Upload
seanku
View
244
Download
0
Embed Size (px)
Citation preview
8/16/2019 Radware SOTU Report
1/18
SHARE THIS WHITEPAPER
STATE OF THE UNION | SUMMER 2015
ECOMMERCE PAGE SPEED
& WEB PERFORMANCEContent and Speed: The Magic Formula
https://plus.google.com/share?url=http://www.radware.com/summer-sotu2015/http://twitter.com/intent/tweet?text=Radware%20State%20of%20the%20Union%20Ecommerce%20Page%20Speed%20&%20Web%20Performance%20Report%20%20-%20Download%20It%20Now%20-%20http://www.radware.com/summer-sotu2015/http://www.facebook.com/sharer/sharer.php?u=http://www.radware.com/summer-sotu2015/http://www.linkedin.com/shareArticle?mini=true&url=http://www.radware.com/summer-sotu2015/
8/16/2019 Radware SOTU Report
2/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
Table of Contents
Content and Speed: The Magic Formula ..................................................................................................................... 3
Website Performance: A Look at the Top 100 Ecommerce Sites ................................................................................ 4
Key Findings .................................................................................................................................................................. 5
The Fastest Sites Are Delivering A Better User Experience .........................................................................................5
Finding 1: Most Site Owners Are Failing To Meet Customer Expectations For Speed ...............................................6
Among the Top 100 Pages, Median Time to Interact Is 5.5 Seconds ...................................................................6
Finding 2: Page Size and Complexity Typically Correlate to Slower Load Times........................................................6
The Median Page is 1905 KB in Size and Contains 169.4 Resources .................................................................. 6
Finding 3: Most Sites Fail to Take Advantage of Core Image Optimization Techniques .............................................7
48% of the Top 100 Retail Sites Fail to Compress Images ................................................................................... 7
What Can the Fastest Sites Teach Us about Web Performance? ................................................................................ 9
1. Faster pages are smaller. ..................................................................................................................................10
2. Faster pages have a faster Time to First Byte (TTFB). ..................................................................................... 10
3. Faster pages understand their critical rendering path and know what content to defer. ................................ 11
Reining in Hero Images: IKEA and Sierra Trading Post ............................................................................................. 11
IKEA: Realizing Major TTI Savings by Improving Image Handling ....................................................................... 11
Sierra Trading Post: Trading Slow-Loading Images for Fast-Loading Ones ........................................................ 13
14 Best Practices to take Your Site from “Slow” to “Go” ............................................................................................ 14
Learn More ................................................................................................................................................................... 16
Conclusion: Web Performance Optimization Leads to Speed Gains ......................................................................... 16
Give Customers What They Want And What They Need ..................................................................................... 16
Methodology ................................................................................................................................................................ 18
Sources. .......................................................................................................................................................................18
About Radware ............................................................................................................................................................ 18
8/16/2019 Radware SOTU Report
3/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
Content and Speed: The Magic Formula
Ecommerce is big business. How big? Well, according to estimates from Statista, ecommerce
sales for 2015 will be in the neighborhood of 1.7 trillion U.S. dollars, and that total is expected
to grow to 1.92 trillion in 2016
1
– more than the 2014 GDP of Canada.
All of this economic activity hinges on Internet usage, enabled by websites. It stands to reason that this mechanism
of buying and selling would be a well-oiled machine, nessed and optimized to handle user interactions with the
greatest eciency and speed possible. Yet, the trend of slow, bulky websites continues.
There’s a push-pull dynamic at play: while customers are expecting more website content and complexity, studies
have shown 57% of site visitors will abandon a web page after just 3 seconds if they are unable to interact.2 As noted
by Forrester Research, “Site performance and download times are some of the most critical aspects of ecommerce
that correlate to conversion rates.”3 Serving more content to customers is expected, but the goal is delivering more
content, faster. That’s the “magic formula.”
Failing that, you might need a dierent formula for calculating revenue lost due to abandonment of slow websites.
Assuming your site:
• averages 100,000 visitors per day
• with a 2% conversion rate (a common conversion rate for ecommerce sites is ≈ 2%)
• and those conversions average $54 each (90% of ecommerce shoppers spend an average of $54 per order) 4
You would ideally be looking at $108,000 in daily revenue. But if 57% of your total visitors leave, they take those
potential dollars with them – to the tune of $61,560 daily, or $22,469,400 annually. Here’s the equation, where x =
visitors and y = average revenue per customer, multiplied by 365 days:
(.57x)(.02)(y)365 = Lost Revenue
That’s a huge amount of money to leave on the table. If you know your site’s stats, substitute them in the equation – it
could be a frightening amount.
With so much at stake, it’s important to get to the underpinning factors behind site complexity and user engagement,
and requires going deeper than just measuring a page’s total load time. The primary metric we’ve been utilizing is the
Time to Interact (TTI), the time by which the page’s central hero image and key surrounding assets have loaded for
the user. While the understanding of interaction time is evolving as new research emerges, it continues to provide a
relatable metric by which we can benchmark and assess the on-going state of website performance.
