Digital Marketing and Search Advertising 1 Digital marketing

Digital Marketing and Search Advertising 1. Digital marketing 2. Search Analysis 3. Panel data analysis © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 1

Online Technologies and the Transformation of Marketing © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 2

Growth of the Internet of everything (now) Contraptions (~2009) Community (~2003) Commerce (~1991) Communications (~1980) © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 3

The Evolution of Online Functionality at Leading Companies Real time business Functionality Web-centric enterprise Process management Transactions Dynamic content Static content 1992 1997 2000 2003 2007 2011 © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 4

Emerging Global Real-Time Enterprises Focused on Customers e-Operations Supplier network Extranet Real-time Enterprise Company’s digital infrastructure Intranet & Internet e-Marketing Customer/ Market network Internet Customer-driven Activities and flows Real-time is an organizational capability, powered by customer-centered information and process management architectures, for anticipating and responding to changing customer and market needs as fast, or faster than, those changes. © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 5

Growing Interest in Digital Marketing Each series is indexed to 100 for its maximum value during this period, and indicates the interest in the topic based on Google searches conducted on those terms. The absolute number of searches for “Digital marketing” is much smaller than the absolute number of searches for “Marketing. ” From www. google. com/trends (accessed October 7, 2016) © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 6

Online Advertising © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 7

Broadcast – Temporal Multiplexing Output Bandwidth 100% Temporal Multiplexing Advertising Content Input time © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 8

Online -- Spatial Multiplexing Output Spatial Multiplexing Bandwidth 100% Advertising Content Input time © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 9

Display Advertising Banner: 468 x 60 Skyscrapers © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 10

Search Advertising Text Ads Organic Listings Product Listing Ads Text Ads © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 11

Advertising: Traditional versus Online Traditional Online • Typically one-way • Can by two-way • Temporal multiplexing • Spatial multiplexing • Content and Ad generally separated • Content and Ad generally comingled • Tradeoffs between reach, richness, and frequency • Consumers reactive • Weak link between advertising and response measurement -- Recall, Attitudes, Intentions • Fewer tradeoffs between reach, richness, and frequency • Consumers proactive • Stronger link between ad and response measurement – Impressions, Clickthroughs • Adaptive ads • Static ads © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 12

Customer Decision Process for a Beauty Product l l l l Post UGC Post tweets Post Youtube videos Facebook likes l l l Internal trigger (e. g. , going on a date) Replenishment (order online? ) Impulse purchase (windows display) TV Ads, Catalogs, print ads Display Ads, e-Mail …. . Search for Ideas and Offerings Need Recognition l Inform Others buy Post Purchase Behaviors Advocate Beauty Care Decision Process Post-purchase Experience Incentives l Please salesperson l Discounts l Points l Samples l Order online, pick up at store l …. l l l Evaluate Alternatives l l Enjoy shopping and trying new stuff (at store) Search engine (paid ads) Watch Youtube videos Follow tweets. . Look for UGC (e. g. , reviews, Beauty Talk) … Purchase Decision l l Experiment and learn (help from sales staff) Product displays at store UGC (ratings/reviews) …. Source: Adapted from Rayport and Jaworski © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 13

© Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 14

Growing Roles of Search and Social Media in Marketing in the US Top 10 Web Brands (Home & Work, USA) (December 2007, Nielsen) Top 10 Web Digital Media Properties (USA)* (August 2016, com. Score) Unique Desktop Audience (‘ 000) Hours: Min/ Person 1. Google 117, 719 1: 11 1. Google sites 240, 704 2. Yahoo! 113, 203 3: 06 2. Yahoo! sites 213, 333 3. Microsoft 98, 338 0: 49 3. Facebook 206, 333 4. MSN / Windows. Live /Bing 96, 448 1: 53 4. Microsoft sites 188, 837 5. AOL Media Network 90, 850 0: 48 5. Amazon sites 186, 256 6. You. Tube 68, 582 2: 09 6. Comcast/NBC 164, 245 7. Fox Interactive 57, 591 0: 34 7. AOL 159, 599 8. Amazon 59, 624 2: 01 8. CBS Interactive 158, 303 9. e. Bay 59, 374 1: 15 9. Apple 153, 163 10. Apple 50, 686 0: 16 10. Linkedin 127, 229 Brand (All sites) Unique Multiplatform Audience (‘ 000) *: Includes both Desktop and Mobile Aug 2016 data from http: //www. comscore. com/Insights/Rankings © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 15

Basic Search Analytics 1. Search engines 2. Search advertising 3. Basic Search analytics © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 16

Growing Importance of Search Engines and Search Advertising © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 17

Huge Amount of Search Queries Everyday From: http: //www. comscore. com/Insights/Rankings (accessed October 9, 2016) About 60% of the searches result in ad displays. © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 18

Online Ad Revenue Growth Half-Year US Online Ad Revenues In US billions (current dollars) H 1 2006 -H 1 2015 $30 27. 5 23. 1 $25 20. 1 $20 17 14. 9 $15 10 $10 11. 5 10. 9 12. 1 7. 9 $5 $0 H 1 2006 H 1 2007 H 1 2008 H 1 2009 H 1 2010 H 1 2011 H 1 2012 H 1 2013 H 1 2014 H 1 2015 Marketing. Charts. com | Data Source: Interactive Advertising Bureau / Pricewaterhouse. COopers © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 19

