JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good...

15
JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1

Transcript of JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good...

Page 1: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

JANE LI , SCOTT B. HUFFMAN, AND AKIHITO TOKUDAJULY 2009

PRESENTED BY : GAURANG JHAWAR

Good Abandonment in Mobile and PC Internet

Search1

Page 2: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

What is Good Abandonment?

An Abandoned query for which user`s information need was successfully addressed by search result page with no need to click on a result or refine the query.

In simple words, it means the snippets that we see under the search results were good enough to provide what the user wanted.

How does Google make rich snippets? 1. Pickup a HTML markup format ( Eg : Microdata, RDFa ) 2. Markup your content ( Google supports Reviews, people, Recipes, Music etc ) 3. Test your markup ( Using Structured data testing tool ) Or use Data Highlighter is a webmaster tool for teaching Google about the pattern of event-related data on your website using a mouse ( Don`t have to consistently update HTML markup).

Eg : If I am looking for the creators of Batman, I have the Answer in the snippet

2

Page 3: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

Rich Text Snippets

An example of good search abandonment would be to Google a business with the intention of finding the address. The Google result returns the street address, the business telephone number, and even a summary of starred reviews when applicable -- all the information the user needs to make a reservation and find the location on a map. This user won’t be clicking through to the website, but that doesn’t mean that the search has failed.

3

Page 4: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

Why and Where?

To approximate the prevalence of good abandonment and the types of information need that may lead to one.

Conducted across two modalities PC, Mobile devices for three countries USA, CHINA, JAPAN because as we all know results for mobile and desktop queries are different based on Local results, Android users.

Google's aim is to give the results that are most relevant to the person searching. That's why they use your searching history and location, where they can, to personalize the results.

4

Page 5: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

What did they find?

Queries potentially indicating good abandonment are a significant portion of all abandoned queries. How to find potential queries?

Good Abandonment rate for Mobile is higher than PC across all countries.

Why?

Good Abandonment vary by both, locale and modality. Why?

Hold on to your questions, coz they are about to be answered

5

Page 6: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

Why is Good Abandonment so Important?

Abandoning is usually considered an indicator of user dissatisfaction coz he did not click on any result or did not make an refinement to the query. Which no search engine wants.

Mobile devices :-1. Are clunky, slow, have formatting, content omission issues.2. Need quicker/handier responses (if provided leads to good

abandonment)3. Inherently local to find local address, phone no. etc

And hence every major search engine out there wants to reduce the load on its server and keep the user happy by providing him the desired result as soon and accurately as possible.

6

Page 7: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

Related Studies

Study 1 : Eye-tracking where tasks are divided into navigational and informational tasks. Shows that time spent on a task was reduced when the snippet length was increased.

Study 2 : Analysis of Google XHTML and PDA interface logs to show that less sophisticated user write shorter queries. ( Finding patterns on mobile devices )

Study 3 : European mobile users had shorter average query length, rare use of advanced search, limited search vocabulary and repeated queries.

Study 4 : Search usage is much more focused for average mobile user than average PC user.

7

Page 8: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

Definitions

‘Potential Good Abandonment’ which should mostly be achieved by a search engine result page if there is a dominant information need associated with that query. Is judged by looking at the query itself rather than Google`s result page to generate an upper bound of good abandonment. ( Which the authors guessed would be there in the results page )

In cases where query is Ambiguous but atleast one strong interpretation can be made it is called ‘Maybe Potential Good Abandonment’. Eg: Apple can be interpreted as the company or the fruit – Content based filtering.

‘Likely Good Abandonment’ to make sure if the guess was right after looking at the actual Google results.

(Sampled 400 abandoned queries from Japan, US and 1000 from China for both mobile and PC from Google`s PC and Mobile search logs from a week in Sept-Oct 08)

8

Page 9: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

Task categories and their definition

The authors coded each potential or likely good query on the basis of information need

They compiled a list and found that the prevalence and likelihood of good abandonment across the different categories varied significantly both by locale and modality.

9

Page 10: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

Upper Bound Estimate of Good Abandonment

It refers to the proportion of abandoned queries classified as ‘YES’ or ‘MAYBE’ a potential good abandonment out of all abandoned queries.

Observations – Japan has lowest potential abandonment

for mobile as it is considered to be a more mature market implying larger diversity in search topics and more depth of user`s information need.

Higher potential good abandonment rate in mobile search :-

1. Retrieving web pages is a clumsy experience, hence users mostly have simple searches like weather report, stock quotes and they expect the results on the result page with no more furthur clicking.

2. Mobile searches are dependent on activity at the time, current location, conversation with others which are usually “Quick answer type searches”.

10

Page 11: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

Likely Abandonment Rates

• Is measure of How often search engines provide results, that result in good abandonment for the user, out of all the queries that could potentially lead to good abandonment i.e. how often does the user actually will get the desired result.• YES - queries that actually turned out to be good abandonments

MAYBE - authors weren`t sure if the result pages would have satisfied the user`s information need NO - author thought would be good abandonments but turns out that the information need was not addressed on the result page on that day.

• For mobile search 70% of the potential good abandonments were met on the result page compared to 56% for PC because of less complex need.

• Over 80% potential good abandonments are answered in Chinese mobile searches because of the relatively simple information need.

11

Page 12: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

Some Interesting Results

China`s local search results are so because the author`s found out that the actual search results were poorly addressed on the results page.

Hence, on mobile devices where text entry is difficult and loading takes longer people may usually give up.

Answers are got from advanced search features like Calculator, Definition and News headlines and betterment of search engines have lead to higher statistics for this category.

Fig : Potential good abandonment vs Locale

12

Page 13: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

Some Interesting Results

The authors found that in US users click more often when searching celebrities while Japanese mobile users apparently browse images or news outlines and hence are satisfied by the snippets.

For Stocks, we get more good abandonments in mobile because usually user looks for stock quotes which he gets in the snippets, whereas in PC user may seek more information and graphs.

China has most simple mobile searches for weather. The author correlates this to China being the less mature mobile search markets compared to the other two locales.

13

Page 14: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

Opportunities For Additional Good Abandonment

In the figure the darker the cell, the more headroom/scope it has for improvement.

Categories like Celebrities, Local, Answer have scope for improvement.

The search engines should give directly address “Local”, “Image” information needs by adding more features like “Shortcut Insertions”.

The “Answer” category requires more intelligent snippets and for mobile users it should be longer since loading a webpage is a hasseel.

14

Page 15: JANE LI, SCOTT B. HUFFMAN, AND AKIHITO TOKUDA JULY 2009 PRESENTED BY : GAURANG JHAWAR Good Abandonment in Mobile and PC Internet Search 1.

Conclusion & Future Work

The authors used their judgement to classify queries into potential or likely good abandonment buckets, but It would be ideal to conduct a study with real users to understand the real information needs.

Query abandonment should not be considered as a negative signal as seen.

Evaluation models derived from CLICKS should be improved by taking Good abandonment in account.

Search engines still have headroom for good abandonment.

Can be improved by providing right information on the result page especially for mobile users because user wants the result in the snippets.

Hence, the snippet length and quality should be increased.

15