At first glance, Google Analytics looks like a great tool to measure speed. It seems that it displays the actual loading times your users experience. But it can actually give you misleading data depending on your circumstances.
Sample size
Google Analytics by default samples just 1% of visitors to your shop to measure their speed. If you have 1,000 visitors a month that's just 10 that'd be measured.
This means that for it to be able to answer the question 'how fast is my shop to users?' the number of users in the sample should be statistically significant.
To put it another way, if you ran a coffee shop and had 100 customers in a day you ask just 1 if they were happy. If that 1 was unhappy can we say that the other 99 were unhappy too? No. Increase the sample size (the number of customers you ask) and you begin to get a true picture of how happy customers are.
Shops without a lot of visitors will not have a large enough sample size to use Google Analytics to measure speed.
Data lag
If you make a change today to improve your speed it will not show in Google Analytics straight away. Wait at least one day until you start to gather data.
Then wait until you have gathered enough data to be statistically significant. Similar to the 'sample size' problem, you will need to wait until enough users have been measured to paint an accurate picture of your shop speed.
Factors out of your control
If a user without a broadband internet connection browses your shop and it takes 30 seconds to load your page, Google Analytics will show 30 seconds. The majority of their loading time was not your shop code, server, images etc. it was their own slow internet connection.
Browsing Google for that same user may take 30 seconds, all due to their own internet connection.
These outliers are very common with a small sample in Google Analytics but are smoothed out when taking a large enough time period and number of users into the measurement.
When to use Google Analytics to measure speed
The factors above mean that Google Analytics is:
- Not suitable if you don't have a lot of unique visitors
- Not suitable to measure a before/after speed optimizations
- Suitable for watching longer term trends (month to month) if you have enough visitors to create statistical significance
Comments
0 comments
Please sign in to leave a comment.