Metrics for Non-Transactional Web Sites
How do you measure success or improvement on the vast majority of Web sites where there's no transaction or direct financial benefit?
I have no idea whether this is true or not, but I once heard someone say that 80 percent of all Web sites don’t sell anything. Yet, probably 80 percent of all Web analytics examples and case studies are based on transactional activities such as buying or subscribing to something. You can see why. It’s much easier to demonstrate performance improvement or increases in return on investment when there hard dollar signs are attached.
But how do you measure success or improvement on the vast majority of Web sites where there’s no direct financial benefit? Take, for example, lead generation or public sector sites. How do these site owners know if they’re doing a good job and spending their money wisely and optimally? This is an increasingly important area for the Web analytics industry to get its head around as more and more of these types of Web site owners are looking for ways to demonstrate value from their digital marketing investments.
As with any process in determining success, it’s important to be clear about the objectives. Why do you have a Web site in the first place? What is its role? Then, think about what “good” looks like. What are the characteristics of visitor behaviour that you consider to be valuable on the Web site? Finally, can you measure these behaviours, directly or indirectly?
Let’s take an example based on an experience I’m going through at the moment: buying a car. I’m assuming one of the main objectives of a car manufacturer’s Web site is to sufficiently convince me that they have the right car for me so that I’ll take one for a test drive. I therefore must feel good about the car I’m thinking of buying (a branding effect), find the information to help me decide as to the right model and specification (a search process), and then be able to contact the appropriate dealer so I can take one out for that test drive. A number of different processes are going on which, of course, may take place over multiple visits.
It’s possible to take this typical visitor behaviour and characterise what “good behaviour” looks like. This may be the fact that I looked at certain pages, but it’s more likely that I did certain things in certain sequences, such as downloading a specification sheet and then using the dealer location tool. In trying to understand the effectiveness of the site, it’s likely outputs that report on and analyse sequences of activity are going to be more useful and generate more insight that ones that report binary activity, such as whether someone visited a page or not.
Just observing behaviour doesn’t give you the complete picture. The fact someone visited a specification sheet doesn’t tell you whether someone found the information they were looking for or not. In my car buying example, I was looking for the capacity of the boot (or trunk) of a certain model of car to compare with some other models. I clicked all over the place, spent loads of time reading certain pages. In many ways my behaviour might have been considered valuable with many pages viewed and long dwell times on key pages. But I never found the information I was looking for.
If there’s to be more progress in the measurement of success of non-transactional Web sites, then the role of surveys and other visitor feedback devices must be strengthened as part of the overall mix. The success of a visit where there’s no transaction involved needs to be defined almost on the visitors own terms. The only people who can tell you whether their visit was successful or not are the visitors themselves.
Non-transactional Web sites are generally harder to measure. Recent debates in the Web analytics world are around the development of new metrics and strategies for measuring concepts such as “engagement.” It’s great the debate is going on. We also need to step outside the box and recognise that the behavioural data will always have to be interpreted in the context of visitor perceptions and feedback. I suggest that organisations that are looking to measure success on sites that have little or no transactional activity should be looking to invest as much time and effort in developing a consistent and managed programme of visitor feedback as they do in getting their behavioral data sorted.
