products for people on the move

Mobile Search: what are people

searching for and how do they
search for it
How this affects search bidding, natural
Mobile Internet and web results

How to monetise a mobile site by adding a

search box
 Table of contents

2. Mobile Commerce and Mobile Search

3. Context of growth of search on mobile

5. Is mobile search like fixed internet search? (No, it’s not)

7. What else affects the volume of searches?

8. What actually gets searched?

11. Local search may not be as close as you think!

13. How do people search (do you want a new job)?

15. Do users click on the results? (Yes, they do)

16. What does this mean to the eco-system?

18. How does Mobile Commerce add value to Mobile Search?

19. Contact

1
Mobile Commerce and Mobile Search

As the number of people using the Mobile Internet increases, the

amount of mobile content is growing with it. As a result, there is a
greater need than ever to make mobile content and services easier to
find, both to enhance the user experience and ultimately to maximise
the revenue opportunities.

Mobile Commerce handles over 25% of the searches coming from the
17.38m Mobile Internet users via UK operator portals (see next
section), plus a growing number of other portals seeking to monetise Mobile Commerce
their mobile assets. In total dealing with a quarter of a billion searches handles over 25% of
per annum, a figure doubling each year, as people gravitate towards the searches coming
from UK mobile
the convenience of the Mobile Internet with new generations of smart
phones becoming the norm and better value data bundles/packages operator portals. The
being sold. company deals with a
total of quarter of a
Having built an analytic tool in 2002 to analyse these increasing billion searches per
numbers of search requests and clicks, a growing element of Mobile annum
Commerce’s core business model is the development of monetised
search using paid listings, thus connecting mobile users with
advertisers. This generates additional revenue for the network
operators (through off-portal traffic) and gives advertisers distribution.
The Mobile Commerce solution gives up to a 70% uplift in search
clicks compared to a raw search engine implementation.

Because of this Mobile Commerce has been recognised as the UK’s

leading provider of Search and Local Search services on mobile,
having been named in Deloitte’s top 50 fastest growing UK tech
companies (2008) and the Tech Media Invest Top 100 (2009) which
showcases the UK’s most innovative media and technology
companies.

Current Mobile Commerce customers include Tier One Mobile

Operators Vodafone, O2, T-Mobile, Orange plus Sony Ericsson, and
Microsoft.

2
 Context of growth of search on mobile

The volume of searches we are seeing on mobile devices has

more than doubled in the last 12 months. There are a number of
factors attributable to this increase.

According to the UK Mobile Data Association, there were 17.38

million users using the Mobile Internet in December 2008 (Source:
text.it) up from 16.5 million in May 2008.

The Orange Digital Media Index April 2010 (Source: Orange)

states that in December 2009 Orange had 3.41 million customers
using Orange World (the Orange Mobile Portal) up 5% from
February 2009.

So part of the growth in search terms can be accounted for by new

Mobile Internet users buying flat rate data bundles and having
handsets with improved capabilities that make them easier to use,
plus existing users trying the search function for the first time.
.
However, the average number of searches per unique user is
rising much faster. It has grown from 8 searches per unique user
per month to more than 13 searches per unique user in the last 12
months. This is a clear indication that once users try the search
box they continue to use it in greater and greater volume every
month. This is because of changes in the behaviour of the use of
the mobile phone with users becoming reliant on the Mobile
Internet.

The PC search market is dominated by the branded search

providers but the mobile market remains dominated by the
operator portals for access to the Mobile Internet. Operators
provide a different experience from the dominant PC search sites.

All UK operators now include a search box on the Home Page (and
often on many other pages) of their portal. To the left is the O2
Active portal. Notice that on the screen the user can see links and
icons as well as the search box (which is at the top of the image).

3
This relationship between the position of the icons and the search box
is very important in promoting/encouraging the use of search on
mobile. It is fundamentally different from the experience of using
Google/Bing on a PC where the search box itself tends to sit alone
(see right).

