twinkle100 web performance study

21
Twinkle100 web performance study Find out what impact performance can have on your ranking…

Upload: measureworks

Post on 21-Jan-2018

3.737 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Twinkle100 web performance study

Twinkle100 web performance study Find out what impact performance can have on your ranking…

Page 2: Twinkle100 web performance study

For the second year in a row MeasureWorks researched the

actual web performance online consumers experience when

visiting the top 100 retail websites and the top 30 travel

websites. The performance benchmark is published in the annual

Twinkle100, by www.twinklemagazine.nl

This report provides an overview of the performance achieved by

the Twinkle100 companies and breaks it down into several

components attributing to site speed, thus creating a complete

overview of the state of web performance in the Netherlands.

Enjoy reading!

The MeasureWorks team

Twinkle100 Performance Benchmark

2

1. Why performance

matters?

2. Ranking the Twinkle100

3. Retail100 vs. Travel30

4. Performance detailed

5. Key findings

6. About MeasureWorks

What you can expect:

Page 3: Twinkle100 web performance study

1. Why Performance matters?

Page 4: Twinkle100 web performance study

A fast website leads to more pages viewed, satisfied visitors

and ultimately higher conversion rates [1]. Performance is

therefore not a feature, but an important requirement for

every site owner and online marketer, especially in highly

competitive online markets like retail and travel.

A mere 250 milliseconds, a blink of an eye, can make a visitor

decide to click way [2]. Continuous performance issues can

influence your brand’s online image. But there’s a positive side

to this story. Fast web pages enable visitors to get into a flow

state, in which they start enjoying their visit and will be more

focused to buy (additional) goods. Faster equals more page

views , equals more potential revenue.

[1] Forrester research

[2] Google research

4

“Performance is not a feature, but a requirement”

- Fred Wilson, Union Square Ventures

Every second counts!

Page 5: Twinkle100 web performance study

Just a 1-second

delay

Can cause 7% loss in

conversion

And lead up to 11% fewer

page views

With 16% decrease

in customer satisfaction

Performance is an important

indicator of any user experience.

You can have the most amazing

looking website, but with every

second delay that your visitor

encounters during his/her visit to

your website, you are

unnecessarily and very directly

risking your revenue stream.

49% less likely to visit again

5

46% visit competition

21% decides to tell others

Why Performance

matters?

Page 6: Twinkle100 web performance study

Social media mentions are not only

a great way to interact with your

audience, but also a great way for

monitoring how your website’s

performance is perceived.

12% of all social media

mentions in eRetail are about bad performance

Tweets about bad performance in social media

% negative tweets about bad performance (speed, downtime & usability) per branche, study conducted by idr1 &

MeasureWorks, june 2012

eRetail 88% 12%

95% 5%

96% 4%

eTravel

eFinance

6

Social Performance:

The topic is… You!

Page 7: Twinkle100 web performance study

It takes 1.38 seconds for the Twinkle100

fastest website to appear on your screen

2. Ranking the Twinkle100

Page 8: Twinkle100 web performance study

Based on annual revenue,

Twinkle magazine determines

the candidates for the

Retail100. These candidates

are ranked based on 1) page

load time and 2) availability.

Fastest website in the

Retail100 is HM.com with a

page load time of 1.38

seconds. This compared to

thuisapotheek.nl that closes

the Retail100 with a page load

time of 10.38 seconds.

Rank Site Load time Availability

1 www.hm.com 1,38 100,00%

2 www.kpn.com 1,72 99,81%

3 www.ibood.com 2,02 100,00%

4 www.dolcis.nl 2,27 100,00%

5 www.paylogictickets.n

l 2,32 100,00%

6 webwinkel.ah.nl 2,42 99,59%

7 www.gsmweb.nl 2,42 100,00%

8 www.wehkamp.nl 2,60 99,94%

9 www.budgethouse.nl 2,79 99,22%

10 www.123inkt.nl 2,88 99,72%

Rank Site Load time Availability

91 www.mediamarkt.nl 6,51 99,88%

92 www.zalando.nl 6,74 100,00%

93 www.large.nl 6,77 100,00%

94 www.just-eat.nl 6,86 92,00%

95 www.jeansonline.nl 6,90 99,22%

96 www.gsmwijzer.nl 6,93 99,94%

97 www.sanitairwinkel.nl 7,11 100,00%

98 www.menatwork.nl 7,70 100,00%

99 www.score.nl 8,06 99,88%

100 www.thuisapotheek.nl 10,38 100,00%

Top 10 Bottom 10

8

Average

Top 10

Average

Bottom 10

2,28sec. Speed 7,40sec.

