Website intelligence and return on investment

Half of the money I spend on advertising is wasted; the trouble is, I don’t know which half.

(William Hesketh Lever, the 1st Lord Leverhulme, 1851–1925. A similar quote expressing an identical sentiment has been attributed to John Wanamaker, the US department store entrepreneur and advertising

pioneer, and also to Henry Ford)

What we call results are beginnings.

(Ralph Waldo Emerson)

I notice increasing reluctance on the part of marketing executives to use judgement; they are coming to rely too much on research, and they use it as a drunkard uses a lamp post: for support, rather than for illumination.

(David Ogilvy)

This chapter is about

l How will I know where to invest my digital marketing budget for maximum ROI?

l What information can I track on my website, how is it collected and how can I use it to inform my digital marketing investment?

l How can I track my online advertising that appears on other sites?

l What are key performance indicators (KPIs), and how can I use them to gain insight into my online marketing?

l Why is testing important, and what types of testing should I be doing?

Measuring your way to digital marketing success

The rise of performance-based advertising

Brandt Dainow

The web is constantly evolving, as we all know. It’s a changing environment because people are in the process of migrating from traditional media to online. This migration will not be total -- people will still read books and watch TV – but the emphasis is changing. Print and broadcast media will become alternatives to the new mainstream – online.

We are passing signiﬁcant markers in this migration right now. For example, people aged 17 to 25 now spend more time online than watching TV. I recently spoke to a friend whose son was starting university. He visited the student halls of residence and was surprised to see that none of the students had stereos (an essential of student life in his day) or TVs in their rooms. When he asked about this the students laughed at him -- why would they need those when they had computers and internet connections? To this younger generation the idea of a device that can only handle a single medium, and that isn’t connected to the web, is laughably archaic.

It’s clear that we are evolving new ways of participating in society, new ways of communicating, and new ways of disseminating information. A key dynamic in this process is the transition of the print and broadcast advertising community on to the web. As this occurs, new models of advertising become possible.

Henry Ford once said ‘I know only half of my advertising works. The problem is, I don’t know which half.’ The web solves this problem. The ability to record people’s behaviour online means advertising can be assessed in terms of the behaviour people exhibit after being exposed to an ad. It then becomes possible to pay for the behaviour instead of the mere delivery of the ad. This is called performance-based advertising.

A shift from selling audience to selling behaviour

The dominating trend in the evolution of online advertising is the rise of performance-based advertising. Predictions are that there will be $40 billion in online ad sales in 2008 and that 50 per cent of this will involve performance-based payment. This represents a shift from selling audience to selling behaviour.

The traditional form of advertising involves selling audience. In print and broadcast, advertising rates are largely determined by the number of people who will be exposed to the ad. As traditional media employees moved online they took this model with them, selling ‘impressions’. Banner advertising is traditionally sold this way.

Impression-based advertising simply consists of placing an ad somewhere on a reader’s computer screen, in a manner similar to placing an ad somewhere on the page of a magazine. Performance- based advertising involves changing the emphasis from views to actions. Instead of paying the outlet to deliver my ad, I will pay it for delivering people.

The most common forms of performance-based advertising are PPC advertising and afﬁliate networks. Google’s AdWords is a classic example of performance-based advertising. Advertisers pay not for exposure but for the people Google sends to the advertisers’ sites.

Even where performance-based advertising is not the obvious basis upon which the advertising is being sold, it is often the way in which it is assessed. Mark Read is director of strategy at WPP UK, one of the world’s leading marketing communication organizations. According to Read, many of WPP’s clients, especially in ﬁnance and automotive, convert the metrics from their ad outlets back into performance metrics.

‘It doesn’t matter how people sell the ad space; it’s bought on a performance basis whether they realize it or not’, says Read.

Read is very much in favour of this shift to performance-based advertising. ‘The advantage of performance-based advertising is that it converts ad spend from a line expense to a cost of goods sold. As such, the expenditure is potentially inﬁnite. . . The secret of Google’s success was to convert ad spend from line of business to cost of sale’, Read says.

In other words, the potential income from an ad outlet is much greater than is possible with impression-based advertising. Performance-based advertising obviously represents better value for the advertiser, but it can also represent better value for the seller.

Ben Regensburger, president of DoubleClick Germany, agrees. ‘If you know your audience and your inventory well you can make more money from performance-based ads than simple impressions, especially in ﬁnance’, he says.