Either way, customers bounce when they have to wait beyond a certain point, and examining the real-world
performance of the top 100 ecommerce sites, as ranked by number of site visits, reveals that only a few of these top
retail sites are currently meeting customer expectations for both content and page speed. Data from this research
shows that websites have continued to grow in complexity, making web performance optimization (WPO) automation
tools essential to delivering rich content quickly enough to convert browsers into buyers. The fact that most retailers
are increasingly responsive to customer demands for content, rather than speed, is discounting the importance of
user experience in buying decisions.
Remember – the magic formula's goal is delivering more content, FASTER. That’s the bottom line, and it will aect
your bottom line.
8/16/2019 Radware SOTU Report
4/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
Website Performance: A Look at the Top 100 Ecommerce Sites
In July 2015, we tested the home pages of the top 100 leading ecommerce web sites, as ranked by Alexa.com.
Top 100 E-Commerce Sites by Alexa Ranking
1 Amazon (US) 26 B&H Photo and Video 51 Humble Bundle 76 Ralph Lauren
2 Ebay 27 Staples 52 Living Social 77 Ulta
3 Netix 28 Forever 21 53 J Crew 78 Blu-Ray.com
4 Amazon (UK) 29 Sky 54 Kmart 79 Neiman Marcus
5 Walmart 30 Zappos 55 DX 80 L.L. Bean
6 Etsy 31 Ticketmaster.com 56 TigerDirect.com 81 CarGurus
7 Best Buy 32 JC Penny 57 Cambridge University Press 82 Land's End
8 IKEA (US) 33 Barnes & Noble 58 Oce Depot 83 Harbor Freight Tools
9 Home Depot 34 Bed Bath & Beyond 59 Dick's Sporting Goods 84 NET-A-PORTER
10 Steam 35 Walgreens 60 trade Me 85 ASDA
11 Target 36 Game Stop 61 Urban Outtters 86 Saks Fifth Avenue
12 Groupon 37 Victoria's Secret 62 HSN 87 Anthropologie
13 Newegg 38 Google Shopping 63 DirecTV 88 Pet Smart
14 Lowe's 39 Wiley 64 Sierra Trading Post 89 GNC
15 Macy's 40 Legacy.com 65 Sony 90 Pottery Barn16 Kohl's 41 QVC 66 Frys.com 91 ModCloth
17 Gap 42 AutoTrader 67 Cabela's 92 MapsofIndia.com
18 HM (US) 43 Rakuten.com 68 Mango US 93 CafePress
19 Nordstrom 44 Sam's Club 69 YOOX.com 94 SiriusXM
20 Sears 45 Cars.com 70 REDBUBBLE 95 Finish Line
21 CostCo 46 Shuttery 71 SHOPBOP 96 CarMax
22 6PM.com 47 CVS 72 Blomingdales 97 Backcountry
23 Nike (US) 48 Sephora 73 American Eagle Outtters 98 PetCo
24 Bodybuilding.com 49 REI 74 Drugstore.com 99 Vitacost.com
25 Overstock 50 iHerb 75 M&S 100 Adorama
We examined page load times and collected page metrics ranging from page
size and composition to the overall complexity of each web site, as well as the
adoption of key performance best practices. Analyzing this data and comparing it
to historical data reveals the trends – both good and bad.
This research was conducted to:
• Examine the performance and page composition of the top ecommerce sites,
• Show how these sites load for real people in real-world scenarios via dominant
browsers and common connection types, and to
• Predict the direction of site performance as shown by current and past trends
in terms of page size, composition, and adoption of performance best practices.
Except where otherwise noted,
the results discussed in this report
reect the stats generated by tests
conducted through version 43 of
Google’s Chrome browser. At the
time of conducting this research,
Chrome was the most widely used
browser in the United States,
with a market share of 39.54%,
according to StatCounter.
Alexa’s methodologycombines a site'sestimated average ofdaily unique visitorsand its estimated
number of pageviews over the past3 months, and thoseestimates are basedon direct sources tothe greatest extentpossible, includingcertied trac data
from sites electing topartner with Alexa.4
www.alexa.com/about 5
8/16/2019 Radware SOTU Report
5/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
Key Findings
Our key ndings are outlined in detail in this report and summarized below:
1. Most site owners fail to meet consumers’ expectations for the 3-second page speed mark. Among the top
100 ecommerce sites, only 12% rendered feature content in fewer than 3 seconds.
2. Page size and complexity typically correlate to slower load times. The median page is 1945 KB in size and
contains 169 resource requests. The median Time to Interact (TTI) is 5.5 seconds, which is considerably slowerthan users’ reported wait-time threshold of 3 seconds.