622 kb Size 1579 kb

49 Objects 113

4 Third party 9,5

The bottom 10 websites are

3x slower compared to the

top 10

2. Retail100 ranking

Page 9: Twinkle100 web performance study

Average

Top 10

Average

Bottom 10

2,68sec. Speed 8,56sec.

714 Kb Size 1547 kb

56 Objects 113

8 Third party 5

Top 5 Bottom 5

The top5 travel is half the page

size compared to the bottom5

9

Ranking Site Load time Availability

1 www.sunweb.nl 1,7 100,00%

2 www.oad.nl 2,7 99,63%

3 www.elmarreizen.nl 2,8 98,03%

4 www.arke.nl 3,1 100,00%

5 www.landal.nl 3,1 100,00%

Ranking Site Load time Availability

26 www.d-reizen.nl 5,9 99,84%

27 www.belvilla.nl 6,4 99,78%

28 www.worldticketcenter.nl 7,6 100,00%

29 www.centerparcs.nl 7,6 99,50%

30 www.nshispeed.nl 15,3 99,25%

Equal to the Retail100, the

Travel30 candidates are

selected based on annual

revenue. Again, these

candidates are ranked based

on 1) page load time and 2)

availability.

Clear winner, same as last

year, in the Travel30 is

sunweb.nl with a page load

time of 1.7 seconds. This

compared to nshispeed.nl,

where it takes almost 16

seconds before you can click

to the next page.

Travel30 ranking

Page 10: Twinkle100 web performance study

With an average speed of 4.64

seconds the Twinkle100 is almost

30% faster than the average

Alexa2000 website [1]

3. Retail100 versus Travel30

[1] Source: Fall State of the Union, Strangeloop, 2012

Page 11: Twinkle100 web performance study

The Retail100 beats the

Travel30 in a head-to-head

competition. It’s a close call,

but with an average page load

time of 4,57 seconds it’s

slightly faster than the

Travel30 with 4,70 seconds.

The contrast in down time is

striking, where the Travel30

beats the Retail100 with 19

hours of total downtime.

Travel30

Retail100

19,2 hour

16,3 hour

4,70 sec.

4,57 sec.

Visitors of the Travel30 experience

18% more downtime

11

Travel30

Retail100

Head-to-Head race

Average page load time (sec.)

Average downtime (hours)

Page 12: Twinkle100 web performance study

The first view represents what

a first-time visitor to the page

will experience. This means

that all content is served to

the browser as new content,

creating a first impression of

the delivered performance. As

long as this visitor does not

clear his browser cache, every

time he returns to your site it

will be served as a “repeat

view”.

With the repeat view you are

able to create a faster

experience because several

elements can be cached

locally in the browser cache.

The more caching, the less

items will have to be

downloaded, essentially

creating the illusion of blazing

fast websites.

Re

tail

10

0

First view: 4,57s

Tra

ve

l30

First view: 4,70s

On average 46% of all content is

cached

12

First View versus

Repeat view

Repeat view: 2,70s

Repeat view: 2,73s

Page 13: Twinkle100 web performance study

To understand why a page loads in

a certain time, we need to look at

the composition of each individual

web page. As each page consists

of several page objects such as

images, JavaScript, and CSS files.

The more objects, the more time it

will take to compose the webpage

itself.

Why is this important? Each

resource makes a round trip from

the user’s browser, which requests

the file from the host server, which

in turn delivers the file to the

browser. Each of these round trips

can take milliseconds to download,

but go up quickly when pages

contain dozens of objects.

In 2011 the average Twinkle100

site contained 88 objects. 88

reasons for your website to

become slow!