Any change made within the operator portal or the position of the
search box has a direct impact on the number of searches for a
particular term. For instance, if there is a display ad or clickable link for
Hotmail close to the search box and visible on the screen at the same
time (see O2 image on Page 3), the number of searches for that and
associated terms will drop when the link is displayed, and rise when
the link is removed or moved lower down on the portal. This does not
mean that the service itself becomes more or less popular but it helps
to explain the rise and fall in certain search terms.

Our experience shows that the higher the search box is placed on the
portal, the more it gets used, but this should not surprise anybody! If
placed below the fold users have a greater tendency to click on the
icons/links they can see.

Very simple things increase search volumes. Much has been written
over the years on bill shock. This is when a user accesses the Mobile
Internet to later discover that costs are significantly higher than
expected. Although the industry has done much to educate users, and
flat rate tariffs and daily plans have been introduced, simple things like
clearly stating that searching is free has an immediate effect on the
number of searches made.

This is how the “free web search” it is promoted on the Orange World
portal (right).

4
 Is mobile search like fixed internet search? (No, it’s not)

Mobile Search is not PC search and this drives many of the

differences in how services are accessed.

The significant factors are:

1. Ease of text input

2. Ease of navigation
3. Bookmarking

A mobile phone has a much smaller text input device than the
keyboard of a PC. This makes input errors much more likely.

Mobile phone screens are small so the user sees less information
before there is a need to pan left/right or scroll down. This makes
navigation around pages, or down through different levels on a
It is often much
portal, a greater challenge. So it is often much simpler to enter a
simpler to enter a term
term into a search box rather than attempt to find something on a
into a search box
portal, which may be several keystrokes away. Therefore, the
rather than attempt to
number of searches for items/content that are already somewhere
find something on a
within the portal is much higher than for the PC, where the eye can
portal, which may be
scan many pieces of information on a large screen, and the user
several keystrokes
can click directly on that link.
away
Most people are familiar with bookmarking a site on a PC. It’s very
easy to add favourites and when you want to visit them, just click
the drop down menus and you get there quickly. When using a
bookmark, there is no need to enter a search term.

Bookmarking on a mobile is much less prevalent. Although it is

possible to do, users tend not to – bookmarks can be hard to find
and with a regular change of handsets, users need to transfer their
bookmarks or re-bookmark every time they change phone, which
happens on a much shorter cycle than a PC upgrade.

5
So if it is hard to bookmark sites and hard to navigate around a portal
due to the screen size, users are much more inclined to use search to
get to the content that they visit on a regular basis. This is exactly the
behaviour that is seen and why Facebook was the single most
searched term of 2009.

Another way to allow users to input search terms without having the
need to enter text is to display tag clouds of commonly searched terms
on the appropriate pages of a portal. For instance a tag cloud
containing the most popular search terms on a games page may
include Monopoly, Pacman, Tetris and others. This allows the user to
click on a search term from the tag cloud and the system then
automatically sends that term to the search providers and draws back
and displays the appropriate results. It is easy, quick and convenient
for the user but the search term is generated by a click on a link rather
than the direct entry of a search term.

This is a screen grab of a tag cloud (see right) comprising commonly

searched games terms, with the search box underneath allowing the
user to “search for anything”.

6
 What else affects the volume of searches?

Having established there a number of User Interface elements that

affect the overall volume of searches, what are the external factors
that affect the specific search terms the user inputs?

Users on mobile don’t show the same propensity to search for

trending topics as they do on PC but there are certain one-off
subjects that have shown a surprising number of searches.

An example of this type of query would be “eclipse”, which became

a top 10 search term around the last UK eclipse of the moon.

Date and time are also strong drivers of search terms. Examples of
annual events that get heavily searched but only when they are
taking place are: The Boat Race, The Derby (the horserace not the
city) and The Eurovision Song Contest.

