Brisbane Web Analytics and conversion consultancy | Metricks


Don’t install Google Analytics yet – read this first!

by Jon

This isn’t a plug for a rival web analytics tool, more a recognition that whilst it is tempting to install a popular web measurement tool and be blown away by big numbers, fascinated by bizarre keywords and mesmerised by pretty graphs it is more important to first develop a measurement strategy.

Recognise this scenario?

Person with huge workload and responsibility “Hi we had a campaign last week, can you tell us how many hits the website had?”

Person responsible for monitoring web analytics tool “I’m afraid we no longer measure hits, but I can tell you 1235689 people visited the site and viewed 1515151 pages. Of these visitors, 53% came from search. 93% of these were from Google, with 17% of traffic from the keyword “awesome thingymajig”, whilst the biggest referrer was massivehugereferrer.com and the main component of direct traffic was thought to be the CEO’s weekly broadcast email that goes out every fortnight. Of the 70% of new visitors, 83% were from Australia, 10% from the US and 3% from India…is there more data I can give you?”

Person with huge workload and responsibility “Er, that’s great, but no thanks!”

Although measurement of websites has matured, there still exist disconnects between the figures the business are interested in, and reports generated from the tool of choice in your business.

Therefore it is important to establish a measurement strategy that puts your web data in context with all other business activities. Some of the points to consider include:

• Getting in early – planning stage!
• What are the goals? You may only be able to provide intent
• Context around measurement – past, present, future
• Use business areas language – avoid death by percentages
• Recommend action – otherwise do not report
• Slice and dice the data – segments, audience, dates

It is far easier to get measurement right when all variables have been planned for, implemented and collected rather than the above scenario where briefs have not included evaluation, goals or targets. Trying to evaluate significance once an event that required measurement has occurred is pretty tough. It is normally possible to present aggregate figures, but not the really specific recommendations that can be seen in an analysis of planned measurement.

Establish what the goals are of the activity. For a new website, what is it for? Is there cost data you can use to measure success against? Having goals establishes a baseline with which you can compare all activity on your site. Goals don’t necessarily have to be the end point either, a goal on a website could be phoning up to place an order – in which case an number should be provided that is unique to the site, and some facility added to track from site tom phone (an order number for example).

Context is essential for painting a picture around the data. Try this example:
You receive an Excel report with three sheets, with a single column titled visits and a number on each sheet
sheet 1 – 20 visits
sheet 2 – 100,000 visits
sheet 3 – 5,230,600 visits

Which is the best? Without context it is difficult to decipher (though odds on the biggest figure would be chosen as a measure of success). For the purposes of this example, add an element of time. Sheet 1 is visits in the last second, sheet 2 the last year and sheet 3 since the site was launched in 1996. Then consider that sheet 1 is a site that only has a target audience of 40 (very exclusive!) people worldwide, and these are separate visits. Sheet 2 is a major worldwide shopping site. Sheet 3 is from a popular search engine in the 90’s that no longer functions and this data is 10 years old!

A much different picture emerges than was initially realised. A useful tip is to document significant events and also be in touch with the people who have an interest in your site and other channels to avoid being blindsided by unexpected changes in the data.

In the initial scenario above, the Person with huge workload and responsibility was completely bamboozled by the long description full of percentages. Rather than rely on top 10 tables and torturing sentences into accepting far too many percentages than is good for them, have a strategy where the data is interrogated to answer business questions, but the answer is provided clearly and in the most appropriate format. For example the question “how much cash has the website generated” would ideally simply require an answer with a number (preferably big) and a $ in front of it. Even though it is tempting to provide a 20 page powerpoint referencing a 50 sheet Excel document with reams of calculations and pivot tables in it, don’t do it. Of course document all calculations, but in the end try and provide the clearest, most undeniable answer.

Recommendations should be approached in the same way – based on the data, but not necessarily requiring the reader to wade through everything. They should be contextually based, consider all angles and consider the end reader. (for example, the manager of the call centre might not be too happy to read an analysis that recommends all level 1 call information be written up as FAQs on the website, as that could negatively impact the managers budget – instead highlight how this would enable call centre staff to concentrate on more complex calls and handle online chat)

Once these factors are discussed planned for and requirements for what needs measuring have been developed, you can then go ahead and install Google Analytics with a much better idea of what functionality is required – the above would give you:
What constitutes a goal, and its value
Whether Ecommerce code is required (tip: use it anyway on goal pages)
Information to use in notes within the application for significant events
Any filters, segments and code customisations that are required.

This will give a much more robust Google Analytics implementation that provides website data that is quicker and easier to get to and analyse.

Then all you have to do is combine with data from search, social marketing, email, campaigns etc to provide the answer to life the universe and everything. Well not quite, but that’s another story…

No comments on ‘Don’t install Google Analytics yet – read this first!’