Travel30 Retail100

0

50

100

150

200

250

ww

w.t

ui.n

lw

ww

.co

olb

lue.

nl/

ww

w.f

on

q.n

lw

ww

.wo

rld

tick

etce

nte

r.n

lw

ww

.jean

son

line.

nl

ww

w.b

izzt

rave

l.nl

ww

w.z

alan

do

.nl

ww

w.m

edia

mar

kt.n

lw

ww

.vac

ance

sele

ct.n

lw

ww

.co

smo

x.n

lw

ww

.bel

villa

.nl

ww

w.s

anit

airw

inke

l.nl

ww

w.g

ree

tz.n

lw

ww

.th

eph

on

eho

use

.nl

ww

w.s

alla

nd

.eu

ww

w.v

od

afo

ne.

nl

ww

w.b

ol.c

om

ww

w.p

ath

e.n

lw

ww

.bax

-sh

op

.nl

ww

w.c

entr

alp

oin

t.n

lw

ww

.om

od

a.n

lw

ww

.frs

.nl

ww

w.d

rogi

ster

ij.n

etw

ww

.mo

der

n.n

lw

ww

.red

coo

n.n

lw

ww

.nsh

isp

eed

.nl

ww

w.d

ejo

ngi

ntr

a.n

lw

ww

.klm

.co

m/

ww

w.la

rge.

nl

ww

w.k

ado

wer

eld

.nl

ww

w.b

adka

me

rmar

kt.n

lw

ww

.deb

ijen

korf

.nl

ww

w.n

ecke

rman

n.n

lw

ww

.eci

.nl

ww

w.c

ame

ran

u.n

lw

ww

.co

ren

do

n.n

lw

ww

.tic

kets

ervi

ce.n

lw

ww

.nrc

lux.

nl

ww

w.c

ente

rpar

cs.n

lw

ww

.pab

o.n

lw

ww

.t-m

ob

ile.n

lw

ww

.men

atw

ork

.nl

ww

w.s

core

.nl

ww

w.4

lau

nch

.nl

ww

w.k

lee

rtje

s.co

mw

ww

.just

-eat

.nl

ww

w.v

oet

bal

dir

ect.

nl

ww

w.o

tto

.nl

ww

w.f

leu

rop

.nl

ww

w.m

anag

emen

tbo

ek.n

lw

ww

.myc

om

.nl

ww

w.e

bo

oke

rs.n

lw

ww

.tra

nsa

via.

com

/hv/

nl-

nl/

ho

me

Vlie

gwin

kel.n

lw

ww

.ro

om

po

tpar

ken

.nl

sto

re.a

pp

le.c

om

/nl

ww

w.h

ema.

nl

ww

w.a

zert

y.n

lw

ww

.zes

ter.

nl

ww

w.g

sm.n

lw

ww

.sky

tou

rs.n

lw

ww

.tak

eitn

ow

.nl

ww

w.t

hu

isb

ezo

rgd

.nl

web

sho

p.t

eleg

raaf

.nl

ww

w.v

akan

tiev

eilin

gen

.nl

ww

w.g

smw

ijzer

.nl

ww

w.k

ras.

nl

ww

w.s

eeti

cket

s.n

lw

ww

.co

nra

d.n

lw

ww

.ch

eap

tick

ets.

nl

ww

w.b

lokk

er.n

lw

ww

.exp

edia

.nl

ww

w.n

ewyo

rkp

izza

.nl

ww

w.ja

ckjo

nes

.co

mw

ww

.elm

arre

izen

.nl

ww

w.s

nel

ton

er.n

lw

ww

.th

uis

apo

thee

k.n

lw

ww

.alt

ern

ate.

nl

ww

w.k

linge

l.nl

ww

w.m

exx.

nl

ww

w.o

utl

etp

lan

et.n

lw

ww

.alb

elli.

nl

ww

w.r

yan

air.