TV is also an influential driver of search terms and searches – The

X Factor, Big Brother, Dancing on Ice, and we expect, Over The
Rainbow, to follow the trend – all generate significant numbers of
searches but only when the program is showing. These terms tend
to last for a few weeks increasing dramatically towards the final
During a screening of episode and then disappear until the following season. During the
The Apprentice there screening of The Apprentice (Series 3) in 2007 there was a high
was a high volume of volume of searches for Nigella Seeds. This took place at the same
searches for Nigella time as they were being featured in one of the tasks in the series
Seeds. This took place and people watching the program used their mobiles to work out
at the same time they what contestants were looking for.
were being featured in
Events are another driver of search terms – every time Joe
one of the tasks and
Calzage fought the number of people looking for information on the
people watching the
fight rose. In the same way, the search term England is seen
program used their
whenever England play a fixture. However, it doesn’t appear to
mobiles to work out
matter if the sport is football, rugby, cricket or something else – the
what contestants were
terms that come in are England or England Result.
looking for
News also drives searches – the death of a celebrity always leads
to searches as people look for information. 2010 has seen a leap in
the number of searches for Malcolm McLaren following his recent
death. The number of people searching “nats” and airlines has
risen following the volcano.

Finally, the day of the week has a significant effect on the volume
of searches for some obvious categories. Lottery search is largest
after a draw – Wednesday and Saturday – and particularly large
after a big Euromillions draw. From the time these requests are
made it is clear that users are looking for the result rather than to
buy a ticket using their mobile in advance of the draw.

Sport is heavily searched over the weekends.

7
Train information is primarily a weekend search activity on mobile.
What actually gets searched?

Mobile Commerce started analysing search terms in 2004. Originally

we started with a list of the terms. Customers could look at their top
terms, click and see the next 50 etc by day, week, month and year.
Although this is interesting in itself, it actually hides the detail of what
people are looking for and how they are looking for it.

For example, on the right are the top 20 most searched terms of 2009
(in order). Note: Adult terms have been excluded but two would have
made the list. 1. Facebook
2. Google
In total, 29% of all searches by volume were for these twenty terms. 3. Bebo
The concentration of popular terms is caused by the input, navigation 4. Youtube
and bookmarking issues raised earlier. 5. www.facebook.com
6. Ebay
However, in the top 20 there are 4 different ways of accessing
7. Ebuddy
Facebook, and 2 of accessing both Google and YouTube.
8. Facebook.com
So the real intelligence lay in collecting together all the terms where 9. Hotmail
the search intent was the same and tagging this into a category thus 10. MSN
creating a taxonomy into which all search terms could be placed. This 11. Yahoo
makes the information easier to see and understand at a macro level. 12. Lottery
At the category level this is what users looked for in 2009. 13. You Tube
14. Face Book
15. Flirtomatic
16. Free Games
17. Yahoo Mail
18. Plenty of Fish
19. Google.com
20. twitter

8
 Looking closer, “Single User Search” accounts for 42% of ALL
searches and the largest category by far. Single user search
actually represents the very end of the long tail. These are
searches where only one user has entered that particular search
term – an example would be “[your user name] facebook login”.

The percentage of single user searches continues to rise which is

further demonstration of the acceptance of the mobile phone for
search. These terms may be entered multiple times but are from
just from one user.

“Uncategorised searches” are those searches when more than one

user has entered the search term but none of the terms are large
enough to be further categorised.
1. facebook
2. www.facebook.com “Other” contains heavily search terms where the term does not fit
3. facebook.com into the logical structure of the taxonomy.
4. face book Each category is split into its component parts e.g. social
5. facebook login networking includes Facebook, Bebo and others. Sport is divided
6. facebook/s by the type of sport e.g. football, and football includes all football
7. www.facebook.co.uk clubs and other football related searches.
8. facebook.co.uk
9. facebook mobile Through this taxonomy it is possible to see what users are actually
10. m.facebook.com looking for, and one click lets our customers see how each of the
categories is made up.

The Facebook category is made up from almost 700 different

search terms.

On the left are the top 10 Facebook search terms (in order).

9
Much lower down the list comes “face booklogin” (incorrect parsing),
“facebooj” (a simple error to make on a mobile device) and “log in face
book” (another common method of parsing).