3. Site owners are not taking advantage of core optimization techniques. Despite the fact that images compose
50% to 60% of the average page’s total size, 48% of the top 100 sites received an “F” score from webpagetest.org
for image compression. At the bottom of the list, the slowest page had a Time to Interact of 34.1 seconds.
The Fastest Sites Are Delivering A Better User Experience
Among the top 100 ecommerce sites, the fastest home pages are dened
by their ability to display meaningful, interactive content. Meaning, the
time at which the document onload event res is not always the most
meaningful measure of a site’s performance.
In the chart below, we provide the Time To Interact (TTI) alongside the full load
time for each page in order to show a distinction between the two metrics. For
example, while Amazon.com has a full load time of 30.058 seconds, it has a
TTI of 2.7 seconds. Below the 3 second target, the TTI indicates Amazon.com
delivers a better user experience.
PERFORMANCE
METRICS DEFINED
Time to Interact (TTI) – The
point at which primary page
content renders and
becomes interactive (e.g.,
feature banners with function
call-to-action buttons).
Load Time – The point at
which all of a page’s resource
– from images to third-
party-party scripts – have
downloaded and rendered.
From a user experience
perspective, TTI is a valuable
performance metric, as itindicates when a page begins
to be usable.
MEASUREMENT
NOTE
In our research we strived to understand the most meaningful metrics that most accurately
reect user experience. While TTI and Load Time are valuable, we also see evidence of
behavioral variation between users. For example, some users appear to have a tendency to
wait for a page to stop rendering visually before they attempt an action – even though the
page may have been ready for interaction earlier. We are currently conducting a study
due for release in October 2015 that looks to explore this tendency.
Adorama
Zappos
GoogleShopping
SierraTrading Post
Etsy
LivingSocial
Trade Me
Amazon
Blu-ray.com
CarMax
B&H Photoand Video
DX
1 . 7 T T I
TTI LOAD TIME
1 3 . 2 5 6
2 . 1 T T I
8 . 7 4 9
2 . 1 T T I
4 . 1 2 5
2 . 3 T T I
6 . 2 7 8
2 . 5 T T I
9 . 9 5 8
2 . 5 T T I
1 0 . 3 8 3
2 . 5 T T I
1 5 . 5 3 3
2 . 7 T T I
3 0 . 0 5 8
2 . 7 T T I
4 . 3 9 7
2 . 7 T T I
9 . 6 2 6
2 . 8 T T I
3 . 5 9 9
2 . 8 T T I
2 1 . 2 3 2
8/16/2019 Radware SOTU Report
6/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
Finding 1: Most Site Owners Are Failing To Meet Customer Expectations For Speed
Among the Top 100 Pages, Median Time to Interact Is 5.5 Seconds
Time to interact (TTI) is a critical measurement, indicating the rst moment at which users can engage and respond
to the site’s call to action or the key action for that page.
Research has shown for some time that that 57% of consumers will
abandon a page that takes longer than 3 seconds to load. Given the
relationship between a page’s render time and business outcomes, it
is clear a Time to Interact of 5.5 seconds – .3 seconds slower than our
testing showed in February 2015 – is failing to meet the needs of both
consumers and site owners.
In addition, we found:
• Only 12% of the top 100 retail sites rendered feature content in
fewer than 3 seconds.
• 14% of the top sites took 10 or more seconds just to becomeinteractive, an increase from 9% in February 2015.
• At the bottom of the list, the slowest page had a Time to Interact
of 34.1 seconds.
Finding 2: Page Size and
Complexity Typically Correlate
to Slower Load Times
The Median Page is 1905 KB in Size
and Contains 169.4 Resources
While page size is part of the problem, page
complexity is arguably an even greater
performance challenge when each of these
169 page resources represents an individual
server call. Since Summer 2014, the median
number of resources for a top 100 ecommerce
site has grown by 69.4% — from a median of
100 resources, or individual server calls, to a
whopping current median of 169.4.
5.4s
26%
5.5s
72%
5.6s
73%
TOP 100RETAIL SITES
10s
34.1s
Slowest TTI
1905KB
169.4RESOURCES
SUMMER
2015
SUMMER
2014
1677KB
100RESOURCES
69.4%INCREASE
8/16/2019 Radware SOTU Report
7/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
For most sites, the greatest drain on performance is the need to complete dozens of network round-trips to retrieve
resources such as style sheets, scripts, and images. Only 20% of the time required to display a typical web page
is consumed by loading the page’s HTML. The remaining 80% is spent loading the additional resources needed to
render the page and performing client-side processing.
Each page resource makes an individual round trip from the user’s browser, which requests the le from the host
server, and in turn delivers the le to the browser. Each round trip can take 65-145 milliseconds (or more) for desktop
browsers – numbers that add up quickly when a typical page contains more than 100 resources.
Not only does each server call introduce an incremental performance slowdown, it also increases the risk of page
failure. For example, poorly executed CSS can create a host of performance problems, ranging from style sheets that
take too long to download and parse to improperly placed CSS les that block the rest of the page from rendering.
