Solutions Architects on the Performance Specialist Team at Akamai closely work on performance analytics, we do a lot of value confirmations for our customers we show how Akamai Web Performance Solutions benefit our customer's websites. As part of our job, we get a lot of requests from customers who want to compare how Akamai stacks up against their current content delivery network (CDN) infrastructure as well as that of other vendors.
In our role we tend to see a lot of customers using Google Analytics to measure website performance and although Google analytics is a great tool to collect data and information about your website, used by itself, Google Analytics may not be the best judge of web performance.
Let's take a closer look at why we believe Google Analytics alone isn't enough to get the full picture regarding website performance.
Working with Google Analytics Site Speed
The Site Speed module of Google Analytics helps website owners understand how quickly users are able to see and interact with their content; identify areas of improvement; and then track the extent of those Improvements. But every time we make an observation, we have to be aware that the network conditions, testing locations, and browsers are different.
In order to make the most accurate comparison, we have to make sure that the runs happen in parallel, from the same location and from the same browser.
However, there is a limitation where we cannot conduct A/B testing.
Ideally we should never compare past data with present data, if we are comparing our website's performance then the network conditions, duration, frequency, test locations should be same.
Below are some points which need to be considered while working with Google Analytics:
1. Limited Sampling - Google takes a 1% sample from the complete traffic. For example if we have approx. 10000 requests then only 1% of 10000 (i.e. 100) would be considered for reporting. Surely 1% sample points cannot represent the complete traffic pattern, if your site attracts huge traffic. Although customization is available, but it requires technical know-how. In order to have a fair comparison make sure your network conditions, agents, frequency and platform is same. With that kind of sampling rate, we cannot ensure the same networking conditions or platform.
2. Everything is average - As you know outliers can influence the average and GA site speed always present the average stats. Hence this adds to the inconsistency.
3. Validate the data - The best way to validate the data is to download the raw data and analyze the page load times. Make sure no failed data points are included in the sample, if you find some data points reporting 0 time then obviously those points will influence the average, and hence should be discarded from the calculation. Moreover it is not possible to drill down to a data point level and see the waterfall charts, and there is no way to apply filters based on domains. Analysis is very limited.
4. Round off inconsistency - Google Analytics uses seconds as their unit and has only 2 digits after the comma, especially for the smaller timings (Redirection time, Domain lookup time, Server connection time, Server response time, page download time). This can result in significant rounding errors (on top of low default sample rate).
5. Profiling of data - It is very important to make profiles based on desktop, mobile, geographies, bandwidth etc. in order to understand data in a better way, because the test conditions should be same.
Conclusion - Google Analytics is a very good tool to understand the basics of your website performance. But, if you want some in-depth analysis or want to enable alerting mechanism when the site breaks, then come talk to us. The best way to reach out to us is via your sales rep. Akamai Web Performance Solutions include some awesome tools which can help you truly understand your website performance.
Written by Reeti Verma and Shashank Bhardwaj