The understanding of how subjects are searched is critical to 1. Pacman

companies wanting to get their mobile assets discovered either by 2. Pac Man
bidding or through natural search. 3. Pacman game
4. Pac-man
In another example, Pacman is searched using hundreds of different 5. Packman
search terms, the top ten being listed on the right. 6. Pac man game
For advertisers to get discovered they need to bid on not only these 7. Pack man 10
but also the other hundreds of Pacman terms, so they can be 8. Free pacman
displayed to the long tail of people interested in Pacman. 9. Ms pacman
10. Free pacman game
In addition, the advertisers must tag their mobile pages correctly to
ensure their results are found high up in the natural search results for
each of the terms.

It may seem odd that Google is the second most searched term on
mobile but the term holds a similar position on all operators’ portals
regardless of whether the search box is branded as Google, a
competitor (usually Yahoo! in the UK) or is white-label.

At first glance a conclusion could be drawn that users want a Google

experience on their phone but according to Google’s 2009 Zeitgeist,
Google is the 8th most searched term put into the Google search box
on a PC! Some things are just not that easy to explain.

10
 Local search may not be as close as you think!

There has been much coverage that local search on mobile is very
significant. Google themselves claimed that 33% of all searches
on mobile have a local context (Diana Pouliot, Director of
Advertising, Google at the Mobile Marketing & Advertising Event,
Las Vegas 2010). Our figures show a much lower level of local
searches when a search is made directly from the search box.

The single top mobile search term with a direct location context is
Halifax. However, it is clear from the click pattern that users are not
seeking information about the northern town (Pop. 82,000 –
Source 2001 census) but information on the bank with the same
name. We would not classify this as a location contextual search.

Another example is the term Derby. But as most clicks are made in
early June, it is clear that the searchers are looking for information
1. Manchester united about the horse race rather than a location query. There is another
2. Man Utd race that features two place names, so it can get even more
3. Man united confusing: The Kentucky Derby.
4. Manutd.com
5. Mufc It can be argued that sport also has a locational element in a
search.

Manchester United was the most searched football club in 2009

(being searched in >200 different ways) the most common being
(see left).

In addition searches starting with any of the above terms also are
suffixed by: fixtures, scores, ringtones, club crests, team news,
players names, v [another team] (e.g. Chelsea v Man U) etc.

Whilst all of these terms have a location element (Manchester)

they probably should not be regarded as local searches. This is
supported by other search terms including city names. For all
searches including the term Sheffield 43% are football related and
Sunderland where 58% are football related.

11
Cinema is a term that does have a strong local search context. In a
single month we would see in excess of 2,000 different search terms
with the word cinema or one of the large cinema chains in the search
string. Detailed analysis shows that search strings can generally be
divided as follows:

[chain] [townname] - e.g. Vue Sheffield

[chain] cinema/’s/s [townname] - e.g. Vue cinema Sheffield

[townname] [chain] - e.g. Sheffield Vue

[townname] cinema/’s/s - e.g. Sheffield cinema

[townname] cinema listing/s - e.g. Sheffield cinema listings

[townname] [chain] cinema - e.g. Sheffield Vue cinema

There are 5 times as many searches when the chain name is placed
before the town name than when it is placed after it.

Very few people local search using the term London (the most popular
being: London Marathon, Transport for London and London Tube Map)
as London locations tend to be searched by area, Peckham, Highgate,
Hammersmith etc
Cinema is a term that
In fact, of all “location searches” 55% include either a city or area
does have a strong
name, 17% have a Point-of-Interest (e.g. O2 Arena, Wembley
local search context.
Stadium), 15% use a full postcode and 13% a partial post code.
In a single month we
Local search turns out to be broadly similar from city to city. Again, see in excess of
looking at searches that include the term Sheffield, they can be 20,000 different search
categorised in the following ways (see chart below). Other cities show terms with the word
very similar percentages. Exceptions are cities with a local airport (e.g. cinema or one of the
Bristol) when the airport terms account for some 10% of searches e.g. large cinema chains in
Bristol airport, Bristol airport parking etc the search string

