Engage-Digital

Lately I have been coming across a series of small but potentially significant shortcomings in the web analytics tools I have been using. I only unearthed them as I was playing around with different ways of getting the same output, in other words sense checking the data I was looking at.

The trouble is if I hadn’t done that I might have taken the first answer as gospel and moved on, unaware that the insight I was inferring was based on incorrect data. I’m talking about the kinds of things that are genuinely obscure and might not impact that many users but for those that do they could make a big difference.

Given that the industry’s mantra is broadly ‘rubbish data = zero insight and worse, possibly damaging action taken as a result’, it seems as though the suppliers of web analytics tools have a responsibility to document and make freely available the shortcoming of their products especially where those shortcomings could have a serious impact. Of course the trouble that would be suicide.

It’s a catch 22, keep quiet and hope that you don’t do too much damage to your client’s site’s or take a more responsible standpoint, ’fess up and deal with implications that come with it.

Bounce rate is a kind of standard bearer metric for measurement in web analytics, it’s up there with conversion. Even novices in web analytics know what bounce rate is, and when asked what should be the objective regarding bounce rate you’d have to be a lunatic to say anything other than “try to reduce it”. But increasingly, I think there are some misconceptions about this metric. Remembering that bounce rate applies to both entry pages and referring sources of traffic, two thoughts that come to mind are as follows:

1. When thinking about bounce rate in the context of entry pages it is hardly surprising that it’s lower on the home page than on a product page. I’ve seen very few sets of data in which the situation is reversed. I think this is because the home page offers a bigger target, i.e. there are more options for a visitor when they arrive on the home page than if they arrive on a highly specific product page. Trying to reduce the bounce rate on a product page is worthy and will yield results for sure but don’t expect to get it down to home page levels.
2. Lower bounce rate = more actual conversions. I’m not disagreeing with this but increasingly as I look at weekly and monthly trended data for various clients I see examples where BOTH bounce rate and conversion actually go UP against a particular referral source. This in turn quite often results in higher yield volumes as well.

This isn’t to say that it’s OK to let the bounce rate metric rocket up, retaining more visitors at the same conversion rate WILL of course produce greater conversion yield volumes. It’s simply to say that it’s not necessarily worth freaking out the minute a bounce rate shows signs of edging upwards. I would always look at the surrounding metrics to build a bit of additional context before deciding on a course of action.

The old adage ”quality not quantity” comes to mind here. A higher bounce rate might mean lower visitor retention but if more of those visitors are converting and the overall conversion yield is going up then that might prompt another old adage “if it isn’t broke….”

I see that Eric Peterson has written a nice white paper on what he refers to as ”The Coming Revolution in Web Analytics”, it can be downloaded  here. In it he talks about the future of web analytics and in particular what he calls 3rd generation web analytics.

I won’t paraphrase what he says about 3rd gen WA (really better to read the white paper) except to say that he no longer describes it as  web analytics as it starts to move into the realm of general business and customer intelligence.

Among the thoughts that occurred to me whilst reading his white paper, two stood out:

1. What he says about 3rd generation web analytics / business / customer insight at a practitioner level sounds a lot like the kind of thing econometricians have being doing for years although initially they weren’t including heavy weight web data. But the kind of modelling he talks about is already being done by econometrics units in many media and advertising agencies.
2. What he describes as being 3rd gen (web) analytics is no doubt advanced stuff but I think the problem will be uptake. I think that there will be a small cadre of trail blazers who will get heavily into all the mechanisms he outlines such as cross channel data mining, predictive modelling, confidence analysis of customer segmentation etc. but the trouble is this will be for the big guys who can afford it and even then precisely because it is complex it will probably only be taken up by organisations who have the right people in the driving seat. i.e. despite having a highly capable team of analysts and statisticians, a C level director who’s head is filled with many things may find it hard to champion this stuff unless s/he has a reasonable to good understanding of it his or herself. For the rest of the world – i.e. the majority of small and mid-sized businesses it’s still a long way off.

None of that is to say that A) Peterson isn’t correct in his overview or B) it isn’t what should happen, just that the main obstacle may be one of human resource more than technology, much as it has been up until now.

Search engine optimisation is seen by many as the Holy Grail of online acquisition. If done properly it’s considered to be the cheapest and most effective way of driving high quality traffic to a web site. Many people obsess and probably loose sleep over their favourite search term(s) reaching the cherished number one spot in Google’s “organic” rankings. A tweak in Google’s algorithm can have a significant impact on some small and mid sized businesses. When Google sneezes others get the flu.

Understanding the impact of organic search is the first useful step to improving rankings and driving increased volume of organic traffic. All web analytics tools provide reports which show the top referring search terms and phrases but in most cases these reports will need some configuration before they become useful.

To begin with if a pay per click campaign is being run it’s normally necessary to split out paid for search traffic and organic search traffic. This is done in most cases by setting up the paid for search traffic as a campaign in your analytics tool. Most tools will then automatically do the rest of the work for you by splitting out the two sources of traffic and dumping them in separate reports.

From here it becomes possible to view which organic search terms and phrases drive the greatest volume of traffic, which are the best converters, which pages they drive traffic to and what the bounce rate is. Being able to directly compare organic and pay per click in this way helps highlight gains that can be cross pollinated between PPC and SEO.

Most sites have a few terms and phrases that drive the core of the organic traffic but they also have a longer tail of terms and phrases that when combined form a significant but often ignored source of high quality traffic. Exporting this data means that these little gems in the long tail can be wheedled out based on their high conversion rate and nurtured until they make a greater contribution to overall volume.

Other measurement tools such as Hittails and Advanced Web Ranking will help in identifying search terms to focus on and monitoring actual positioning in search engine rankings.

Finally, due to the long term nature of search engine optimisation patience and trended data are key to happiness and enlightenment. It’s unusual to move up the search engine rankings and drive more traffic to a site the day after making a few changes. Quite often it can take months to rise up or fall from grace.

IMPLEMENT AND CONFIGURE YOUR ANALYTICS TOOL PROPERLY

Bad data is worse than no data at all. Decisions made using bad data can have a negative impact on performance and revenue and that can often be worse than sticking with the status quo.

Thanks to Google and Microsoft a wealth of potentially valuable data about how visitors interact with your site is now available for free. Implementing Google Analytics or Microsoft Gatineau takes little more than ½ an hour and from the moment they are implemented they will start collecting data about traffic to your site including a range of useful metrics to help guide the decision making process.

However, to get the best from these web analytics tools they should be correctly configured post implementation. This involves filtering out unwanted traffic such as your own visits which can skew the data; setting up campaign tracking to understand which elements of acquisition strategy are the most cost efficient; setting up funnels which will identify which levels in the visitor journey through your website have the highest attrition points – knowing this helps focus attention on the most critical points and reduces the chances of wasting precious budget on design updates made on the wrong areas. Additionally setting up internal search tracking is a key part of understanding what visitors are specifically looking for when they arrive on your site, it cam also help inform you pay per click marketing strategy again reducing wastage.

Web analytics data is often the starting point from which wider analysis stems so making sure configuration of your web analytics tool is as good as it can be is critical.