com

/nl

ww

w.k

om

ple

tt.n

lw

ww

.ass

em.n

lV

liegt

icke

ts.n

lw

ww

.vlie

gtic

kets

.nl

ww

w.b

cc.n

lw

ebw

inke

l.ah

.nl

ww

w.k

ijksh

op

.nl

ww

w.n

esp

ress

o.n

lw

ww

.to

ple

nze

n.n

lw

ww

.oad

.nl

ww

w.b

ud

geth

ou

se.n

lw

ww

.fu

turu

msh

op

.nl

ww

w.ic

entr

e.n

lw

ww

.web

pri

nt.

nl

ww

w.d

-rei

zen

.nl

ww

w.v

acan

sole

il.n

lw

ww

.vd

.nl

ww

w.p

aylo

gict

icke

ts.n

lw

ww

.flin

nd

al.n

lw

ww

.bel

tego

ed.n

lw

ww

.am

ster

dam

gold

.nl

ww

w.w

ehka

mp

.nl

ww

w.a

nw

b.n

l/w

ebw

inke

lw

ww

.op

waa

rder

en.n

lw

ww

.12

3in

kt.n

lw

ww

.ko

op

jesk

amp

ioen

.nl

ww

w.la

nd

al.n

lw

ww

.gsm

web

.nl

ww

w.d

om

ino

s.n

lw

ww

.ven

te-e

xclu

sive

.nl

ww

w.v

ksh

op

.nl

ww

w.w

ijnvo

ord

eel.n

lw

ww

.do

lcis

.nl

ww

w.n

eck.

nl

ww

w.w

orl

dti

cket

sho

p.n

lw

ww

.wif

ish

op

.nl

ww

w.le

nsw

ay.n

lw

ww

.sch

eere

nfo

pp

en.n

l/w

ww

.plu

tosp

ort

.nl

ww

w.b

abyp

ark.

nl

ww

w.a

rke.

nl

ww

w.p

laye

xper

t.n

lw

ww

.su

nw

eb.n

lw

ww

.hm

.co

m/n

lw

ww

.del

l.nl

ww

w.k

pn

.co

m/

ww

w.ib

oo

d.c

om

/nl/

nl/

HTML

CSS

Javascript

Images

With 65% of all objects, images

are the biggest part of any website

13

Twinkle100: Side by side comparison

4

5

13

57

Total objects: 87

4

6

16

61

Total objects: 96

Objects per page

# O

bje

cts

Retail100 Travel30

Page 14: Twinkle100 web performance study

0

500

1000

1500

2000

2500

3000

3500

htt

p:/

/ww

w.s

kyto

urs

.nl

htt

p:/

/sto

re.a

pp

le.c

om

/nl

htt

p:/

/web

sho

p.t

eleg

raaf

.nl

htt

p:/

/web

win

kel.a

h.n

lh

ttp

://w

ww

.12

3in

kt.n

lh

ttp

://w

ww

.4la

un

ch.n

lh

ttp

://w

ww

.bel

villa

.nl

htt

p:/

/ww

w.a

lbel

li.n

lh

ttp

://w

ww

.alt

ern

ate.

nl

htt

p:/

/ww

w.b

izzt

rave

l.nl

htt

p:/

/ww

w.a

mst

erd

amgo

ld.n

lh

ttp

://w

ww

.an

wb

.nl/

web

win

kel

htt

p:/

/ww

w.n

shis

pee

d.n

lh

ttp

://w

ww

.ass

em.n

lh

ttp

://w

ww

.aze

rty.

nl

htt

p:/

/ww

w.v

acan

sole

il.n

lh

ttp

://w

ww

.bab

ypar

k.n

lh

ttp

://w

ww

.bad

kam

erm

arkt

.nl

htt

p:/

/ww

w.b

ax-s

ho

p.n

lh

ttp

://w

ww

.bcc

.nl

htt

p:/

/ww

w.b

elte

goed

.nl

htt

p:/

/ww

w.v

acan

cese

lect

.nl

htt

p:/

/ww

w.b

lokk

er.n

lh

ttp

://w

ww

.bo

l.co

mh

ttp

://w

ww

.nec

kerm

ann

.nl

htt

p:/

/ww

w.b

ud

geth

ou

se.n

lh

ttp

://w

ww

.cam

era

nu

.nl

htt

p:/

/ww

w.c

entr

alp

oin

t.n

lh

ttp

://w

ww

.co

nra

d.n

lh

ttp

://w

ww

.wo

rld

tick

etce

nte

r.n

lh

ttp

://w

ww

.co

olb

lue.

