2. Agenda
SearchSense Team
The Market
Problem
Impact
SearchSense Business Proposition
Business Proposition Hypothesis
Experiment
Hypothesis Results
Demo
User Behavior Adaptation
3. SearchSense Team
Naveen Grover (Team Lead) naveengrover2000@gmail.com
Hery Zo Rakotondramanana heryzo@gmail.com
Lokesh KUMAR email.lokesh.kumar@gmail.com
Sonya Fox sonya.fox@gmail.com
Lem Griffin lemuel.griffin@gmail.com
Kamal Pasha Shaikh shaikhkamal@gmail.com
Samir Carecho digiwise@gmail.com
Jonathan Tanner ronin@ronindesign.net
4. The Market | Non Search Engine Market
Although Search Engines Are Still the Major Players, Search Isn’t Confined to the Engines Anymore
Site search (non-search engine search) continues to show growth, but still represents a smaller percentage of
overall searches.
Source: comScore qSearch 2.0; “Search Engines” defined as properties falling under the Search/Navigation category in qSearch , January 2011
5. The Market | Long Tail of Search
Addressing discoverability Seventy percent of queries in commercial searches are “long tail”
queries, the sheer magnitude of which defies the labor-intensive efforts used for “head” terms.
http://www.seomoz.org/blog/illustrating-the-long-tail
6. The Market | Smart Phone Local Search and Buying Behavior
Source : 5th Annual 15miles/Localeze Local Search Usage Study Conducted by comScore February, 2012
7. The Market | Primary Source for Local Business Information
Consumers utilize multiple media sources when conducting a local search.
15miles/comScore Local Search Usage Study, 2010
8. The Market | Reasons for Dissatisfaction
Local Business Information Provided
Source: 15miles/comScore Local Search Usage Study, 2010
9. The Market | Site Search & Shopping Behavior
Source : The Value of Retail Search and Position Conducted by comScore Sponsored by Searchandise Commerce and iProspect July 2010
10. The Market | User Behaviors - Number of Clicks
The Value of Retail Search and Position Conducted by comScore Sponsored by Searchandise Commerce and iProspect July 2010
11. The Market | Site Search Page Perception of Top Section
Source : The Value of Retail Search and Position Conducted by comScore Sponsored by Searchandise Commerce and iProspect July 2010
12. Problem | The Predicament
From bad to worse – a small problem takes the user way off track
Source : India famous IndiaTimes Shopping Site , http://shopping.indiatimes.com/ Accessed 18 May 2012
13. Problem | The Predicament
OR
People Database
Brad Pitt
Queries against collection: Forest Whittacker
Find all entries for “Forrest Whitaker” George Bush
Find all entries for “Arnold Angelina Jolie
Schwarzenegger” Arnold Schwarzeneger
Find all entries for “Brittany Spears”
…
…
…
14. Problem | The Predicament
Search Problem
Actual queries gathered by Google*
Correct Spelling 488941
Incorrect Spelling 146258
% of Problem 30%
Errors in queries
Errors in data
Search - Bring query and meaningful results
* Source www.google.com/jobs/britney.html closer together
15. Problem | The Predicament
Search failure isn’t pretty… From bad to worse – a small problem takes the user way off
track
There is only one right way to search, but many wrong way
… Which side you are !!!
16. Problem | The Predicament
Does this sound familiar?
It’s either all or nothing - zero results or thousands
Your site statistics show that many users try a few searches then
exit the site without any other action.
In an online store, visitors frequently do not find what they are
looking for, even when the item is available. The problem is
usually caused by the eCommerce on-site search.
Users who found something useful by browsing on a previous
visit can’t find it again when they return and use search. They are
especially frustrated because they know the page exists.
17. Problem | Impact
eCommerce Site Search - the impact of negative search experiences
Poor search = lost revenue
Users who conduct site searches are almost three times more likely to purchase something
while visiting a site – (WebSideStory study)
Users who had SUCCESSFUL site searches are twice as likely to convert (Enlighten study)
Half of all add-to-cart actions happened after a search. (Enlighten study)
Users who had NULL-RESULTS site searches were three times as likely to leave (Enlighten
study)
Estimate - Up to 20% of the gains in user experience during a site redesign can be attributed to
search improvements – “Laura Ramos of Forrester”
Alternative Search Properties - Apple moves into the top five with 120% intensity growth and
14% searcher growth over the past year. eBay leads with 855MM searches and 19.0 searches
per searcher in January 2011; Facebook.com has the largest number of searchers (65MM).
