTER 16 NEXT
BACK TO CHAPTER 14