Attracting Buyers with Search, Semantic, and Recommendation

Attracting Buyers with Search, Semantic, and Recommendation

Technology

Prepared by Dr. Derek Sedlack, South University

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Learning Objectives

Organic Search and Search Engine Optimization

Pay-Per-Click and Paid Search Strategies

A Search for Meaning—Semantic Technology

Recommendation Engines

Using Search Technology for Business Success

Using Search Technology for Business Success

How Search Engines Work

Search Engine: an application for locating webpages or other content on a computer network using spiders.

Spiders: web bots (or bots); small computer programs designed to perform automated, repetitive tasks over the Internet.

Bots scan webpages and return information to be stored in a page repository.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Using Search Technology for Business Success

Chapter 6

Crawler search engine

Web directory

Hybrid search engine

Meta-search engine

Semantic search engine

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Using Search Technology for Business Success

Web Directories

Typically organized by categories.

Webpage content is usually reviewed by directory editors prior to listing.

Page Repository: data structure that stores and manages information from a large number of webpages, providing a fast and efficient means for accessing and analyzing the information at a later time.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Using Search Technology for Business Success

Chapter 6

Figure 6.5 Components of crawler search engines (Grehan, 2002).

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Using Search Technology for Business Success

Chapter 6

Figure 6.6 Search engines use invested indexes to efficiently locate Web content based on search query terms.

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Using Search Technology for Business Success

Why Search is Important for Business

Enterprise search tools allow organizations to share information internally.

An organizations’ ability to share knowledge among employees is vital to its ability to compete.

Information is not always in the same format.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Using Search Technology for Business Success

Why Search is Important for Business

Structured data: information with a high degree of organization, such that inclusion in a relational database is seamless and readily searchable by simple, straightforward search engine algorithms or other search operations.

Unstructured data: “messy data” not organized in a systematic or predefined way.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Using Search Technology for Business Success

Security Issues

Limited access to certain data via job function or clearance.

Request log audits should be conducted regularly for patterns or inconsistencies.

Enterprise Vendors

Used to treat data in large companies like Internet data but include information management tools.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Using Search Technology for Business Success

Recommendation Engines

Attempt to anticipate information users might be interested in to recommend new products, articles, videos, etc.

Search Engine Marketing

A collection of online marketing strategies and tactics that promote brands by increasing their visibility in search engine results pages (SERPs) through optimization and advertising.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Using Search Technology for Business Success

Search Engine Marketing

Basic search types:

Informational search

Navigational search

Transactional search

Strategies and tactics produce two outcomes:

Organic search listings

Paid search listings

Pay-per-click (produce click-through rates)

Social media optimization

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Using Search Technology for Business Success

Mobile Search

Technically configured mobile sites

Content designed for mobile devices

Business search

Focused search

Filetype

Advanced search

Search tools button

Search history

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Using Search Technology for Business Success

Real-time Search

Google Trends

Google Alerts

Twitter Search

Social Bookmarking Search

Page links tagged with keywords

Specialty Search: Vertical Search

Programmed to focus on webpages related to a particular topic and to drill down by crawling pages that other search engines are likely to ignore.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Using Search Technology for Business Success

What is the primary difference between a web directory and a crawler based search engine?

What is the purpose of an index in a search engine?

Describe the page-ranking method most commonly associated with Google’s success.

What is the difference between search engine optimization and PPC advertising?

Describe three different real-time search tools.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Suggested Answers:

1. Crawler search engines rely on sophisticated computer programs called “spiders,” “crawlers,” or “bots” that surf the Internet, locating webpages, links, and other content that are then stored in the SE’s page repository.

 

Web directories are categorized listings of webpages created and maintained by humans. Because websites are only included after being reviewed by a person, it is less likely that search results will contain irrelevant websites.

 

2. An index helps search engines efficiently locate relevant pages containing keywords used in a search.

 

3. The methods by which search engines determine page rank vary and the specific algorithms they use are often carefully guarded trade secrets. In some cases, a search engine may use hundreds of different criteria to determine which pages appear at the top of a SERP. Google, for instance, claims to use over 200 “clues” to determine how it ranks pages (Google.com, 2014). According to Dover and Dafforn (2011), all these factors can be grouped into two categories: relevance and popularity (or “authority”).

