Reviewing Atlas Page Performance - how should you do this? tyGraph Analytics has a window which will show load times, but SharePoint Page Diagnostics Tool (browser plug-in) also measures page performance. We'll be explaining some of the differences between these tools within this article.
We’ve undertaken some testing in Builqo. Please see some comparisons in the table below
Page |
tyGraph (Seconds) |
Diagnostics (seconds) |
Difference |
DPT - Legal |
2.012 |
2.538 |
- 0.526 |
Global policies & procedures |
2.126 |
2.679 |
-0.553 |
PRJ – Bank of Laska |
1.994 |
4.083/3.557/3.076/4.014 (4 tries) 3.6825 (average) |
-1.688 (average) |
DemoPoCsHome |
1.936 |
2.318 |
-0.382 |
Bid-bank of Laska – design of smart bank in NY retail park |
2.268 |
2.904 |
-0.636 |
Atlas Config |
3.248 |
2.765 |
+0.483 |
Bid Hub |
2.087 |
2.778 |
-0.691 |
We also looked at the most extreme cases shown within tyGraph.
Quickest page load in tyGraph is 1.042secs
INT – PM Executive Search |
1.042 |
2.510 |
-1.468 |
Slowest is 5.696 (which was an OOTB SP site)
OOTB SP Site Test |
5.696 |
1.240 |
+4.456 |
Slowest Atlas page load is 5.187
MTR – Smart Build Inc – IP Dispute |
5.187 |
2.575 |
+2.612 |
Review of Data
On average, for Builqo, for ClearPeople users, Diagnostics tool is presenting data as 0.5 seconds slower, but occasionally on some sites it can show the page as loading 2.5 seconds quicker than the average load time shown in the tyGraph statistic. Your production and users will receive different results as our environment and our testing is not indicative of your performance on your Atlas pages and M365 tenant.
Review of Tools
The tyGraph reporting on Average Page Load time is based on the average time reported to Azure Application Insights when the user's browser sends the request until the browser's page load event is called. Page load times for users can be influenced by the quality of their internet connection, VPN usage, web browser (settings, extensions, number of tabs open) and device performance - among a host of other variables. This information, along with some more detail, can be found here on support.tygraph.com.
They also state that if one page has consistently slow performing pages, the Microsoft Page Diagnostics tool for SharePoint should be used to investigate, as this helps to show how to optimise pages which might improve the page load times but also shows a more accurate view of how the page is performing for that user at that specific time, instead of a user-wide average. The page diagnostics tool analyses pages against a pre-defined set of rules and displays detailed information when results for a test fall outside of the baseline value.
It’s important to note that a slow load time from a user, perhaps who has a weak internet connection, can increase the average load time, and that the diagnostics tool can often show slower times, as browser calls are still being made and measured in the back-end, whilst the page, visually and functionally, has already loaded for the user. This might be why the above table shows the diagnostics results being slightly slower.
Summary
tyGraph shows an average time broken down across the user base where variables are mitigated (however, are still influential) and results are shown as an average. This is useful as some users have quicker and some have slower experiences. This result is a good gauge of the typical time it should take for the page to load across the entire user-base, but is not necessarily an accurate reflection of how quickly or slowly a page might load for you or for another user or set of users.
The SharePoint Diagnostic tool will measure that specific page performance for you at that moment in time. If you test 4 times in a row in quick succession you will get a different result each time because of the sensitivity of the variables involved in digital platform performance, web connection, hardware, and M365, among a long list of other things.
For example, the below Builqo tenant ‘PRJ – Bank of Laska’ workspace was tested with the SharePoint Diagnostics tool 4 times in a row, and, in the same 5 minutes, the time went from 4 seconds down to 3 seconds and then back up to 4 seconds (25% difference), and the average of that testing was 1.5 seconds above the tyGraph average load time. If we apply our assumption and logic from the above testing that Diagnostics is 0.5 seconds slower on average due to back-end calls, tyGraph report on average load time would still show this workspace +1 second faster than what the diagnostics tool shows, proving that during testing the page was loading more slowly for me than the average user.
Page |
tyGraph (Seconds) |
Diagnostics (seconds) |
PRJ – Bank of Laska |
1.994 |
4.083/3.557/3.076/4.014 (4 tries) 3.6825 (average) |
As the diagnostics is a live personal view, and the tyGraph reporting is a more accurate representation of what most other users might be experiencing, they will most likely show different results and have different use cases when analysing performance. Both tools are accurate in their own right but are not necessarily representative of another user’s experience.
ClearPeople’s Recommendation
- Use tyGraph for finding average page load times. Anything over 7 seconds average should be considered slow and further investigations can be made within this tool outlined in the tyGraph support page (e.g. geolocation of slower users).
- Use SharePoint diagnostics tool for reviewing pages reported as slow by several users in different locations, and undertake thorough testing following recommendations made by ClearPeople in these two articles (here and here)
- A combination of both tools should be used for any granular review or investigation or review, but keep in mind what each tool is showing.
We recommend you define some thresholds for review (leveraging both tools for insight, but ignoring yellow highlights in tyGraph average loads), and when average load time is higher than 10 seconds and number of Page Views is above 50, only in that case investigate if there could be a problem in the specific page by testing the page with Page diagnostics or investigating if the high average might correspond with specific geolocation or set of users instead.
Thanks for reading!
Comments
0 comments
Please sign in to leave a comment.