nl/

htt

p:/

/ww

w.c

osm

ox.

nl

htt

p:/

/ww

w.d

ebije

nko

rf.n

lh

ttp

://w

ww

.del

l.nl

htt

p:/

/ww

w.c

ente

rpar

cs.n

lh

ttp

://w

ww

.do

lcis

.nl

htt

p:/

/ww

w.d

om

ino

s.n

lh

ttp

://w

ww

.klm

.co

m/

htt

p:/

/ww

w.d

rogi

ster

ij.n

eth

ttp

://w

ww

.eci

.nl

htt

p:/

/ww

w.f

leu

rop

.nl

htt

p:/

/ww

w.la

nd

al.n

lh

ttp

://w

ww

.flin

nd

al.n

lh

ttp

://w

ww

.fo

nq

.nl

htt

p:/

/ww

w.f

rs.n

lh

ttp

://w

ww

.fu

turu

msh

op

.nl

htt

p:/

/ww

w.a

rke.

nl

htt

p:/

/ww

w.g

ree

tz.n

lh

ttp

://w

ww

.gsm

.nl

htt

p:/

/ww

w.v

akan

tiev

eilin

gen

.nl

htt

p:/

/ww

w.g

smw

eb.n

lh

ttp

://w

ww

.gsm

wijz

er.n

lh

ttp

://w

ww

.hem

a.n

lh

ttp

://w

ww

.kra

s.n

lh

ttp

://w

ww

.hm

.co

m/n

lh

ttp

://w

ww

.ibo

od

.co

m/n

l/n

l/h

ttp

://w

ww

.icen

tre.

nl

htt

p:/

/ww

w.ja

ckjo

nes

.co

mh

ttp

://w

ww

.jean

son

line.

nl

htt

p:/

/ww

w.c

ore

nd

on

.nl

htt

p:/

/ww

w.ju

st-e

at.n

lh

ttp

://w

ww

.kad

ow

erel

d.n

lh

ttp

://w

ww

.kijk

sho

p.n

lh

ttp

://w

ww

.kle

ert

jes.

com

htt

p:/

/ww

w.k

linge

l.nl

htt

p:/

/ww

w.k

om

ple

tt.n

lh

ttp

://w

ww

.ko

op

jesk

amp

ioen

.nl

htt

p:/

/Vlie

gtic

kets

.nl

htt

p:/

/ww

w.k

pn

.co

m/p

rive

/ho

me.

htm

htt

p:/

/ww

w.e

bo

oke

rs.n

lh

ttp

://w

ww

.larg

e.n

lh

ttp

://w

ww

.vlie

gtic

kets

.nl

htt

p:/

/ww

w.le

nsw

ay.n

lh

ttp

://w

ww

.man

agem

entb

oek

.nl

htt

p:/

/ww

w.m

edia

mar

kt.n

lh

ttp

://w

ww

.tu

i.nl

htt

p:/

/ww

w.m

enat

wo

rk.n

lh

ttp

://w

ww

.mex

x.n

lh

ttp

://w

ww

.mo

der

n.n

lh

ttp

://w

ww

.exp

edia

.nl

htt

p:/

/ww

w.m

yco

m.n

lh

ttp

://w

ww

.nec

k.n

lh

ttp

://w

ww

.nes

pre

sso

.nl

htt

p:/

/ww

w.d

-rei

zen

.nl

htt

p:/

/Vlie

gwin

kel.n

lh

ttp

://w

ww

.new

york

piz

za.n

lh

ttp

://w

ww

.nrc

lux.

nl

htt

p:/

/ww

w.o

mo

da.

nl

htt

p:/

/ww

w.o

pw

aard

eren

.nl

htt

p:/

/ww

w.o

tto

.nl

htt

p:/

/ww

w.o

utl

etp

lan

et.n

lh

ttp

://w

ww

.pab

o.n

lh

ttp

://w

ww

.pat

he.

nl

htt

p:/

/ww

w.p

aylo

gict

icke

ts.n

lh

ttp

://w

ww

.pla

yexp

ert.

