6: Defining Site Goals, KPIs, and Key Metrics

This chapter goes into greater details about determining the goals of a site and defining the metrics that can be applied to reflect its effectiveness in achieving those goals.

Defining Overall Business Goals

Web analytics should not be an abstract exercise, done just for its own sake. It is a practical task, done to understand the way in which people interact with a specific Web site, and that site is operated by a business with specific purposes in mind. While there was a time in which a company would launch a Web site "just to have one," their approach has since become less random and more mercenary: the site must accomplish goals for the business. And as such, the task of the analyst is to demonstrate how well the site is doing in terms of accomplishing those goals.

It's suggested that the failure of sites during the dot-com bust stemmed either from sites that had no specific goals, or sites whose goals were misguided. A "pure Internet" company often had no goals that would constitute a sustainable business model, and even traditional businesses considered their Internet operations to be a separate matter from their "real" business. Those that survived the bust knew, or quickly learned, that a Web site must have goals, and must be run as a business, or it's not worth doing.

As such, the goals of the Web site must derive from the goals of the firm. They should also be consistent with the ways in which the firm seeks to accomplish its goals in other channels, and should have the same temporal horizon (the site should not be restricted only to short-term goals such as a sales promotion), in order for the site to support the business is a meaningful and sustainable fashion.

If the company's goals are vague or ill-defined or badly communicated, that is a problem, and quite a serious one, but beyond the scope of Web analytics. Chances are that a company without a clear plan has much bigger problems that cannot be fixed by technology.

Defining Site Goals: The Conversion Funnel

Once the goals of the business are understood, you can seek to identify ways in which the Web channel can support them. The author refers to the "sales process" as four steps - awareness interest, consideration, and purchase (EN: How this is different from the AIDA model of marketing is beyond me, sounds like a rip-and-tweak) - and consider whether the Web might be useful - it may be all stages, or it may be just one of them.

The author explains a bit about the AICP model in a bit too much detail, and identifies ways in which the Web can contribute to each: Online advertising can create awareness, a multimedia campaign can be used to generate interest, information provided on a site can help in consideration, and an online "store" can help in making an actual purchase.

(EN: This discussion is myopic, focused entirely on product sales. I expect that this is meant to be used as an example, though the author does not say as much, nor are any other examples provided. It is also quite shallow, in that it perceives "the Web" to be no different from broadcast media in which companies control the message. That's not to suggest the advice above is invalid - its' fairly good - merely that it's limited in its depth and breadth of vision and the failure to mention this explicitly might be misleading by its omissions and implications.)

The conversion "funnel" assumes a diminishing audience - a large number of people will see an ad, some (not all) will click through, some (not all) will enter a sales flow, some (not all) will purchase - and the "not all" represents a number of customers who will drop out along the way. This also assumes that a greater percentage of users will remain on track (and fewer will drop out) if you can make improvements at any given steps.

(EN: The funnel model is broken. There are various minor criticisms that seem to be splitting hairs, but the one that really crushes the funnel theory is the fact that it is based on the assumption that moving a person further along the funnel makes them more likely to buy at the end - the counterargument being that it merely moves a person who will not buy further along the process. He feels no greater commitment to purchase and, ultimately, is all the more annoyed if it takes him longer to discover that he doesn't value what you're trying to talk him into buying. So the danger in pushing non-buyers through a funnel to make a short-term sale is in the long-term and more permanent damage it does to the reputation and credibility of your firm. The non-buyer may be a not-today buyer - and not only have you galvanized him against ever coming back, you've given him incentive to dissuade other would-be buyers from engaging with you. This is another instance where short-term thinking can be detrimental to long-term success.)

Understanding Key Performance Indicators (KPI)

Once the general goals of the site are defined, you can then approach the task of defining the key performance indicators (KPI). The author acknowledges that the KPI method of organizational management cane into fashion, then suffered a great deal of criticism, but implies that the problem is no inherent to KPIs, but to the specific way in which they are defined. A bad measurement yields bad results, but does not invalidate the value of measuring.

It's noted that KPI tend to be tied to long-term goals rather than short-term; are specific to an organization rather than standard for an industry; are measured repeatedly over time rather than just once; and are on an organizational rather than departmental level. There are other definitions, and the notion is often abused in practice (not every metric is a KPI).

The KPI are derived from organizational goals - the mission statement - and to be meaningful, they should have a fairly simple and fixed calculation (it does not change over time) that is based on observable and reliable numbers. Many of them tend to be financial in nature (keep expenses to X% of revenues) though some can be attitudinal (maintain a customer satisfaction "good" or better for X% of customers who respond to the after-sales survey). Goals for business units, departments, and individuals can derive from the KPI for the organization (in order to maintain customer satisfaction, the shipping department must process 95% of orders the same day they are received).

A KPI is generally guided to reach a specific target - but in order to be effective, it cannot be an arbitrary one. You should consider the current level of performance, potential variations, to ensure that the goals are achievable - nobody is motivated to attempt a goal that is too lofty, and if a goal doesn't motivate behavior, it serves no purpose. While the core of the KPI should remain constant (what it is designed to measure, how it is calculated), the target of a KPI can be adjusted over time, up or down, based on what is a reasonable expectation.

(EN: A note here on progression is that many KPI seem to be ever-increasing. If a company hits the target, the target is increased. This practice results in diminishing returns - improving from 95% to 96% customer satisfaction costs more than it generates in revenue - as well as dysfunctional behavior as the level becomes less and less feasible given limited resources. Ultimately, the question to ask is "how good is good enough" - and more companies set the bar too high than too low, which wastes resources, harms employee morale, and may even run off customers.)

Common KPIs for Different Site Types

The author goes into laundry-list mode, showing specific KPI of various kinds of sites. He concedes that these are archetypes, and it's unusual for a company site to be focused exclusively on one function to the exclusion of all others, so it may be necessary to adopt KPI from various categories

For an e-commerce site, in which the customer is meant to purchase items online, the KPI pertain to the activities related to the purchasing:

For a lead generation site, the customer does not purchase an item, but instead provides contact information that can later be used to close the sale. To some degree, sites that require the user to sign up for an account might also be subjected to the same metrics.

A "customer service site" can be defined in any number of ways - it may be getting technical support for a product, servicing an account (checking bank balance), or any activity for which the customer would have otherwise contacted the company for assistance. The value of such a site is presumed to be the cost saved versus having the same volume of customers call the company (which is a much more expensive channel).

(EN: The "value" of service sites is difficult to measure, and to defend metrics. It can be argued, for example, that many people who use the Web would not have bothered calling and figured it out on their own; and then counter-argued that a stymied customer is a damaged relationship and loss of future sales, etc. It's slippery and subjective.)

Content sites deliver information to site visitors by means of text, image, videos, etc. and derive revenue either from advertising or subscription fees, so some of the KPI are relevant to one more than the other.

"Branding" sites are those the "involve experiential content" that is designed to give the user a positive impression of the company that may eventually lead to purchasing. As the goal is to effect a changer in attitude, it can be difficult to measure in terms of behavior, though given that the goal is exposure, the level of engagement can be considered a success indicator. Hence:

(EN: the author has suggested some of the KPI that might be used for different kinds of sites, but the enumeration above very limited and superficial, and some of these measures are arguably not the best metric or the best approach to generating a number that represents a desired behavior. As such, consider the above to be examples, and keep in mind that more consideration is necessary to determine what metrics are appropriate to a given site.)