 

4. Businesses utilize search engine optimization (SEO) to improve their website’s organic listings on SERPs. No payments are made to the search engine service for organic search listings.

 

Pay-per-click (PPC) advertising refers to paid search listings where advertisers pay search engines based on how many people click on the ads.

5. Google Trends—Trends (google.com/trends) will help you identify current and historical interest in the topic by reporting the volume of search activity over time. Google Trends allows you to view the information for different time periods and geographic regions.

Google Alerts—Alerts (google.com/alerts) is an automated search tool for monitoring new Web content, news stories, videos, and blog posts about some topic. Users set up alerts by specifying a search term (e.g., a company name, product, or topic), how often they want to receive notices, and an e-mail address where the alerts are to be sent. When Google finds content that match the parameters of the search, users are notified via e-mail. Bing has a similar feature called News Alerts.

 

Twitter Search—You can leverage the crowd of over 650 million Twitter users to find information as well as gauge sentiment on a wide range of topics and issues in real time. Twitter’s search tool (twitter.com/search-home) looks similar to other search engines, and includes an advanced search mode.

15

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Learning Objectives

Organic Search and Search Engine Optimization

Pay-Per-Click and Paid Search Strategies

A Search for Meaning—Semantic Technology

Recommendation Engines

Using Search Technology for Business Success

Organic Search and Search Engine Optimization

Search Engine Optimization

Keyword conversion rates: the likelihood that using a particular keyword to optimize a page will result in conversions*.

Ranking factors

Reputation or popularity

PageRank: Google’s algorithm based on the assumption that people are more likely to link a high-quality website than poor-quality site.

Backlinks: external links that point back to a site.

Relevancy

User Satisfaction

Conversions: when a website visitor converts to a buyer

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Organic Search and Search Engine Optimization

Inbound marketing

An approach to marketing that emphasizes SEO, content Marketing, and social media strategies to attract customers.

Outbound marketing

Traditional approach using mass media advertising.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Organic Search and Search Engine Optimization

Black Hat SEO

Gaming the system or tricking search engines into ranking a site higher than its content deserves.

Link spamming: generating backlinks toward SEO, not adding user value.

Keyword tricks: embedded high-value keywords to drive up traffic statistics.

Ghost text: text hidden in the background that will affect page ranking

Shadow (ghost or cloaked) pages: created pages optimized to attract lots of people through redirect.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Organic Search and Search Engine Optimization

Search engines use many different “clues” about the quality of a website’s content to determine how a page should be ranked in search results. These clues fall into three primary categories: Reputation or Popularity, Relevancy, and User Satisfaction. Explain the rationale for using each of these three categories as an indicator of a website’s content quality.

Backlinks were a key factor in Google’s original PageRank algorithm. Explain what a backlink is and why Google has reduced its emphasis on backlinks and instead uses many other additional factors in its ranking algorithm?

Explain why so-called black hat SEO tactics are ultimately short-sighted and can lead to significant consequences for businesses that use them.

How do organizations evaluate the effectiveness of their search engine optimization (SEO) strategies and tactics?

Explain why providing high quality, regularly updated content is the most important aspect of any SEO strategy.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Suggested Answers:

1. One way of assessing the quality of a website is to use measures of popularity. This is based on the assumption that websites with good content will be more popular than sites with poor quality content. On the assumption that people are more likely to link to high-quality websites than poor-quality sites, one measure of popularity is the number of backlinks—external links that point back to a site.

 

Search engines attempt to determine if the content on a webpage is relevant to what the searcher is looking for. As with quality, the search engine cannot determine relevance directly, so algorithms have been developed to look for clues that suggest a site might be relevant. Factors which affect relevancy:

Keywords related to the search topic suggest relevant content.

Page titles: Words in the page title that are related to the topic suggest relevant content.

Relevant phrases in text: In addition to keywords, search engines look at the words and phrases on the page to determine relevance.