The drawbacks of performance-based metrics

Christoph Schuh is CMO of Tomorrow Focus AG, one of Germany’s leading digital content providers. Tomorrow Focus is the largest supplier of German-language content on the web. As a media owner, someone whose income is based on selling advertising, Schuh can see issues with performance-based advertising.

‘The danger in buying performance is that it ignores the value of repeated exposure and of time-delayed responses’, he says. ‘I think performance buying to a single-response dimension will become insufﬁcient; we need to develop behavioural targeting.’

Notice that Schuh is not opposed to performance-based advertis- ing. He simply wants to see performance assessment become more sophisticated.

One of the most common problems encountered when dealing with performance-based advertising is disagreement between the advertiser and the publisher over the numbers. Web metrics systems are still fairly primitive, and the web analytics community has yet to establish clear procedures for measurement. As a result, advertiser and publisher systems can often disagree about exactly how many people have been delivered.

‘Currently, ad people don’t understand the metrics’, says Schuh. ‘This makes resolving disputes extremely difﬁcult.’

Addressing this issue requires training advertising salespeople in web analytics so they have a language in which to communicate and so they understand what it is their web analytics systems are telling them.

In addition, resolving discrepancies between the advertiser’s and the publisher’s numbers usually involves a technical conversation about how the data are processed on both systems. The ﬁeld of web analytics lacks standards, and the few standards that do exist are rarely implemented consistently within analytics software. If both systems are measuring the same thing in the same way, the numbers will match to within a few percentage points.

But discrepancies occur because the two systems are measuring things differently or using the same terms for different things. If the respective technicians explain to each other what their systems are measuring, and how, it is usually possible to adjust the numbers to match. This requires that advertising and marketing people have access to their web analytics technicians, and have the training to be able to communicate with them.

Much of this can be avoided if the methodology for performance assessment is agreed upon before the deal is signed. Once again, this requires that sales staff have sufﬁcient training to participate in such conversations, and that, where necessary, they can call on their technicians for assistance.

The publisher dilemma

For a publisher, performance-based advertising represents both an opportunity and a threat. As Christoph Schuh says, ‘You have to understand your website better than your client. . . You have to understand the behaviour of your readers in the conversion funnel. . . You need an e-commerce unit within your editorial team.’

Once, the editorial focus was purely on producing content that would appeal to a large swath of the population – appeal to enough readers and the advertisers would follow. In the early days of the internet, we thought this was all we needed to do. Jim Barksdale, president and CEO of Netscape until the company merged with AOL, said in 1995: ‘Don’t worry about how to earn money online. Simply get a big enough audience and the money will come to you.’

This was true for a while, but advertisers are wising up. They’re not interested in mere numbers; they want behaviour. This presents publishers with a dilemma. We all know you can’t make money selling content to readers – they won’t pay for it. The presence of huge quantities of free information on the web has devalued the perceived value of all information in the eyes of the online community.

How information is measured

Imagine if you could tell not just how many people were visiting your website (your trafﬁc) but for all individual visitors where they came from (both on the internet and geographically), what browser and operating system they were using at the time, what keywords they used to ﬁnd your site and on which search engine, the page they arrived at, how long they stayed for, what pages they visited while they were there, which page they ultimately left from and whether or not they came back again.

That would be really useful information, right? It would let you analyse how your users were ﬁnding your site, whether you were giving them what they wanted when they arrived, where they were leaving from, whether your site was optimized for the right search keywords, how effective your different forms of advertising were at driving trafﬁc to your site and what proportion of that trafﬁc was ultimately being converted into sales, leads or whatever other conversion goals you’d deﬁned. In other words, marketing gold dust.

Well, the good news is that with modern website analytics software you can track all of that information and more. Like everything else on the internet, web analytics has gone all ‘2.0’, so you can get all of the data you want, presented to you when you want and in the way that you want.

Website analytics uses information your visitors volunteer

It’s important to note here that there’s nothing underhanded about tracking and analysing website statistics. All we’re doing is looking at the information that’s readily volunteered by the user or the user’s browser. There’s no magic ‘voodoo’ or (as a rule, at least) underhanded cyber- espionage going on here. We’re simply collecting the information that’s routinely recorded when a visitor comes to our website and then using analytics software to aggregate it and present it in a format that lets us view trends and make informed decisions.

There are two main ways of collecting information about your website visitors. You can analyse the web access logs created by your own web server or you can embed some code (called a page tag) in every page on your website that sends similar information to your chosen analytics service provider.

Web server log ﬁles

Every time your web server receives a request for a resource (a ﬁle) on your website it stores details of that request in its server access logs. What exactly is recorded depends on a variety of factors, including the way the server itself is set up, the format of the log ﬁles it produces and the settings of the user’s browser.

