Presentation is loading. Please wait.

Presentation is loading. Please wait.

Benchmark Methodology - Update

Similar presentations


Presentation on theme: "Benchmark Methodology - Update"— Presentation transcript:

1 Benchmark Methodology - Update
February 2017 ©2017 Dynatrace Systems Confidential

2 ©2016 Keynote Systems Confidential
Agenda Current Approach and Challenges Benchmarking Evolved – Key Changes Browser Processing W3C Visual Complete Availability Ranking Metrics Key Questions How are sites selected? How are pages/paths selected? Why the use of Chrome? Why not multiple browsers? Why the use of Mobile Emulation? ©2016 Keynote Systems Confidential

3 Current Approach and Challenges
Modern sites cannot be accurately measured without including Browser Processing Dynatrace has found that many pages have up to 50% of their time spent in Browser Execution Not surprising given extremely high amounts of 1st party and 3rd party JS overhead Not surprising given the complexity of these web applications Current benchmark is massively underreporting performance by excluding this processing time There is less correlation between actual perception of performance and end to end metric End to end is captured when the last network request has returned Includes all third party tags, all lazy loaded content and all content loaded below the fold Sites have so much 3rd party overhead that in most cases push out this end to end metric without any real impact on a site visitor The current benchmark is overreporting load time by only using end to end Multiple browsers are important for QA testing but do not necessarily reveal stark differences in performance The goal of the benchmark is to fairly compare members and to provide a view of a “typical” experience The goal is not to capture every possible user To that end a single browser (especially the most widely used browser – Chrome) is a good proxy for this “typical” experience Current Approach End to end is the core performance metric Browser Processing not included Multiple browsers are used ©2016 Keynote Systems Confidential

4 Benchmarking Evolved – Key Changes
W3C (Navigation Timing) Metrics Browser Processing Visual Complete Single Browser Simpler Rankings ©2016 Keynote Systems Confidential

5 W3C (Navigation Timing) Metrics
By capturing key Navigation Timing metrics, we can slice the data much more precisely than end to end No single metric is perfect at capturing when a page is ready but many are better than end to end At minimum, looking at the LoadEventEnd (or commonly called onload) allows us to exclude all lazy loaded or post loaded third parties ©2016 Keynote Systems Confidential

6 ©2016 Keynote Systems Confidential
Browser Processing The site below is a common example of the time taken in a typical page load that is tied to Browser Processing Browser Processing in this cases means excessive execution – not allowing the network fetching/threading to continue To be very clear, this is not parallel processing which happens for every page and site but excessive execution that serially interrupts network fetching/loading If Browser Processing were excluded (as it is in the current benchmark data) these pages would appear to be much faster than they actually are (Home for example has more time in execution than in network/back-end) Given the amount of processing is completely dependent on the page’s design – its impact on each page/site will be unique ©2016 Keynote Systems Confidential

7 ©2016 Keynote Systems Confidential
Visual Complete Even more useful than the W3C metrics is a measure of above the fold render When is the page rendered above the fold and likely ready to use This metric is “visual complete” Once available in Q3 this will become the primary metric for ranking in the benchmark results Speed Index Visually Complete Response Time ©2016 Keynote Systems Confidential

8 ©2016 Keynote Systems Confidential
Simpler Rankings Data should be easy to understand and the method of calculation transparent Each benchmark will have a single set of data – no longer combining data into a “meta” ranking Those who purchase the benchmark will have the ability to sort this data (in a CSV) based on several metrics Part of the weekly publication First Paint, Onload, Visual Render, End to End Dynatrace will also rank based on what it considers to be the most accurate metric available at the time Likely Onload first and then moving to Visual Render Variability rankings will be removed ©2016 Keynote Systems Confidential

9 ©2016 Keynote Systems Confidential
Key Questions How are sites selected? Dynatrace selects sites based a mix of factors included brand presence, traffic volumes (where known), measurability etc. Our hope is to represent the core sites in a vertical or industry How are pages/paths selected? Many of our benchmarks are Home Pages, as Home Pages are still a primary entry point to the site and can be indicative of overall site performance For our transactional (multi-step) benchmarks, Dynatrace selects common and comparable sets of pages that represent the most important functionally in the measured industry Why the use of Chrome? Chrome continues to dominate in terms of market penetration There are typically not major differences between browsers in terms of error and performance, allowing Chrome to be a good proxy for a “typical user” Why not multiple browsers? If we think about the common performance issues – back-end, network/infrastructure, third parties, core design etc – they do not vary by browser Similar with errors where most errors are tied to infrastructure/application or third party issues and will be captured by ANY browser There are corner cases where both errors (typically functional) or performance can be unique in a specific browser version. The best way to capture this is with Real User (RUM) data or functional testing and not with a benchmark. Why the use of Mobile Emulation? The goal of the benchmark is to help our customers understand how they can improve their site. By removing the high levels of errors and performance variability seen in all mobile networks we provide a accurate proxy for mobile load times without including errors and issues that are outside the site owners control. ©2016 Keynote Systems Confidential


Download ppt "Benchmark Methodology - Update"

Similar presentations


Ads by Google