Amount of text on page that appears relevant: The proportion of relevant text to non-relevant text can influence relevance.

Backlinks from relevant sites and Web directories: Webpages that are listed in relevant categories of Web directories are more likely to be relevant because they were reviewed by human editors.

SERP click through rate (CTR): Searchers are more likely to click on listings that contain relevant content.

Onpage factor: Metadata (such as page titles, page descriptions) and descriptive URLs should reflect the page content. People use the information in search listings to determine if a link contains relevant information. This affects CTR.

Dwell time and bounce rate are impacted by how relevant a website’s content is. Long dwell times and short bounce rates suggest relevant content related to the search.

Search engines want their customers to be satisfied. As a result, SERP ranking is influenced by factors that impact user satisfaction. Factors that are likely to influence a search engine’s user satisfaction rating are:

Dwell time: Users that stay on a site longer are probably more satisfied.

Site speed: Slow page loading time on websites reduces satisfaction.

Reading level: Reading levels that are too high or too low frustrate users.

Hacked sites, malware, spam reduce user satisfaction significantly.

Website satisfaction surveys: Google created user satisfaction surveys that webmasters can embed in their websites. Positive responses to these surveys can improve ranking.

Barriers to content: Making people register, provide names, or fill out forms to get to content has a negative impact on user satisfaction.

Other factors: Too many ads, page-not-found errors, duplicate content/pages, content copied from other websites, and spam in comment sections all detract from user satisfaction.

 

2. A backlinks is an external link that points back to a site. Google has changed its ranking methods and the assumption is that backlinks are still very important, but not weighted as heavily as they used to be. Using a number of other factors yields a better picture of the relevance of sites to the search.

 

3. Black hat tactics try to trick the search engine into thinking a website has high-quality content, when in fact it does not. The search engines have stronger detection systems in place and when they are discovered, Google and other SEs will usually punish the business by dramatically lowering the website’s rank so that it does not show up on SERPs at all.

 

4. Using Web analytics programs like Google Analytics, companies can determine how many people visit their site, what specific pages they visit, how long they spend on the site, and what search engines are producing the most traffic. More sophisticated SEO practitioners will also analyze keyword conversion rates, or the likelihood that using a particular keyword to optimize a page will result in conversions (i.e., when a website visitor converts to a buyer).These are just a few of the many metrics used to measure the effectiveness of SEO strategies.

5. Perhaps the most important action an organization can take to improve its website’s ranking and satisfy website visitors is provide helpful content that is current and updated regularly. When SEO practices are combined with valuable content, websites become easier to find in search engines but, more importantly, contribute to building brand awareness, positive attitudes toward the brand, and brand loyalty.

20

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Learning Objectives

Organic Search and Search Engine Optimization

Pay-Per-Click and Paid Search Strategies

A Search for Meaning—Semantic Technology

Recommendation Engines

Using Search Technology for Business Success

Pay-Per-Click and Paid Search Strategies

Pay-Per-Click

PPC advertising campaigns:

Set an overall budget

Create ads

Select associated keywords

Set up billing account information

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Pay-Per-Click and Paid Search Strategies

Paid Search Advertising Metrics

Click through rates (CTR): used to evaluate keyword selection and ad copy campaign decisions.

Keyword conversion: should lead to sales, not just visits.

Cost of customer acquisition (CoCA): amount of money spent to attract a paying customer.

Return on advertising spend (ROAS): overall financial effectiveness.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Pay-Per-Click and Paid Search Strategies

Quality Score

Determined by factors related to the user’s experience.

Expected keyword click-through-rate (CTR)

The past CTR of your URL (web address)

Past effectiveness

Landing page quality

Relevance of keywords to ads

Relevance of keywords to customer search

Ad performance on difference devices

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Pay-Per-Click and Paid Search Strategies

What would most people say is the fundamental difference between organic listings and PPC listings on a search engine?

What are the four primary steps to creating a PPC advertising campaign on search engines?

In addition to the “bid price” for a particular keyword, what other factor(s) influence the likelihood that an advertisement will appear on a search results page? Why don’t search engines just rely on the advertisers bid when deciding what ads will appear on the search results page?

