“Set up the measuring and we will base our decision upon its outcome.”
Logical, isn’t it? The trouble is that single numbers must be read within a certain context, so they can actually make some sense instead of just being confusing.
The majority of the online tools which are able to monitor the behavior of a website’s users offer some wonderful tables and occasionally some graphs. Thus we find it senseless to pay someone to analyze this data and to deduce the consequences and make conclusions.
Why does this not necessarily have to be true? See the examples below.
Causes and consequences
Have you ever heard:
”Nobody comes to your website from a mobile phone. There is no point in redesigning the website for mobile devices.“
Could that be, by any chance, the other way round?
”People using a mobile browser who come to website with hard-to-read, tiny text and small buttons are more likely to stop using that website and switch to a similar one which will work much better on their phones. It is also probable that the search giant Google will tend to post in a search the mobile device-friendly pages first. “
And there are more cases when people confuse the cause with the consequence.
Correlation does not mean causation
A study by Searchmetrics showed that the most popular websites are often mentioned on Facebook, they either have many likes or multiple shares (or both). However, that does not necessarily mean that the number of likes or shares determines what position a website will have within a web search (Google, Bing etc.).
One can see that these two things (in this case: the number of likes and the position within a web search) do not have to be a cause and a consequence of one another. Actually, both facts’ causes may be the same. A high quality website can be shared/liked by many users spontaneously, whereas the position within a search results page is given by numerous ranking factors.
A/B testing and the chance
Put simply: A/B testing is a process whereby the people visiting a certain website are put into two groups. Each group is shown a website which looks slightly different and we make a note of what effect this change has on the number of responses (conversions) made (e.g. ordering a brochure, signing up for a newsletter etc).
Let’s say we have 2000 total page views (and thus each group has 1000 page views). On the original version of the web page 40 people signed up, whereas on the slightly redesigned version 50 people agreed to receive the newsletter.
Good, so that is a 25% increase, right? No, what if it was only a chance coincidence? We have to consider the statistical significance. In other words, the difference in the response to the website is so insignificant that if we repeat the experiment once more we might get a completely different outcome.
Narrowing the focus
“Please, see the list of keywords attached and create a table based on the number of searches.”
If you are a professional in a certain area, you are logically familiar with numerous terms which the general public has never used.
If the majority of your customers consists of the general public, then you are better off using simple language. Professional terms or jargon may be put in brackets (parentheses in American English) or mentioned in a special section for professionals in your field. A few practical examples:
- Parity instead of equality (in economics)
- Onset of a diseaseinstead of start, beginning (in medicine)
If you expand your horizon, you might find out that many visitors are searching for some more commonly used phrases (keywords) which you, as a professional, may hardly ever use.
External influences
“A month ago we changed the website and since then we have seen a 20% increase in the number of visits and sales. The impact of this change has been more significant than we expected.”
Could not the reason be that it is Christmas time, and so the overall demand is rising?
I am purposely showing a basic example. There can be multiple causes which may affect the result of changing a website.
- Seasonality of goods or services (need for an annual comparison)
- Online campaign (should be easy to recognize)
- Offline campaign (slightly difficult to analyze, but still possible)
- Brand new products (e.g. MP3 players replacing the old discmans)
- Boom, a sudden increase in demand (e. g. you may benefit from a rival’s advertising campaign)
Averages vs. segments
“The average length of a website’s visit is 2:54 minutes.”
or
”During one visit an average web visitor usually views about 3.25 pages.“
Is that too much or too little? What do these figures really show you?
Talking about averages, it is always difficult since they try to relate different types of visitor´s behavior. Just one group with unusual characteristics can distort the outcome.
It is more interesting to ask, for example:
- Is there a significant difference in the views per visit of visitors who have been attracted by a campaign (compared to other sources)?
- What is the average time spent viewing a page in the Product section compared to reading the Blog (both on the same website)?
The information related to a certain segment of visitors allows us to compare and at the end of the day it will all make better sense.
It is always good to ask for a website user’s aim in order to get the right answer within a given context.
- Did the customer who was searching for the company’s address go straight to the Contact page and then decide to leave? Good. However, the percentage of people leaving the website has immediately increased.
- Has our blog reader only seen one article? What can we do to attract him to read more? – Maybe we could suggest some related articles?
In order to get the right result it is necessary to use the right calculation method. And thus, a quiz question to conclude:
How does Google Analytics calculate the time which the user has spent on a website if he/she has only visited one page?