nl

htt

p:/

/ww

w.p

luto

spo

rt.n

lh

ttp

://w

ww

.red

coo

n.n

lh

ttp

://w

ww

.sal

lan

d.e

uh

ttp

://w

ww

.san

itai

rwin

kel.n

lh

ttp

://w

ww

.sch

eere

nfo

pp

en.n

l/h

ttp

://w

ww

.sco

re.n

lh

ttp

://w

ww

.see

tick

ets.

nl

htt

p:/

/ww

w.s

nel

ton

er.n

lh

ttp

://w

ww

.dej

on

gin

tra.

nl

htt

p:/

/ww

w.t

akei

tno

w.n

lh

ttp

://w

ww

.th

eph

on

eho

use

.nl

htt

p:/

/ww

w.t

hu

isap

oth

eek.

nl

htt

p:/

/ww

w.t

hu

isb

ezo

rgd

.nl

htt

p:/

/ww

w.t

icke

tser

vice

.nl

htt

p:/

/ww

w.t

-mo

bile

.nl

htt

p:/

/ww

w.t

ran

savi

a.co

m/h

v/n

l-n

l/h

om

eh

ttp

://w

ww

.to

ple

nze

n.n

lh

ttp

://w

ww

.vd

.nl

htt

p:/

/ww

w.c

hea

pti

cket

s.n

lh

ttp

://w

ww

.ro

om

po

tpar

ken

.nl

htt

p:/

/ww

w.e

lmar

reiz

en.n

lh

ttp

://w

ww

.ven

te-e

xclu

sive

.nl

htt

p:/

/ww

w.v

ksh

op

.nl

htt

p:/

/ww

w.r

yan

air.

com

/nl

htt

p:/

/ww

w.v

od

afo

ne.

nl

htt

p:/

/ww

w.v

oet

bal

dir

ect.

nl

htt

p:/

/ww

w.w

ebp

rin

t.n

lh

ttp

://w

ww

.weh

kam

p.n

lh

ttp

://w

ww

.oad

.nl

htt

p:/

/ww

w.w

ifis

ho

p.n

lh

ttp

://w

ww

.wijn

voo

rdee

l.nl

htt

p:/

/ww

w.w

orl

dti

cket

sho

p.n

lh

ttp

://w

ww

.zal

and

o.n

lh

ttp

://w

ww

.su

nw

eb.n

lh

ttp

://w

ww

.zes

ter.

nl

Pag

esiz

e i

n K

b

1072

1135 Travel30

Retail100

The largest websites is 10x bigger

compared to the smallest website

14

Not only many objects can make

your website slow, also the size of

each object can make all the

difference between fast and very

slow.

Why? Each resource makes a trip

from a host server to your web

browser the time it takes depends

heavily on the size of each object.

The bigger the object the more

time it will take for transfer, the

same as sending a large

attachment via email. The larger

the attachment, the more time it

takes to leave your inbox (or to

download it).

For your webpage it’s the same.

The bigger the size, the longer it

takes for content to download into

your browser. The more large

objects, the problem multiplies.

Thus the end user is forced to wait

for his content to appear on the

screen.

Page Size

Twinkle100: Page Size comparison Retail100 Travel30

Average page size (Kb)

Page 15: Twinkle100 web performance study

89%

51% 44%

31% 27%

24% 23% 23% 21% 19% 17% 16% 14% 14% 13% 13% 13% 12% 12% 12%

Facebook is most used as 3rd

party social domain

15

Top20 most used Third party scripts in Retail100 and Travel30 sites

Third party content, like images,

adverts, analytics or social sharing

buttons, are the fastest growing

area of content for a web page.

In 2011 almost 20% of every

webpage consisted of third party

content. Although this type of

objects usually adds appealing

content to your site, the trouble

with 3rd party performance is that

that quality is out of your hands.

Since third party content is only

assembled in the browser of your

end user, with limited control and

visibility.

You depend on the service level of

the third party supplier provides for

everything to work, so basically the

speed and availability of your site

could depend on the performance

of another vendor.