How do webpage factors influence the effectiveness of PPC advertisements?

Describe four metrics that can be used to evaluate the effectiveness of a PPC advertising campaign.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Suggested Answers:

1. Paid advertisements receive preferential page placement, but most major search engines differentiate organic search results from paid ad listings on SERPs with labels, shading, and placing the ads in a different place on the page.

 

2. There are four steps to creating a PPC advertising campaign on SEs.

1. Set an overall budget for the campaign.

2. Create ads—most search engine ads are text only.

3. Select keywords associated with the campaign.

4. Set up billing account information.

 

3. In addition to selecting keywords and setting bid prices, advertisers also set parameters for the geographic location they want their ad to appear in and time of day. These factors allow for additional customer targeting designed to help advertisers reach the consumers most likely to purchase their products.

 

A quality score is determined by factors related to the user’s experience. Ads that are considered to be more relevant (and therefore more likely to be clicked on) will cost less and more likely run in a top position.

 

Relevant ads are good for all parties—the search engine makes more money from clicked ads, the advertiser experiences more customers visiting its site, and the customer is more likely to find what he or she is looking for.

 

4. The effectiveness of PPC ads is heavily influenced by factors on the webpages that ads are linked to. For instance, sometimes companies create product-oriented ads, but then link to the main page of their website instead of a page with information about the product in the ad. Other factors include landing page design, effectiveness of the call to action, and the quality of the shopping cart application. A PPC campaign will not be very effective if the website is not attractive to consumers once they reach it.

 

5. Click through rates (CTRs)—By themselves, CTRs do not measure the financial performance of an ad campaign. But they are useful for evaluating many of the decisions that go into a campaign, such as keyword selection and ad copy.

 

Keyword conversion—High CTRs are not always good if they do not lead to sales. Since the cost of the campaign is based on how many people click an ad, you want to select keywords that lead to sales (conversions), not just site visits. PPC advertisers monitor which keywords lead to sales and focus on those in future campaigns.

 

Cost of customer acquisition (CoCA)—This metric represents the amount of money spent to attract a paying customer. To calculate CoCA for a PPC campaign, you divide the total budget of the campaign by the number of customers who purchased something from your site. For instance, if you spent $1,000 on a campaign that yielded 40 customers, your CoCA would be $1,000/40 5 $25 per customer.

 

Return on advertising spend (ROAS)—The campaign’s overall financial effectiveness is evaluated with ROAS (revenue /cost). For example, if $1,000 was spent on a campaign that led to $6,000 in sales, ROAS would be $6,000/$1,000 5 $6. In other words, for every dollar spent on PPC ads, $6 was earned.

25

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Learning Objectives

Organic Search and Search Engine Optimization

Pay-Per-Click and Paid Search Strategies

A Search for Meaning—Semantic Technology

Recommendation Engines

Using Search Technology for Business Success

A Search for Meaning—Semantic Technology

Semantic Web

Meaningful computing using metadata: application of natural language processing (NLP) to support information retrieval, analytics, and data-integration that compass both numerical and “unstructured” information.

Semantic Search

Process of typing something into a search engine and getting more results than just those that feature the exact keyword typed into the search box.

Metadata

Data that describes and provides information about other data.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

A Search for Meaning—Semantic Technology

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

A Search for Meaning—Semantic Technology

Web 3.0

Developed by W3C.

Resource description framework (RDF)

Used to represent information about resources

Web ontology language (OWL)

Language used to categorize and accurately identify the nature of Internet things

SPARCQL protocol

Used to write programs that can retrieve and manipulate data scored in RDF

RDF query language (SPARCQL)

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

A Search for Meaning—Semantic Technology

Semantic Search Features and Benefits

Related searches/queries

Reference results

Semantically annotated results

Full-text similarity search

Search on semantic/syntactic annotations

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

A Search for Meaning—Semantic Technology

Semantic Search Features and Benefits

Concept search

Ontology-based search

Semantic Web search

Faceted search

Clustered search

Natural language search

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

A Search for Meaning—Semantic Technology

List five different practical ways that semantic technology is enhancing the search experience of users.