Smart optimization can alleviate much of this, without sacricing a page’s aesthetics.
Finding 3: Most Sites Fail to Take Advantage of Core Image Optimization Techniques
48% of the Top 100 Retail Sites Fail to Compress Images
Image compression is a core performance technique
that minimizes the size (in bytes) of a graphics le without
degrading the quality of the image to an unacceptable
level. Compressing image les lightens a web page’s
overall payload. Fewer bytes mean reduced bandwidth
and faster pages.
Currently, images typically comprise between 50%
to 60% of a page’s total weight, making them fertile
territory for optimization. However, only a scant 3% ofthe top 100 ecommerce sites received an ‘A’ grade for
image compression.
Reducing an image’s le size has two key benets:
• Improved time to interact with content reduces
likelihood of site bounce. Reducing the amount of
time required for images to be sent over the internet or
downloaded oers engaging content, faster.
• Faster loading time rewards frequent site visitors with improved user experience. Increasing the number
of images that can be stored in the browser cache can improve page render time on repeat visits to the same page.
Consumer expectations of image quality have never been greater. As mobile technology continues to push the visual
delity envelope, consumers also expect those same images to render quickly on their smartphones and tablets. In
order to remain competitive, site owners must meet consumers’ demand for large, high-resolution product images
while at the same time ensuring those images don’t slow down the overall loading time.
AB
C
D
F
N / A
3%
8%
13%
13%
48%
15%
8/16/2019 Radware SOTU Report
8/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
Consider the results from one of our tests (Note: "MIME types" are used to denote the type of information that a le
contains, such as .html meaning an HTML page):
Content Breakdown by MIME Type (First View)
Requests MIME Type Bytes
198 IMAGE 9249019
68 JS 1246979
57 HTML 191252
12 OTHER 105237
9 CSS 304677
1 FONT 21227
0 FLASH 0
On Chrome, out of the site’s 383 requests, 57.8% were for images, which accounted for over 83% of the site’s
overall size of 11,088 KB. That’s the lion’s share of the site’s bandwidth. With a TTI of 18 seconds, they are
nowhere near the three-second threshold. Their start render time was just over 15 seconds and their load time on
Chrome was 53.46 seconds.
Here’s the data from WebPageTest.org:
Performance Results (Median Run)
Document Complete Fully Loaded
LoadTime
FirstByte
StartRender
SpeedIndex
DOMElements
Time Requests Bytes In Time Requests Bytes In Cost
First View(Run 1)
53.46s 0.197s 15.99s 19446 1750 53.46s 338 9.136 KB 65.304s 383 11.088 KB $$$$$
Repeat View(Run 2)
8.526s 0.184s 5.39s 9562 1746 8.526s 76 983 KB 10.273s 89 1,018 KB
The results were similar on Firefox, with a load time of 77.633 seconds, and on Internet Explorer 11, where it took
91.321 seconds – the slowest time we recorded on any of the sites across all three browsers. At 32+ seconds before
interaction is possible, we’re looking at a TTI that’s 10x slower than the 3-second goal.
BYTESREQUESTS
CSS
FONT
HTML
IMAGE
JS
OTHER
8/16/2019 Radware SOTU Report
9/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
What Can the Fastest Sites Teach Us about Web Performance?
Only 12% of the top 100 retail sites rendered feature content in fewer than 3 seconds, while 14% took 10 seconds or
longer to become interactive. Taking a deeper look at the fastest and slowest sites in our research will show what the
top performing ecommerce websites have in common, where they dier, and what insights we can derive from this.