12
 How do people search (do you want a new job)?

In most categories there are specific ways in which very large

numbers of search terms are made. Despite the high volume Jobs
gets searched in very few ways (in order):

• Job in [place] e.g. Job in Salisbury

• Jobs in [place]
• Job [place]
• Jobs [place]
• [place] job
• [place] jobs
• Job in [county]
• Jobs in [county]
• Job [county]
• Jobs [county]

Of course, understanding how users are searching is critical for

Understanding how
any company wanting to bid on as many appropriate search terms
users are searching is
as possible and also get discovered through natural search.
critical for any
company wanting to The same is true for searching for weather; however a time
bid on as many element also comes into play i.e. 5 day, 7 day, 3 month etc.
appropriate search
terms as possible and Here we find searches are made the following ways:
also get discovered
• Weather [place &/or time period] e.g. weather Luton 5 days
through natural search
• Weather at [place &/or time period]
• Weather for [place &/or time period]
• Weather forecast for [place &/or time period]
• Weather forecast in [place &/or time period]
• Weather forecast [place &/or time period]
• Weather in [place &/or time period]
• Weather report [place &/or time period]

13
When looking for hotels, people search in the following ways:

• By hotel name 53%

• Hotels in [place] 14% e.g. Hotels in Reading
• [place] hotels 8%
• [place] hotel 7%
• Hotel [place] 7%
• Hotels [place] 6%
• [hotel chain] hotel [place] 2%
• Cheap hotels in [place] 2%
• By web address www. 1%

Music is also heavily searched. However, searchers are looking for a

broad number of different things around their favourite artists.

Once again there are several common themes in the way that these
searches are made, and the percentage split changes depending upon
what is band is announcing (in order):

• Band name (on its own)

Then band name plus:

• Pictures, pics, images, photo/s, wallpaper/s

• Video, vid/s)
• Album/s or single/s
• Downloads, mp3/s, theme/s, full track, song/s, tune/s)
• Ringtones
• Concerts, gig/s, live, ticket/s, tour/s)
• Lyrics

14
 Do users click on the results? (Yes, they do)

The Mobile Commerce system analyses what users actually click

on. Of course this depends greatly on the User Experience. How
results are presented back to the user by our customers can be
dramatically different between implementations. Click rates also
alter depending on the search term entered.

The over-whelming
number of clicks are
for mobile web results
showing that this is
the users’ preferred
option on mobile as The number of clicks for the sponsored search results increase
these sites have been when the number of results shown to the user increases, so long
optimised for mobile as those results are prominently displayed towards the top of the
devices results set. It reduces if the results are shown at the bottom of the
results set.

Although it appears roughly an equal split between those clicking

on mobile web results and web results, more analysis reveals that
more than 50% of all clicks from Single User Search terms are for
web results.

This means that the over-whelming number of clicks for commonly

searched terms are for mobile web results, showing that this is the
users’ preferred option on mobile as these sites have been
optimised for mobile devices giving a much better user experience.

15
What does this mean for the eco-system?

This paper brings some important learnings to both publishers and

brands.

Our customers for the Monetised Search API tend to be publishers

seeking to monetise their digital assets by adding a search box to their
mobile site. The customer decides which search terms to send to
Mobile Commerce and we return various sets of results to them. They
can display their own results at the top of the results set.

The publisher decides the user experience they want to offer their
customers through crafting a balance between their own results, Sponsored search is
sponsored search results, mobile web results, web results and images. much more effective
Our experience allows us to provide information on the optimal level of as an advertiser can
sponsored results to return and our algorithms generate up to a 70% bid on those keywords
increase in clicks over a raw search engine implementation. that are directly
relevant to the
Brands can get discovered on mobile in a number of different ways
services it has and link
including banner ads, sponsored search results and natural (Mobile
users directly to that
Internet and web) results.
part of their product
Banner ads generally only allow advertisers to bid on clusters of offering. To do this the
categories – e.g. an advertiser will buy placement for a term that they advertiser has to
believe to be appropriate for their product. However, the number of understand how users
categories is limited which means the relevance is reduced which are searching and bid
limits the click rate/effectiveness and hence the cost per click. on those permutations

Sponsored search is much more effective allowing an advertiser to bid

on keywords that are directly relevant to the services it has, and link
users directly to that part of their product offering. To do this the
advertiser has to understand how users are searching and bid on
those permutations.