Measurability of Online Advertising Confidence in Ability to Measure ROI, by Channel Based on a global survey of more than 500 client-side marketers/advertisers and agency /consultants (67% from the US) November 2015 “How would you describe your ability to measure ROI from marketing 100% 4% 90% 6% 5% 80% 22% 15% 17% 8% 70% 36% Don't Know 60% Poor 50% Fair 40% Good 30% 18% 17% 9% 29% 30% 36% 40% 67% 45% 20% 44% 29% 10% 22% 0% Paid Search SEO E-mail Display Media Social Media Marketing. Charts. com | Data Source: SEMPO © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 20

Potential Benefits of Search Advertising • Potential for reaching “motivated” customers. • Potential for communicating longer, more contentrich information. • Quicker and less costly to change content seen by customer. • Segment-of-one customization of content (e. g. , landing page customization). • Integrates advertising with selling (closed-loop marketing). • Can collect info about who views ads to gauge impact (Advertising and ad research co-mingled). © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 21

Introduction to Ad. Words © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 22

Some Terminology • Keyword: Words and phrases that an advertiser bids on. • Search term: What a customer types into the search box (or speaks to Siri). • Paid Search: This is what “surrounds” the organic search results at Google. • Display Advertising (Ad. Sense): An advertisement that is shown to you on a web page to where you are taken when you click on a link from organic listing – here Google inserts Ads on the pages if the publisher is part of the Google Ad network. © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 23

Overview of How Google Ad. Words Works WWW 1 1 Cached Web Pages 1 Distributed Web Crawler 1 Web map Keyword Auction Process 1 2 Keyword Index Database Keyword Ad Database 5 8 Dotted lines indicate browser gets display components from different sources through Google. com. 4 4 Page-rank Process These activities take place in the background. The other activities occur as a result of a user query. 6 6 Google. com 3 7 8 Web Sites/ Landing pages These sources contain relevant ads 8 User’s Browser Customer initiates process Numbers represent typical sequence of events starting with 1. © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 24

Search Advertising Text Ads Organic Listings Paid Product Listing Ads Paid Text Ads © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 25

Basic Structure of Ad. Words Advertising • Ad campaign (theme) • Ad group (an Ad associated with any keyword in an Ad group will take customers to the same landing page) • Keywords − Ad should be relevant to the query and product/service − Ad copy is important to entice user to click on the ad • Match type (Broad, Exact, and Phrase) © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 26

Basic Structure of Ad. Words Advertising Account www. electrostore. com Campaign 1 Campaign 2 Televisions Cameras Ad Group 1 Ad Group 2 3 D TVs HD TVs Compact Cameras SLR Cameras Keywords 2 -3 Ads © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 27

Paid Search -- Broad Match • Broad match (Keyword: kitchen knives) Ø Ad may then show for these search terms: − kitchen knife − Knives kitchen − Cutlery on sale − Kitchen utensils in New York − Chef’s knife © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 28

Paid Search -- Phrase Match • Phrase match (Keyword: “kitchen knives”) Ø May show whenever the keyword is present somewhere in the search term. Examples: − Best kitchen knives − Kitchen knives in New York Ø But not for these search terms: − Knives for kitchen (there is an extra word in between) − Knives kitchen (the words are reversed) − Kitchen knife (this is the singular) − Hunting knives (the words are different) © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 29

Paid Search -- Exact Match • Exact match (Keyword: kitchen knives) Ø May show for the following search terms (allows close variants): − kitchen knives − Kitchen Knives − kit chen knives © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 30

Paid Search Negative Keyword Match • Negative match (Keyword: cheap) Ø Your ad will never show for these search terms: − Cheap kitchen knives − Where can I buy cheap knives There also some more advanced options such as Single word phrase match, and Negative broad match. © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 31

Measuring Effectiveness of Online Advertising © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 32

“Funnel” Analytics for Assessing Ad. Word Effectiveness 1000 Impressions 10 $$ Spent Clicks 1% 10 % $$$ Got Clickthrough Rate (CTR) 2 1% 2% 100 Conversions Conversion rate ROI © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 33

Some Advertising Metrics • Impressions or Exposures − The number of times a page containing the ad is viewed (measured as the number of times an ad rotates through or pops-up on a web site). • Cost-per-thousand impressions (CPM) − The cost of gaining 1, 000 impressions/exposures. For example, if a banner ad has 30, 000 impressions at a web site at a CPM of $5, the total cost of the ad is $150. • Click-through-rate (CTR) − The number of visitors delivered to a site by a particular advertisement or the percentage of people who click on a web banner and visit a site, computed as a percentage of impressions. − An ad that generates only 1000 exposures, but gets 40 clicks is more efficient than one that generates 10, 000 exposures but only generates 100 clicks. © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 34