That said, a server log ﬁle will typically contain the following kind of information for every browser request it receives:

l the unique IP address of the user’s computer;

l a time stamp showing the date and time of the request;

l the URI of the requested resource;

l a status code conﬁrming the result of the request;

l the ﬁle size of the returned resource;

l the URL of the referring page;

l other information supplied by the ‘user agent’ (typically browser type/version, language and operating system).

From this information, website analytics software can derive a host of useful information. For example, from the IP address it can determine where in the world the user is browsing from; the referring URL can tell it whether the user entered the site directly, was referred from a link on another site or came from a search engine and, if so, what search

query string the user entered to reach your site. The IP address may also help track an individual user’s path through the site during the user’s visit (although using the IP address to track this can be unreliable and is typically augmented by the use of cookies).

Log ﬁle analysis software comes in all shapes, sizes and ﬂavours. Some merely read raw log ﬁle data, aggregate it and spit out the results as crude tables for you to trawl through manually. Others format the information to make it more readable and present it in easy-to-comprehend graphs and tables. The most sophisticated let you deﬁne your own summary reports, segment your audience, deﬁne and track conversion goals and analyse conversion ‘funnels’ (the navigation steps a user follows to complete a ‘conversion’ goal), and will integrate data from other sources to give you a complete picture of your website’s overall performance.

Most web hosts provide some form of basic log ﬁle analysis software as part of their hosting package, so it’s worth looking at this ﬁrst to get a basic idea of the kind of information that’s available in your log ﬁles. There is a wide range of software options available, ranging from free open source offerings (like Webalizer, AWStats and Analog) to costly enterprise-level solutions from leading industry players like WebTrends and Omniture, and a host of other options in between.

Page tagging and hosted solutions

The second method is a process known as page tagging. This involves putting a small piece of code on every page of your site that you want to track. Whenever a visitor requests the page the code sends information (gleaned in much the same way as that recorded in the server log ﬁle) and sends it to your chosen provider. This form of tracking has become very much in vogue with the rising popularity of the software-as-a-service (SAAS) concept, and has been fuelled by the introduction of free, powerful and highly conﬁgurable analytics services like StatCounter (www.statcounter.com) and the very popular Google Analytics (www. google.com/analytics).

Once the code is installed on your pages these services will start collect- ing data and can provide you with a wealth of easily accessible and highly conﬁgurable information. You just have to make sure the tracking code is included on every page of your site that you want to track (including new pages you may add over time) and your service provider will look after the rest.

Augmenting information using cookies

Using a user’s IP address to uniquely identify visitors is inconsistent and inaccurate for a variety of reasons, no matter which of the tracking methods above you choose to use. For instance, a large number of internet users (AOL users being the primary example) may share a single IP address assigned by their internet service provider (ISP). That means that, if your analytics solution relies purely on IP address to identify unique visitors, it will count these multiple users as the same visitor, skewing your metrics. Similarly, users are often assigned a dynamic IP address that changes each time they log on to their ISP, so a person returning to your site may be counted as a new visitor rather than a returning one.

To get around these limitations and to remember site settings that help to improve the user experience, many sites and third-party tracking services employ HTTP cookies to identify individual users. Cookies are small ﬁles that are sent to users’ browsers and stored on their local hard drives. Typically they store a unique ID that allows the site (or tracking service) to identify a returning visitor, store site preference and personalization settings that enhance the user experience, and help you to track that visitor’s navigation around your website.

Cookies get a bad press because of the potential privacy issues associated with what are called persistent third-party cookies, or cookies that are set by a domain other than the one you’re visiting and that persist beyond the scope of your existing browser session. In theory these cookies could be used to track visitor behaviour across multiple websites on different domains, building up a picture of users’ behaviour as they surf the web. That’s perceived as a bad thing, because large ad-serving and tracking companies can potentially use cookies to build up proﬁles of user behaviour across all the websites that they serve without explicit consent from the user.

In practice, though, cookies tend to be largely harmless, enable websites to deliver a better user experience to their customers, and allow more accurate tracking of website statistics. The vast majority of internet users’ browsers are set to accept cookies by default, but it’s important to note that some choose to reject cookies out of hand, others accept them only for the duration of their current browser session and then delete them, while others choose to accept ﬁrst-party cookies (cookies originated by the domain they’re visiting) but reject third-party cookies (originated by any other domain). All of these factors can affect the accuracy of your statistics – by how much will depend on the proﬁle of your target audience and their acceptance of HTTP cookies.