Third Party Content

Page 16: Twinkle100 web performance study

The PageSpeed Score

indicates how well optimized

your web page is, thus how

fast it can be. The higher the

score, the better optimized

your web page is to deliver a

solid performance according

to optimization rules from

Google. A low score indicates

room for improvement.

Note: The PageSpeed Score

does not measure the time it

takes for a page to load.

79

75 Travel30

Retail100

70

72

74

76

78

80

12-2011 08-2012

16

Average Google page speed score

Google knows best

Twinkle100: Page Speed comparison Retail100 Travel30

Retail100 websites show a

steady increase in page speed

score

Page 17: Twinkle100 web performance study

Twinkle100 websites are getting

bigger and slower, down 9%

compared to 2011[1]

4. Key Findings

[1] Source: Fall State of the Union, Strangeloop, 2012

Page 18: Twinkle100 web performance study

2011 2012

3,9

3,4

2011 2012

1150

1025

2011 2012

9288

2011 2012

24

19

Speed (sec.) Size (Kb) Objects (#) 3d Party (#)

After rev iewing the

per formance of the Reta i l100

and Travel30 the genera l

conclus ion is that we are

slowing down. Pages are

get t ing b igger wi th more

content and more th i rd

par t ies being used. This

resul ts in s lower webpages

and lower avai lab i l i ty.

The overall performance

is down 9%

compared to 2011

18

Trending

Page 19: Twinkle100 web performance study

19

Key Findings Performance of Twinkle100 is slower compared to

2011 The average Twinkle100 site keeps on growing. More objects

and larger content will result in slower webpages. The difference

between #1 (1.70 seconds) and #30 (15,30 seconds) in Travel30

best shows the impact: 268Kb and 38 objects compared to

1191Kb and 108 objects. Retail100 beats the Travel30 in every category An interesting finding is that the Retail100, on average, beats the

Travel30 in every category researched. Not only does this result

in a slower page load time average, but webpages are bigger in

size and contain more objects/third parties.

Websites rely more on Third Party Content The usage of third party content keeps growing, from 19% in

2011 to 24% in 2012. Next to the usual suspect like analytics and

affiliaties we see a growth in social and testing/optimization

tooling. This raises some concern because performance is

impacted by the quality of delivery from these third parties. Do

you know what their SLA is?

Page 20: Twinkle100 web performance study

Web Performance is not just something technical, it ’s something every online

consumer experiences day in day out when surfing the (mobile) web, at

home, in the car or at work. Most of the times it ’s fun, but sometimes a

struggle with content squeezed on a mobile screen, flash adverts, slow

websites or failure to load pages.

The setup of our performance tests are to measure web performance from a

end-users perspective:

- Via webpagetest.org which simulates page load times from a real user’s

perspective using real browsers, in this case Internet Explorer 8;

- Average internet connection with a download speed of 10Mbps and an

upload speed of 1.5 Mbps[1];

- All measurements are performed 24 hours per day from December 1st

2011 to August 1st 2012.

- All candidates for the Retail100 and Travel30 are measured during the

entire period, so all are compared equal.

- For equal comparison only the homepage is measured, since in most

cases this is the first page a first time visitors visits.

All data analysis is done based on raw data, which means that no averages

are used to calculate performance. Based on the revenue per year provided

by Twinkle magazine we linked actual performance (speed & downtime) to

revenue risked.

[1 ] This is accord ing to recent research by Akamai representat ive rate for the Dutch in ternet users.

Akamai State of the Internet Q1 2012 (http://www.akamai.com/stateoftheinternet/)

20

Methodology

Page 21: Twinkle100 web performance study

Want your own performance report or more details? Contact us via [email protected] and/or @measureworks

MeasureWorks ensures fast and reliable websites and mobile apps. We

are constantly working to analyze and optimize the performance of our

customers online environment, to exceed their business objectives. For

this we create and combine data from different data sources. Combined

with the expertise of our consultants we deliver actionable insights and

optmization.

Our customers, like Bol.com, Thomas Cook and GlobalCollect, manage

real-time the impact of performance on their business objectives and

deliver an user experience that exceeds the expectations of online

visitors.

www.measureworks.nl

About MeasureWorks