Here’s some key WebPagetest.org test data from the 10 fastest sites, in descending order:
Note: All measurements are in seconds, except First Byte, which is in milliseconds. Fully Loaded
TTI Load Time First Byte (MS) Start Render Speed Index DOM Element Time Requests KB In
1.7 13.256 234 3.192 4196 787 15.34 165 2,147
2.1 8.749 195 2.094 4171 640 11.31 113 1,467
2.1 4.125 389 0.793 1751 406 4.209 29 550
2.3 6.278 464 1.792 1966 1824 8.998 69 853
2.5 5.958 516 1.69 1731 861 6.772 56 8.27
2.5 10.383 35 2.146 4076 671 10.38 102 1,403
2.5 15.533 861 2.892 4515 2173 16.72 101 1,633
2.7 30.058 216 1.093 14607 2349 35.28 344 5,832
2.7 4.397 324 0.394 1609 4141 60.3 750 8,554
2.7 9.626 588 2.692 2745 869 10.48 114 818
Note: The above represents industry benchmark data resultant from tests conducted on the websites of real companies within the AlexaTop 100 Shopping Sites
TTI LOAD TIME
1 . 7 T T I
1 3 . 2 5 6
2 . 1 T T I
8 . 7 4 9
2 . 1 T T I
4 . 1 2 5
2 . 3 T T I
6 . 2 7 8
2 . 5 T T I
9 . 9
5 8
2 . 5 T T I
1 0 . 3 8 3
2 . 5 T T I
1 5 . 5 3 3
2 . 7 T T I
3 0 . 0 5 8
2 . 7 T T I
4 . 3 9 7
2 . 7 T T I
9 . 6 2 6
1 1 . 3 T T I
1 1 . 2 8 9 1
2 T T I
2 4 . 7 9 6
1 2 . 4 T T I
1 6 . 3 8
1 2 . 6 T T I
1 6 . 0 7 1
1 2 . 7 T T I
2 2 . 2 0 5
1 3 . 5 T T I
2 4 . 7 2 7
1 3 . 9 T T I
1 7 . 0 4 7
1 5 . 1 T T I
1 7 . 8 7
1 8 T T I
5 3 . 4 6
3 4 . 1 T T I
3 6 . 6 6 3
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
10 FASTEST SITES 10 SLOWEST SITES
8/16/2019 Radware SOTU Report
10/18
8/16/2019 Radware SOTU Report
11/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
Among the fastest sites, the quickest TTFB was 35 milliseconds, compared to 1.68 seconds for the bottom score
amongst the slowest sites. This dierence — 1.645 seconds — is signicant Bear in mind that TTFB isn’t a one-time
metric: It aects every resource on the page because its eects are cumulative.
3. Faster pages understand their critical rendering path and know what content to defer.
Deferral is a fundamental performance technique. As its name suggests, deferral is the practice of deferring any page
resources that are not part of a page’s critical rendering path, so they load last.
Site owners with faster pages appear to have a better handle on deferral, which can be inferred from looking at the
dierence between their page size metrics at the “document complete” point versus their size when the page has
fully loaded and all resources – including requests for “invisible” resources such as beacons and other third-party
scripts – have rendered in the browser.
As noted earlier, among the fastest pages in our study, the median page contained 82 resources and was 1040 KB in
size at doc complete. But when looking at full load time, the median page contained 113.5 resources and was 1550
KB in size. In other words, the median number of page resources grew by 38% between doc complete time and full
load time, and the median page size increased by 49%.
Compare this to the 14 slowest pages. The median page size grew by only 7.5% — from 2609 KB to 2807 KB, and
from 145.5 resources to 161.5 resources – between document complete time and full load time. In several cases,
the dierence between the doc complete and fully loaded metrics was either unchanged or only negligibly dierent.
These metrics indicate that site owners have not put a significant effort into leveraging deferral techniques
to optimize the critical rendering path.
Reining in Hero Images: IKEA and Sierra Trading Post
As web performance luminary Steve Souders explains, “Many websites use images as a major design element in the
page – these are called hero images. Because these hero images are critical design elements, it’s important that theyrender quickly, and yet often hero images load too late, frequently getting pre-empted by less critical resources on
the page.”6
IKEA and Sierra Trading Post had challenges in previous ecommerce performance studies. In our Spring 2015 report,
these sites ranked 77th and 93rd, respectively, out of 100.
Both of these sites had similar performance pains in Spring 2015: hero images, also referred to as feature images.
Our latest research, however, nds that both companies have made impressive performance comebacks. IKEA
brought their TTI down from 10.5 seconds to a much lower 3.6 seconds, and Sierra Trading Post’s home page
rendered primary content in fewer than 3 seconds.
How did IKEA and Sierra Trading Post regain their footing and improve their TTI scores?
IKEA: Realizing Major TTI Savings by Improving Image Handling
IKEA managed an impressive turnaround with their feature images and resulting TTI. In Spring 2015, visitors to the
page were confronted with an imageless page until the 10.5-second mark. In our more recent test, Time to Interact
had improved radically, from 10.5 seconds to 3.6 seconds – not yet under the 3-second mark, but headed in the right
direction in a big way.
http://-/?-http://-/?-
8/16/2019 Radware SOTU Report
12/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
1
IKEA.com Spring 2015 Summer 2015
Speed Rank 93 out of 100 19 out of 100Time to Interact 10.5 seconds 3.6 secondsLoad Time 15.98 seconds 11.815 secondsTime to First Byte 170 milliseconds 117 milliseconds
Start Render 3.993 seconds 2.298 secondsPage Size 2644 KB 1908 KB
Resource Requests 87 resources 87 resources
Spring 2015
In the Spring 2015 test, IKEA’s feature image took more than 13 seconds to load. That’s a full 10 seconds beyond the
3-second mark. The site fared much better in our most recent test:
Summer 2015
The same type of feature image loaded in 1.104 seconds this time! The results speak for themselves: IKEA jumped
from 93/100 to come in at number 19 on our list of the fastest pages amongst Alexa’s top 100 ecommerce sites – a
well-deserved gain of 74 slots!