(comScore qSearch State of Search, January 2011)
Site features and overall site performance strongly affect online shopper loyalty. Online shoppers
also look for rich product content, with 67% of online consumers finding this important, while
60% of consumers demand an effective site search. (August 17, 2009 eCommerce Web Site
Performance Today An Updated Look At Consumer Reaction To A Poor Online Shopping
Experience A commissioned study conducted by Forrester Consulting on behalf of Akamai
Technologies, Inc.)
18. SearchSense Business Proposition
SearchSense fill the “linguistic gap”, which improve
the “Findability”. It simply turns more of site visitors
into buyers which means higher conversion rate,
larger orders and higher revenue.
19. Business Proposition Hypothesis
We have tested our Value Proposition for the following Hypothesis
HA 1: Reduction of Failed Searches
“Subjects using the SearchSense models will initiate fewer searches that fail to return an
acceptable selection than subjects using any other search tool.”
HA 2: Reduction of Search Refinements
“Subjects using the SearchSense models will require fewer search refinements to locate the
intended result items than subjects using any other search tool.”
HA 3: Reduction of Search Time
“Subjects using the SearchSense will take less time to complete the assigned searches than
subjects using any other search tool.”
HA 4: No Significant Negative Impact on Precision
“Subjects using the SearchSense will not receive significantly more search results than
subjects using any other search tool.”
20. Experiment
The primary research objective, to measure the effect of
SearchSense, is accomplished by conducting an
experiment in which a demo of SearchSense system
(minimum viable product version) is provided to group
of human subjects and observation being made and
verbal feedback is collected to evaluate the performance
of SearchSense system. Subject is asked to perform the
same misspell search (linguistic problem) to other
publicly available search system as well e.g.
http://shopping.indiatimes.com to compare the results
21. Hypothesis HA 1: Reduction of Failed Searches
HA 1: Reduction of Failed Searches
“Subjects using the SearchSense models will initiate fewer searches
that fail to return an acceptable selection than subjects using any
other search tool.”
Result
Analysis of the observation results demonstrates that for subjects
using SearchSense system, the relevant results are shown even
when subject misspell due to linguistic problem than subjects using
any other publicly available search system e.g.
http://shopping.indiatimes.com.
22. Hypothesis HA 2: Reduction of Search Refinements
HA 2: Reduction of Search Refinements
“Subjects using the SearchSense models will require fewer search
refinements to locate the intended result items than subjects using
any other search tool.”
Result
Analysis of the observation results demonstrates that for subjects
using SearchSense system, the relevant results are shown even
when subject misspell due to linguistic problem hence subject would
require fewer search refinements to locate the intended result items
than subjects using any other publicly available search system e.g.
http://shopping.indiatimes.com.
23. Hypothesis HA 3: Reduction of Search Time
HA 3: Reduction of Search Time
“Subjects using the SearchSense will take less time to complete the
assigned searches than subjects using any other search tool.”
Result
Analysis of the observation results demonstrates that for subjects
using SearchSense system, the relevant results are shown even
when subject misspell due to linguistic problem hence subject would
require fewer search refinements to locate the intended result items
than subjects using any other publicly available search system e.g.
http://shopping.indiatimes.com , which causes reduction in search
time per identified search term.
24. Hypothesis HA 4: No Significant Negative Impact on
Precision
HA 4: No Significant Negative Impact on Precision
“Subjects using the SearchSense will not receive significantly more
search results than subjects using any other search tool.”
Result
We were unable to test this hypothesis as no publicly data available
against which we can test our hypothesis.
25. Demo | SearchSense
One Text Box for Category or Product
Relevant Category and Product First
Search Suggestion can be turned off
Products Listing
Product or Category
Products Category
Last Search Items Shown
Search Suggest Can be turned off
26. User Behavior Adaptation
Adoption is ramping up quickly, but Instant is still only engaged on approximately 22% of queries
Source : comScore qSearch State of Search, January 2011
27. User Behavior Adaptation
Instant appears to be driving users toward shorter queries
Instant is clearly reducing users’ workload, but it’s also shortening the average query
Source : comScore qSearch State of Search, January 2011
Good e-commerce searches do more than just retrieve the right products. They must present the shopper with the right amount of descriptive information about the product to enable selection and decision making. The study showed that during the holidays (a peak time for mission-driven visits as opposed to casual browsing visits) “site searchers converted at a rate of 7.54% while all visitors converted at 2.79%.”