How do metadata tags facilitate more accurate search results?

Briefly describe the three evolutionary stages of the Internet?

Define the words “context,” “personalization,” and “vertical search” and explain how they make for more powerful and accurate search results.

What are the three languages developed by the W3C and associated with the semantic Web?

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Suggested Answers:

1. Grimes (2010) provides a list of practical benefits that could result from semantic search technology:

Related searches/queries. The engine suggests alternative search queries that may produce information related to the original query. Search engines may also ask you, “Did you mean: [search term]?” if it detects a misspelling. (This already happens with some.)

Reference results. The search engine suggests reference material related to the query, such as a dictionary definition, Wikipedia pages, maps, reviews, or stock quotes.

Semantically annotated results. Returned pages contain highlighting of search terms, but also related words or phrases that may not have appeared in the original query. These can be used in future searches simply by clicking on them.

Full-text similarity search. Users can submit a block of text or even a full document to find similar content.

Search on semantic/syntactic annotations. This approach would allow a user to indicate the “syntactic role the term plays—for instance, the part-of-speech (noun, verb, etc.)—or its semantic meaning—whether it’s a company name, location, or event.” For instance, a keyword search on the word “center” would produce too many results. Instead, a search query could be written using a syntax such as the following:

<organization> center </organization>

This would only return documents where the word “center” was part of an organization’s name. Google currently allows you to do something similar to specify the kind of files you are looking for (e.g., filetype:pdf)

Concept search. Search engines could return results with related concepts. For instance, if the original query was “Tarantino films,” documents would be returned that contain the word “movies” even if not the word “films.”

Ontology-based search. Ontologies define the relationships between data. An ontology is based on the concept of “triples”: subject, predicate, and object. This would allow the search engine to answer questions such as “What vegetables are green?” The search engine would return results about “broccoli,” “spinach,” “peas,” “asparagus,” “Brussels sprouts,” and so on.

Semantic Web search. This approach would take advantage of content tagged with metadata as previously described in this section. Search results are likely to be more accurate than keyword matching.

Faceted search. Faceted search provides a means of refining results based on predefined categories called facets. For instance, a search on “colleges” might result in options to “refine this search by. . .” location, size, degrees offered, private or public, and so on. Faceted search tools available today tend to focus on a specific domain, such as Wikipedia or Semidico, a search tool for biomedical literature.

Clustered search. This is similar to a faceted search, but without the predefined categories. Visit Carrot2.org to better understand this concept. After conducting a search, click on the “foamtree” option to see how you can refine your search. The refining options are extracted from the content in pages of the initial search.

Natural language search. Natural language search tools attempt to extract words from questions such as “How many countries are there in Europe?” and create a semantic representation of the query. Initially, this is what people hoped search engines would evolve toward, but Grimes wonders if we have become so accustomed to typing just one or two words into our queries that writing out a whole question may seem like too much work.

 

2. Much of the world’s digital information is stored in files structured so that they can only be read by the programs that created them. With metadata, the content of these files can be labeled with tags describing the nature of the information, where it came from, or how it is arranged, essentially making the Web one large database that can be read and used by a wide variety of applications.

 

The semantic Web will make it possible to access information about real things (people, places, contracts, books, chemicals, etc.) without worrying about the details associated with the nature or structure of the data files, pages, and databases where these things are described or contained (Hendler and Berners-Lee, 2010).

 

3. The first stage was Web 1.0 (The Initial Web) – A Web of Pages. Pages or documents are “hyperlinked,” making it easier than ever before to access connected information.

The first stage was Web 2.0 (The Social Web) – A Web of Applications. Applications are created that allow people to easily create, share, and organize information.

 

The third stage is Web 3.0 (The Semantic Web) – A Web of Data. Information within documents or pages is tagged with metadata, allowing users to access specific information across platforms, regardless of the original structure of the fi le, page, or document that contains it. It turns the Web into one giant database.