8/16/2019 Radware SOTU Report
13/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
1
Sierra Trading Post: Trading Slow-Loading Images for Fast-Loading Ones
In our Spring 2015 test, Sierra Trading Post’s home page featured much slower-loading hero images than the home
page tested for this report, and the dierence in the resulting TTI was a marked one.
In Spring 2015, the page had 31 image requests and a Time to Interact (TTI) of 7.6 seconds; while at the time of this
test, the page had only 7 image requests and a TTI of just 2.3 seconds.
SierraTradingPost.com Spring 2015 Summer 2015
Speed Rank 77 out of 100 4 out of 100Time to Interact 7.6 seconds 2.3 secondsLoad Time 12.041 seconds 6.278 secondsTime to First Byte 407 milliseconds 464 millisecondsStart Render 2.190 seconds 1.792 secondsPage Size 936 KB 853 KB
Resource Requests 62 resources 69 resources
Spring 2015
As you can see in the lmstrip and water fall, the central hero image took more than 8.6 seconds to load – an eternity,
as far as TTI is concerned.
This time around, Sierra Trading Post managed to drastically improve their performance:
Summer 2015
8/16/2019 Radware SOTU Report
14/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
Sierra Trading Post demonstrates the value of optimizing their images to improve page speed. In the Spring 2015
page, the critical performance roadblocks were a few large, rotating images centered around sales. These images
each required between 8 to 9 seconds to download. By contrast, the page measured for this report contains leaner
feature images, each requiring less than 2.5 seconds to download and become available for user interaction.
Both sites show that it is possible deliver to more images and content to customers while also delivering speed and an
improved user experience. But they also previously illustrated, in the testing we did in Spring 2015, the potential pitfallof tying a CTA link to a feature image. As Forrester Research points out, “Good digital customer experience metrics
fulll two purposes: tracking site experience quality and pointing to opportunities for improvement.”7 By identifying their
bottlenecks, IKEA and Sierra Trading Post both improved their performance – and the user experience.
14 Best Practices to take Your Site from “Slow” to “Go”
Throughout this report, there have been numerous references to performance techniques and best practices that
site owners can leverage to optimize and accelerate their page load times – ultimately improving both the real and
perceived user experience for their visitors. Some of these techniques, which we also suggested in our Spring
2015 report, can be implemented manually or via an automated solution, while others can only be performed by
automated solutions.
1. Consolidate JavaScript and CSSConsolidating JavaScript code and CSS styles into common les that can be shared across multiple pages should
be a common practice. This technique simplies code maintenance and improves the eciency of client-side
caching. In JavaScript les, be sure that the same script isn’t downloaded multiple times for one page. Redundant
script downloads are especially likely when large teams or multiple teams collaborate on page development.
2. Minify CodeMinication, which is usually applied to scripts and style sheets, eliminates non-essential characters such as spaces,
newline characters, and comments. A correctly minied resource is used on the client without any special processing,
and le-size reductions average about 20%. Script and style blocks within HTML pages can also be minied. There are
many good libraries available to perform minication, often along with services to combine multiple les into one, whichadditionally reduces requests.
3. Enable Keep-AlivesEnabling keep-alives is one of the easiest “low hanging fruit” on the performance optimization tree, yet a signicant
number of sites fail to do this. TCP connection is the process by which both the user and the server send and
receive acknowledgment that a connection has been made and data can begin to be transferred. Too many TCP
connections will slow down your site. It’s not easy to speed up TCP connection, but you can control how many times
the connection takes place. To enable keep-alives, make sure you have the proper conguration on your servers and
load balancer.
4. Compress Text Compression technologies such as gzip reduce payloads at the slight cost of adding processing steps to compresson the server and decompress in the browser. These operations are highly optimized, however, and tests show
that the overall eect is a net improvement in performance. Text-based responses, including HTML, XML, JSON
(JavaScript Object Notation), JavaScript, and CSS, can all be reduced in size by as much as 70%.
5. Sprite ImagesSpriting is a CSS technique for consolidating images. Sprites are simply multiple images combined into a rectilinear
grid in one large image. The page fetches the large image all at once as a single CSS background image and then
uses CSS background positioning to display the individual component images as needed on the page. This reduces
multiple requests to only one, signicantly improving performance.
8/16/2019 Radware SOTU Report
15/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
1
6. Compress ImagesImage compression is a performance technique that minimizes the size (in bytes) of a graphics le without degrading
the quality of the image to an unacceptable level. Reducing an image’s le size has two benets:
• reducing the amount of time required for images to be sent over the internet or downloaded, and
• increasing the number of images that can be stored in the browser cache, thereby improving page render time
on repeat visits to the same page.
7. Reformat ImagesInappropriate image formatting is an extremely common performance culprit. An image that is saved to the wrong
format can be several times larger than it would be if saved to the optimal format. Images with unnecessarily high
resolution waste bandwidth, processing time, and cache space.