As a rule of thumb, if you use cookies it’s best practice to include an entry in your site’s privacy policy explicitly stating what you use cookies for, what information they contain, and who (if anyone) that information is shared with.

The pros and the cons

Advantages of log ﬁle analysis

l Your server will generally already be producing log ﬁles, so the raw data are already available for analysis. Collecting data using page tags requires changes to the website, and tracking can only begin once the changes have been made.

l Every transaction your web server makes is recorded in the log ﬁle. Page tagging relies on settings in the visitors’ browsers (such as JavaScript being enabled), so a certain (small) proportion of visitors may go undetected.

l The data collected in your log ﬁles are your data, and are in a standard format that makes it easy for you to migrate to analytics software from a different vendor, use more than one package to give you a broader view of your data, and analyse historical data using any log ﬁle analysis program you choose. Page tagging solutions usually mean you’re locked in to the relationship with your chosen provider – if you change providers you typically have to start collecting data again from scratch.

l Your log ﬁles capture visits by search engine spiders and other automated bots as well as human users. Although it’s important that your analytics software can differentiate these from your human visitors, knowing which spiders have crawled your site and when can be important for search engine optimization. Page tagging solutions typically overlook non-human visitors.

l Server logs record information on failed requests, giving you insight into potential problems with your website; page tagging, on the other hand, only records an event when a page is successfully viewed.

Advantages of page tagging

l The tagging code (typically JavaScript) is automatically run every time the page is loaded, so even viewing a cached page will generate a visit. Because viewing a page from a cache doesn’t require communication with the server, log ﬁles contain no records of cached page views.

l It is easier for developers to add customized information to page tagging code to be collected by the remote server (eg information

about a visitor’s screen resolution, or about the goods the visitor purchased). With log ﬁle analysis, custom information that’s not routinely collected by the web server can only be recorded by appending information to the URL.

l Page tagging can collect data based on events that don’t involve sending a request to the web server, such as interactions with Flash, AJAX and other rich media content.

l Cookies are assigned and handled by the page tagging service; with log ﬁle analysis your server has to be specially conﬁgured to use cookies.

l Page tagging will work even if you can’t access your web server logs.

Bear in mind here that you’re not restricted to using one type of analytics solution or the other – you’re free to use both as necessary, extracting the best information from each to suit your particular needs.

Figure 5.1 Google Analytics is a very capable, free, hosted analytics service from Google. The service uses JavaScript page tags to collect information about visitors and allows webmasters to track a wide range of metrics to evaluate the performance of their websites

Hybrid analytics solutions: the best of both worlds

A number of analytics solutions on the market, particularly at the enterprise level, have the ability to combine log ﬁle analysis and page tagging methods within the same analytics suite. These hybrid systems analyse all of the data, and consolidate the information to present a seamless reporting solution to the analyst. While these may offer the most comprehensive analytics solutions, with each data collection method compensating in part for the inadequacies of the other, it’s important to remember that no analytics solution is 100 per cent accurate. Most, however, are near enough to the mark to allow you to spot trends and make informed decisions about your online investment.

Other sources of trafﬁc

With log ﬁles and page tagging you can track pretty much everything that happens on your website – but what about the ads you place on other sites, your afﬁliate or ad-network campaigns, e-mail newsletters, or pay-per-click advertising? The good news is that the new generation of analytics software (even the ‘free’ Google Analytics) allows you to ‘tag’ incoming links from all of your external advertising campaigns so that you can isolate and track visitors who are directed to your site from those campaigns.

All you have to do is add the appropriate parameters to the end of the linking URL, and the analytics software will do the rest. Some will even generate the ‘tagged’ URL for you – you just ﬁll in the parameters and then cut and paste the resulting code into your ad. Some tracking solutions, with the help of cookies and page tags, even allow you to recognize visitors who saw your ad on another site, didn’t click on it, but some time later decided to visit your website anyway. Powerful stuff.

proﬁle of their sites and perhaps to attract the brave pioneers of online advertising to buy space on their site (ﬂashing banner ad, anyone?).

The trouble with ‘hits’ as a unit of measuring page popularity was that it was completely unrepresentative of the number of unique visitors to a site. A ‘hit’ represents a single resource request from a user’s browser (or other user agent) to a web server. For a basic web page containing 10 images, the browser initiates one request for the page itself and 10 separate requests for each of the image ﬁles on that page, or a minimum of 11 ‘hits’ in all – more if there are other design elements being pulled from other ﬁles on the web server. Different pages therefore generate a different number of ‘hits’ depending on their make-up, which means the whole concept of using ‘hits’ as a comparative unit of measure of page popularity is inherently ﬂawed.