Natural results require the correct tagging of sites and must ensure
they can be crawled by the search companies. As the mobile screen is
small, the number of results visible to a user is much less than on the
web. This makes it vital to ensure everything is done through mobile
search engine optimisation so the results are featured on the first page
of the results set.

16
 Much has been written about the poor quality of web sites when
transcoded onto mobile (Bango White Paper, Lost in the Mobile
Maze). If an advertiser is going to go to the trouble of wanting their
site to be discovered on mobile, then they must build mobile
variants that optimise to the different device types – e.g. a touch
screen handset site needs to be designed with strong visual appeal
and buttons that are larger and make it easy to click whereas an
older handset is navigated using the up/down keys/button (Taptu:
Exploring the Touch-Friendly Web).

Finally, Mobile Commerce has looked at many industry sites and

has been very disappointed with some of the intelligence deployed
behind the search box. Common spellings of pacman (the classic
arcade game) are pac-man and pac man. Searching the Nokia OVI
store with an N96, the term “pac man” brings back results, but the
term “pacman” brings back a “Your search did not match any items
compatible with your mobile”! (see left)

Clearly this is a poor user experience and negatively affects the

revenue of everybody in the value chain. This can be resolved very
simply by adding a “Did You Mean” feature that allows users to
click on alternative spellings so relevant results can be returned.

If you search Android Market you find the same shortcoming!

17
How does Mobile Commerce add value to Mobile Search?

Mobile Commerce has a number of services designed to help

publishers, advertisers and ad networks make the most from the
opportunities in mobile search.

These are:

• Monetisation for mobile properties and reporting for customers

• Mobile Internet and web results
• “Did You Mean” and “Also Try”

Mobile Commerce has a monetised search API that allows publishers

to place a search box on their site. The API is configurable so the
publisher can decide on what results they want to display to produce Mobile
Mobile Commerce
Commerce has has
an optimal experience for their consumers. a
a monetised
monetised search
search
API
API that
that allows
allows
The Monetised Search API also supports “Did You Mean” and “Also publishers
Try” functionality, which interprets requests and gives an alternative if publishers to to place
place aa
search
search boxbox on
on their
their
relevant. site.
site. It
It generates
generates up
up to
to
70% more clicks
70% more clicks thanthan
For Ad Network Aggregators, Mobile Commerce provides a search
a
a raw
raw search
search engine
engine
feed. This means that Ad Network Aggregators can offer their
implementation
implementation.
customers both Display Advertising and Search Advertising.

All customers have access to on-line reporting which has been used
by most customers to improve their portals by adding terms/services
that are regularly searched making their portals more relevant to their
specific audience.

Mobile Commerce also supplies dynamic tag clouds so within a set of

results for a subject like “games” a list of the most search games
(Pacman, Tetris etc) can be displayed making it much easier for users
to select a game that they are probably familiar with and other users
have been searching.

The service is available for UK, USA, Germany, Japan and Ireland.

18
 Press enquiries:

Keith Wildman
PagetBaker
keith@pagetbaker.com
+44 (0) 7889 158 060

39 Great Russell St
London WC1B

www.pagetbaker.com
facebook/pagetbaker
twitter.com/pagetbakeruk

Contact details:

Mobile Commerce Ltd

info@mobilecommerce.co.uk
+44 (0)1285 883 050

103 Cirencester Business Park

Love Lane
Cirencester
Glos GL7 1XD
United Kingdom

www.mobilecommerce.co.uk
twitter.com/sjspage – on mobile search terms and usage

19