As a general rule of thumb, these are the optimal formats for common image types:
• Photos – JPEG, PNG-24
• Low complexity (few colors) – GIF, PNG-8
• Low complexity with transparency – GIF, PNG-8
• High complexity with transparency – PNG-24
• Line art – SVG
8. Ensure That Feature Images Are Optimized to Load Early and Quickly As discussed earlier in this report, site owners should be aware of the usability consequence of delaying the
rendering of feature content: a user who experiences instantaneous page rendering spends 20% of their viewing time
within the feature area of a page, whereas a user who endures an eight-second download delay spends only 1% of
their total viewing time looking at the featured space on a page.
9. Rethink the Design and Location of Call-to-Action Links in Feature GraphicsWhile the accepted design convention has been to position CTA buttons at the bottom of feature banners, this
convention does not always serve the best interests of users or site owners, as shoppers must wait for theimage to fully render before taking their next action on the page, as we illustrated earlier. The simplest solution:
Reposition the CTA.
10. Defer Rendering “Below the Fold” ContentEnsure that the user sees the page quicker by delaying the loading and rendering of any content that is below the
initially visible area, sometimes called “below the fold.” To eliminate the need to reow content after the remainder of
the page is loaded, replace images initially with placeholder tags that specify the correct height and width.
11. Defer Loading and Executing Non-Essential ScriptsMany script libraries aren’t needed until after a page has nished rendering. Downloading and parsing these scripts
can safely be deferred until after the onload event. For example, scripts that support interactive user behavior, such
as “drag and drop,” can’t possibly be called before the user has even seen the page. The same logic applies to scriptexecution. Defer as much as possible until after onload instead of needlessly holding up the initial rendering of the
important visible content on the page.
The script to defer could be your own or, often more importantly, scripts from third parties. Poorly optimized scripts
for advertisements, social media widgets, or analytics support can block a page from rendering, sometimes adding
precious seconds to load times.
8/16/2019 Radware SOTU Report
16/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
1
12. Use AJAX for Progressive Enhancement AJAX (Asynchronous JavaScript and XML) is a technique for using the XHR (XMLHttpRequest) object to fetch data
from a web server without refreshing the page where the code is running. AJAX enables a page to display updated
data in a section of a page without reconstructing the entire page. This is often used to respond to user interaction,
but it can also enable your application to load a bare-bones version of a page quickly, and then to ll in more detailed
content while the user is already viewing the page.
13. Preload Page Resources in the Browser Auto-preloading is a powerful performance technique in which all user paths through a website are observed and
recorded. Based on this massive amount of aggregated data, the auto-preloading engine can predict where a user
is likely to go based on the page they are currently on and the previous pages in their path. The engine loads the
resources for those “next” pages in the user’s browser cache, enabling the page to render up to 70% faster.
Note that this is a data-intensive, highly dynamic technique that can only be performed by an automated solution.
14. Implement an Automated Web Performance Optimization Solution While many of the performance techniques outlined in this section can be performed manually by developers, hand-
coding pages for performance is specialized, time-consuming work. It is a never-ending task, particularly on highly
dynamic sites that contain hundreds of objects per page, as both browser requirements and page requirements
continue to develop. Automated front-end performance optimization solutions apply a range of performance
techniques that deliver faster pages consistently and reliably across the entire site.
Learn More
If you’re reading this report, then it’s evident that you value your website’s performance and want to achieve the best
results for both your customers and your company. You can make a dierence in your site’s performance.
We urge you to examine your own manual and in-house automated techniques against this report’s best practices,
and then evaluate where there’s room for further optimization and improvement.
To learn more from seasoned experts about how to improve your current practices, or for a benchmark assessment
on where you stand against current industry leaders, send us an email at: [email protected] and include ‘WPO
Assessment’ in the subject line of your request. We’ll be happy to take a look and share our ndings with you.
Conclusion: Web Performance Optimization Leads to Speed Gains
Give Customers What They Want And What They Need
Ecommerce site owners are striving to give their customers the content they want, and most are sacricing
performance and revenue in the battle of “pretty vs. fast.”
These top online retailers are responding to customer demands for content but are still discounting the importance of
user experience in buying decisions. They need to ask themselves: What is the value of the 57% of total site visitors
who abandon a slow website, and what can be done to retain them?
The good news is there are signicant opportunities for every site to optimize pages and ne-tune performance. Site
visitors can be converted into buyers by taking a more serious approach to front-end optimization.
Making intelligent tweaks to shorten the interaction time of the elements that matter the most to a website – from
feature images to search boxes and social media buttons – will help improve the overall eectiveness of the site.
mailto:info%40radware.com?subject=WPO%20Assessmentmailto:info%40radware.com?subject=WPO%20Assessment
8/16/2019 Radware SOTU Report
17/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
While page growth and complexity present critical web performance challenges, addressing these issues smartly
with the proper optimization and automation tools can nesse signicant site performance gains.