Thankfully, while ‘hits’ are still recorded in web server log ﬁles (and you’ll see the individual records if you browse through the raw text of a server log ﬁle), modern web analytics software aggregates the data into far more useful metrics. Today you’re far more likely to hear about ‘page views’ or ‘page impressions’ (which mean pretty much what they say – a complete page delivered to a user’s browser), or ‘visits’.

With modern software you can automatically collect, collate and compile the information being volunteered by your users’ browsers and then use sophisticated graphical interfaces to analyse trends, present the data in a range of different ways, and conduct comparative analysis that can help inform your digital marketing investment. ‘Hits’ may well be dead as a unit of measuring website success, but take the information revealed by your web analytics solution, use it to understand your online market, and take action based on what you’ve learned, and your website itself is sure to be a big hit!

Measuring what’s important to you

While web analytics allows you to have your ﬁnger on the pulse of your website, to monitor the beating heart of your online audience, you can easily get overwhelmed by the sheer volume of information available to you. You want to spend your time focusing on what’s important to your business, not wading through reams and reams of information that may offer interesting insight but does little to further your strategic goals.

You need to determine what exactly you need to measure. What metrics are important to your business? What will help you extract maximum value from the visitor information available? If you can’t answer those questions, how can you make the strategic decisions that will drive your online business forward?

The concept of key performance indicators (or KPIs) is nothing new, and has been common in the world of business analysis for many years. KPIs are used to distil key trends from complex, often disparate pools of data and to present them as a series of clear, unequivocal indices – a snapshot of how your organization (or website, in our case) is performing at any given time. KPIs do ‘exactly what it says on the tin’: they indicate progress (or lack of it) in areas that are key to your website’s performance.

Why KPIs are important

The real value of KPIs is that they let you extract meaning from your website data at a glance. Without them, it’s all too easy to drown in the proliferation of data your web analytics solution churns out. It’s a classic case of not seeing the wood for the trees.

By deﬁning and measuring your KPIs you’re creating a regular snap- shot that allows you to monitor your website’s performance over time. You know that if this KPI is going up it means one thing, if that one’s going down it means another, and so on. Your KPIs not only give you an immediate sense of the overall health of your website, but also help to highlight potential problems and point you in the right direction before you delve deeper into your data looking for solutions.

Choosing effective KPIs

The main difference between the metrics you select as your KPIs and all the other metrics you can get out of your web analytics software is that the KPIs should be the ones most critical in measuring your site’s success.

In their document Web Analytics: Key metrics and KPIs (G Creese and J Burby, Washington, DC, 2005), the Web Analytics Association (WAA) deﬁnes a KPI in the context of web analytics as:

KPI (Key Performance Indicator): while a KPI can be either a count or a ratio, it is frequently a ratio. While basic counts and ratios can be used by all Web site types, a KPI is infused with business strategy — hence the term, ‘Key’

— and therefore the set of appropriate KPIs typically differs between site and process types.

Another thing to note is that the terms ‘KPI’ and ‘metric’ are often used interchangeably. This is misleading because, although a KPI is always a metric, a metric is not necessarily a KPI. So how do you tell the difference?

l KPIs are always clearly aligned to strategic business goals.

l KPIs are deﬁned by management: Decision makers have to identify, deﬁne and take ownership of the key drivers of their organization’s success.

l KPIs are tied to value drivers critical to achieving key business goals: They should represent the ‘deal breakers’ in the pursuit of your organizational goals.

l KPIs need to be based on valid data: You only get out what you put in.

l KPIs need to be quantiﬁable: You have to be able to measure your KPIs in a consistent and meaningful way over time.

l KPIs need to be easy to understand: They should be a barometer of your site’s performance – a quick glance at your KPIs should tell anyone in your organization, from management to trainee, how well your website is performing.

l KPIs can be inﬂuenced by, and used as triggers for, positive action: One of the main values of KPIs is that they immediately highlight where your organization ‘could do better’ and highlight areas where action is required to get things back on track.

Some generic web-based KPIs you may ﬁnd useful

l Conversion rate

l Page views

l Absolute unique visitors

l New vs returning visitors

l Bounce rate

l Abandonment rate

l Cost per conversion (CPC)

Search This Blog

MyTutorials4U

Digital Marketing - unit 3