 

4. Context defines the intent of the user; for example, trying to purchase music, to find a job, to share memories with friends and family

 

Personalization refers to the user’s personal characteristics that impact how relevant the content, commerce, and community are to an individual.

 

Vertical search, as you have read, focuses on finding information in a particular content area, such as travel, finance, legal, and medical.

 

The current Web is disjointed, requiring us to visit different websites to get content, engage in commerce, and interact with our social networks (community). The future Web will use context, personalization, and vertical search to make content, commerce, and community more relevant and easier to access (Mitra, 2007).

 

5. The semantic Web utilizes additional languages that have been developed by the W3C. These include resource description framework (RDF), Web ontology language (OWL), and SPARQL protocol and RDF query language (SPARQL).

32

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Learning Objectives

Organic Search and Search Engine Optimization

Pay-Per-Click and Paid Search Strategies

A Search for Meaning—Semantic Technology

Recommendation Engines

Using Search Technology for Business Success

Recommendation Engines

Recommendation Filters

Content-based filtering: products based on product features in past interactions.

Collaborative filtering: based on user’s similarity to other people.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Recommendation Engines

Limitations of Recommendation Engines

Cold start or new user: challenging since no starting point or preexisting information exists.

Sparsity: unable to create critical mass due to few ratings or similar groups are unidentifiable.

Limited feature content: manual information entry is prohibitive where there are many products.

Overspecialization: narrowly configured results may only recommend the same item, but in different sizes or colors.

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Recommendation Engines

Hybrid Recommendation Engines

Weighted hybrid: results from different recommenders are assigned weight and combined numerically to determined final recommendations.

Mixed hybrid: results from different recommenders presented along-side of each other.

Cascade hybrid: results from different recommenders assigned a rank or priority.

Mixed hybrid: results from different recommenders combines results from two recommender systems from the same technique category.

Chapter 5

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Recommendation Engines

How is a recommendation engine different from a search engine?

Besides e-commerce websites that sell products, what are some other ways that recommendation engines are being used on the Web today?

What are some examples of user information required by recommendation engines that use collaborative filtering?

Before implementing a content-based recommendation engine, what kind of information would website operators need to collect about their products?

What are the four distinct methodologies used by recommender systems to create recommendations?

What is a recommendation engine called that combines different methodologies to create recommendations? What are three ways these systems combine methodologies?

Chapter 6

Copyright © 2015 John Wiley & Sons, Inc. All rights reserved.

Suggested Answers:

1. With a search engine, customers find products through an active search, assuming customers know what they want and how to describe it when forming their search query.

Recommendation engines proactively identify products that have a high probability of being something the consumer might want to buy. Each time customers log into the site, they are presented with an assortment of products based on their purchase history, browsing history, product reviews, ratings, and many other factors.

 

2. Netflix does recommendations of movies for customers similar to movies they already have watched.

Pandora creates recommendations or playlists based on song attributes.

 

3. Many collaborative filtering systems use purchase history to identify similarities between customers. In principle, however, any customer characteristic that improves the quality of recommendations could be used, such as patterns of consumer behavior, interests, ratings, reviews, social media contacts and conversations, media use, financial information, and so on.

 

4. Answers may vary.

Content-based filtering recommends products based on the product features of items the customer has interacted with in the past and the similarity to other products’ features.

 

5. Content filtering, collaborative filtering, knowledge-based systems, and demographic systems.

 

6. Hybrid recommendation engines develop recommendations based on some combination of the methodologies described (content-based filtering, collaboration filtering, knowledge-based, and demographic systems).

Weighted hybrid: Results from different recommenders are assigned a weight and combined numerically to determine a final set of recommendations. Relative weights are determined by system tests to identify the levels that produce the best recommendations.

Mixed hybrid: Results from different recommenders are presented alongside of each other.

Cascade hybrid: Recommenders are assigned a rank or priority. If a tie occurs (with two products assigned the same recommendation value), results from the lower-ranked systems are used to break ties from the higher-ranked systems.

Compound hybrid: This approach combines results from two recommender systems from the same technique category (e.g., two collaborative filters), but uses different algorithms or calculation procedures.

37


Comments are closed.