Site performance is also likely to improve as the new HTTP/2 standard is adopted. This new HTTP protocol enables
transaction multiplexing – enabling the browser to send any number of requests and receive the responses interleaved
and out of order. As a result, the connection between the browser and the server is much more ecient, and the wait
between subsequent requests and replies is nearly eliminated. This all contributes to faster page load times.
This HTTP/2 standard has just been released and we will be monitoring its adoption and making note of emerging
performance gains.
While site owners cannot aect the end-user environment, they do have a great deal of control over the user
experience they deliver. The most high-performing sites contain smaller, leaner pages, understand the critical
rendering path, and know what resources to defer.
Look to our list of top performing ecommerce sites for insight and optimization strategies to start delivering more
content, FASTER.
Take Charge of YourWebsite’s Performance
You don’t have to settle for slow pages. There are a number
of tools and techniques you can leverage to accelerate your site.
Find out more: http://www.radware.com/FastView/
http://www.radware.com/FastView/http://www.radware.com/FastView/
8/16/2019 Radware SOTU Report
18/18
STATE OF THE UNION
Ecommerce Page Speed & Web Performance – Summer 201
Methodology The tests in this study were conducted using an online tool called WebPagetest – an open-source project primarily developed andsupported by Google – which simulates page load times from a real user’s perspective using real browsers.
Radware tested the home page of the top 100 sites from the Alexa Retail 500 three consecutive times. The system automatically clearsthe cache between tests. The median test result for each home page was recorded and used in our calculations.
The tests were conducted on July 16, 2015, via the WebPagetest.org server in Dulles, VA, using Chrome 43 on a DSL connection. In veryfew cases, WebPagetest rendered a blank page or an error in which none of the page rendered. These tests were re-run with the samecriteria, and agged as such, with the results substituted in the list.
To identify the Time to Interact (TTI) for each page, we generated a timed lmstrip view of the median page load for each site in the
Alexa Retail 100. Time to Interact is dened as the moment that the featured page content and primary call-to-action button or menu is
rendered in the frame.
Sources 1 Statista, B2C E-Commerce Sales Worldwide from 2013 to 2018, April 2015
2 PhoCusWright, Consumer Response to Travel Site Performance, June 2010 3 Forrester Research, The Retail eCommerce Metrics That Matter, December 2014 4 RJMetrics, 2014 Ecommerce Benchmark Report, January 2014 5 www.alexa.com/about 6 Steve Souders, “Hero Image Custom Metrics,” May 2015 7 Forrester Research, The Retail eCommerce Metrics That Matter, December 2014
About Radware
Radware (NASDAQ: RDWR), is a global leader of application delivery and application security solutions for virtual,
cloud and software defined data centers. Its award-winning solutions portfolio delivers service level assurance
for business-critical applications, while maximizing IT efficiency. Radware’s solutions empower more than 10,000
enterprise and carrier customers worldwide to adapt to market challenges quickly, maintain business continuity and
achieve maximum productivity while keeping costs down. For more information, please visit www.radware.com.
Radware encourages you to join our community and follow us on: Facebook, Google+, LinkedIn, Radware Blog,
SlideShare, Twitter, YouTube, Radware Connect app for iPhone® and our security center DDoSWarriors.com that
provides a comprehensive analysis on DDoS attack tools, trends and threats.
This document is provided for information purposes only. This document is not warranted to be error-free, nor subject to any other warranties or conditions,
whether expressed orally or implied in law. Radware specifically disclaims any liability with respect to this document and no contractual obligations are formed
either directly or indirectly by this document. The technologies, functionalities, services, or processes described herein are subject to change without notice.
© 2015 Radware, Ltd. All Rights Reserved. Radware and all other Radware product and service names are registered trademarks of
Radware in the U.S. and other countries. All other trademarks and names are the property of their respective owners.
http://www.radwrae.com/http://www.radware.com/Solutions/ApplicationDelivery/http://www.radware.com/Products/Application-Network-Security/http://www.radware.com/https://www.facebook.com/Radwarehttps://plus.google.com/+radware/postshttp://www.linkedin.com/company/165642http://blog.radware.com/http://www.slideshare.net/Radwarehttp://twitter.com/#!/radwarehttp://www.youtube.com/radwareinchttp://itunes.apple.com/us/app/radware-connect/id391124100?mt=8http://security.radware.com/http://security.radware.com/http://itunes.apple.com/us/app/radware-connect/id391124100?mt=8http://www.youtube.com/radwareinchttp://twitter.com/#!/radwarehttp://www.slideshare.net/Radwarehttp://blog.radware.com/http://www.linkedin.com/company/165642https://plus.google.com/+radware/postshttps://www.facebook.com/Radwarehttp://www.radware.com/http://www.radware.com/Products/Application-Network-Security/http://www.radware.com/Solutions/ApplicationDelivery/http://www.radwrae.com/