Analysis of Browser User/Market-share, or So Long! IE8, 9 and 10
Developers love to create new functionality about as much as they dislike the tedious work necessary to get web apps to work well in a wide range of browsers. This work is an expensive time sink. It can slow down the release of valuable features, and it’s a source of frustration that affects your ability to attract and retain top developer talent. So, development teams are constantly asking if they can stop supporting older browsers. The push back from the business is that shortening the list of browsers can shrink the available pool of customers.
There is some good news. First, all browsers IE 9 or newer have significantly better compatibility and performance. Second, the most relevant good news on this front is that Microsoft has decided to end-of-life support for IE 8, IE 9, and IE 10 (with a few exceptions for a few long-term-support versions of Windows). Note, those few exceptions account for less than 0.01% of Pendo event data so I’ve decided to ignore them in this discussion.
This announcement by Microsoft triggered some interesting discussion at Pendo, so we decided to post about it. I had previously written some custom analysis of browser usage as a Pendo customer, so I just adapted that and ran it against events recorded in Pendo. I have published a number of interesting analyses in the agile team development world and this is the first such publication in the space occupied by Pendo clients. I’m sure there will be more. Send me any questions that you would like to see answered across the web app usage landscape, and I’ll address them in future posts.
First, let’s take a look at the overall browser mix comparing the general web site browsing/shopping usage represented by NetMarketShare.com versus the web app usage that is represented by the Pendo data.
The Pendo data comes from over 100 Pendo accounts that had at least 90 days of relatively stable usage. The filter was designed to leave out internal testing usage, or early ramp up that might not be indicative of the steady state usage. Because we have a huge volume of event data, I only looked at the last 7 days of Pendo event data. I found this to be so stable over that time period that a larger time period would yield nearly identical results. A better approach would be to sample a fraction of events over a wider period and trend it over time. I’ll leave that for a follow up post on trends in browser usage.
The first thing that jumps out at me is that IE accounts for 50% of the general market share, but only 31% for web app usage (all Pendo). This is why it’s important to conduct the analysis in the right domain. In this case, the domain is all of Pendo events. But, you could do this same analysis on just your users… or even just a segment of your users. You can easily see the breakdown of browsers for your users in the Pendo dashboard widget.
Microsoft desktop browsers
Now, let’s look at the data applicable to the discontinuance of support for older Microsoft browsers. The browsers that will remain supported are shown in green. The ones being deprecated are shown in red.
|IE 7 + IE 6||1.88%||5.74%|
The good news is that web app usage shows a higher percentage of usage from supported browsers compared to general browser usage. Taking into account the fact that Pendo events have lower occurrence of any Microsoft browser, the impact on the general population of the Microsoft announcement is more than two times that for Pendo customers. The take away is that web app users (at least those instrumented with Pendo) have significantly lower exposure to Microsoft’s support deprecation announcement. That said, the fact remains that 32% of 31% (roughly 10% of all browser usage) is at risk of being impacted by this announcement.
It’s also interesting that the use of IE6/7 is higher among web app usage. We have a couple theories to explain this. I’ll try to do the analysis on that and post again here.
I use “What? So what? NOW what?” as a mnemonic to think about data science. The “What?” and “So what?” aspects are where most data analysis goes, but the key to getting the most out of any data analysis is how it alters your decision making – the “NOW what?”.
I can think of a few decisions that might occur in your organizations as a result of this analysis, assuming the 10% of usage from deprecated browsers will decline sharply going forward:
- If you are starting a new project now, tell your development teams that they don’t need to support any browser earlier than IE 11.
- For ongoing projects, announce the end-of-life of support for browsers before IE 11 to your customers and encourage them to switch, citing the Microsoft announcement. You can use Pendo’s segment feature and filter on “ ‘Most recent browser name’ contains ‘IE’ ” and “ ‘Most recent browser version’ contains ‘10.0’ ”, etc. (Note that Pendo captures both the browser name and version as a string.)
- Depending upon your situation, you may also want to consider an effort to actively remove support for older browsers in your code base. This code is costly to maintain and more importantly it may be impacting your users’ experience. Simply switching to a later version of the libraries that you use would help. For instance JQuery 2.0 only supports IE 9+.
Be sure to send me your thoughts on analysis that you would like to see.