Some Advertising Metrics • Cost-per-click (CPC) − It is the price you pay for each click on your ad. − Determined via a formula that takes into account your maximum bid and bids of those you are competing against, adjusted for quality of your ad. • Cost-per-visitor (CPV) – mainly for display advertising − The cost of a given ad divided by the number of clicks obtained from it. − It combines CPM and CTR to evaluate effectiveness of ad placements − On a website with a CPM of $20 and a CTR of 2%, it costs you $1 per visitor; a website with a $40 CPM and a CTR of 10% costs you $0. 40 per visitor. The higher CPM ad actually has greater value to you. • Cost-per-conversion/Action (CPA) − Cost of a given ad divided by the number of orders (or other measures of conversion such as filling out a call request) received through that ad. CPS is cost per sale. © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 35

Hierarchy of Keywords • Brand keywords – contains brand name in the keyword − Usually will get good conversions (low CPA) • Product keywords (what product is, or does, or problems it solves) − Higher cost per conversion than branded keywords • Competitor keywords – contains competitive terms or competitor brands − Could be helpful if you are not the major player. Higher CPA and also higher CPC (higher CPC because it will get lower Google quality score). © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 36

Hierarchy of Keywords • Substitute product keywords (e. g. airlines and trains) − Usually will have lower Google quality score, higher CPC, and higher CPA. − Lower position in organic listings. • Complementary product keywords − Could get decent CPA in case of strong complementary products (e. g. airlines and rental cars). • Audience/Segment related keywords − May be effective for getting impressions, but not clicks or conversions. © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 37

Driving Demand Through Search Increase conversion rate Visitors Conversion Rate Conversions 1000 2% 3% 20 30 Base © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 38

Driving Demand Through Search Increase conversion rate Visitors Conversion Rate Conversions Increase potential customers (traffic) 1000 2000 2% 3% 1. 5% 20 30 30 Base © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 39

Driving Demand Through Search Increase conversion rate Increase traffic Or both Visitors Conversion Rate Conversions 1000 2000 2% 3% 1. 5% 1. 8% 20 30 30 38 Base © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 40

Dayparting Historical Conversion Rates by Time of Day 2. 50% 2. 00% 1. 50% Average Conversion Rate 1. 00% 0. 50% 0. 00% 0 6 12 18 24 Dayparting is a term used for describing placement of different ads at different rates during different times of a day. Increase bids during times when conversion rate is above average and decrease bids during periods when conversion rate is below average. This can be done using “bid multipliers. ” © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 41

Some Strategic Considerations • Before you spend more money on search advertising, make sure you reallocate current spend optimally. You get infinite ROI by reallocating current spend in a way that increases profits. • In search advertising, setting an overall budget is less important than knowing its profitability! • You need customized analytics to assess profitability. • Use benchmarks (your past data, average industry data) to guide spending increases. © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 42

Some Actions to Take to Improve Paid Search Performance Operational guidelines • Eliminate keywords that are low performing (low CTR, high CPC, or low conversions). • Convert high-performing broad match keywords to exact match keywords. • Modify bidding according to daypart performance. • Limit number of keywords in an Adgroup (it may even be ideal to have one high-volume keyword per Adgroup!) Strategic guidelines • Find keyword themes that are best aligned with your product and your current and potential customers (e. g. branded keywords, keywords that contain “free”). • Rationalize the set of keywords used – minimize keyword creep. © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 43

The Long Tail Phenomenon Conversions Head: A few keywords generate a large amount of traffic and total conversions, but cost more. Tail: These keywords generate low traffic and conversions but bring in different kinds of customers, and cost less Keywords © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 44

Another Example of the Long Tail Source: www. openslatedata. com © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 45

Panel Data Analysis 1. Nature of panel data 2. Potential insights from panel data 3. Example panel data 4. Implementing simple panel data analysis in Excel © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 46

New Types of Marketing Data • Huge amounts of various types of data are generated by online activities of consumers and firms. • Evolution of marketing data: • Static data (from 1970’s) – one time data collection • Panel data (from 1980’s) – typically same entity is observed at several points in time, but data not immediately available for analysis • “Real-time” data available for immediate analysis (from 1990’s) • Text, audio, video and other data (from the turn of the millennium) © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 47

Brief Introduction to Panel Data Analysis © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 48

The General Nature of Panel Data • We observe data for the same entity (e. g. person, keyword, store, industry, etc. ) in several different contexts (usually the different contexts are different points in time). • And, we observe such data for a number of entities (i. e. for a set of panel IDs). © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 49

Reasons for Growing Use of Panel Data Analysis • We can account for explicit characteristics of an entity (e. g. , whether keyword is branded or unbranded) in measuring effects of marketing activities. This reduces “aggregation bias” resulting form ascribing a common effect to a marketing activity across all entities (e. g. across all keywords). • By incorporating additional sources of variation (e. g. characteristics of entities; unique conditions prevalent at each time period) we reduce model error and improve the quality and accuracy of model predictions. • We can understand how effects are changing over time (i. e. we can understand dynamics and transition behaviors such as whether branded keywords are becoming more effective over time). © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 50

Example Panel Data • This is a balanced panel – same number of observations (8) for each unit (Kwd). The units can be anything – consumers, firms, countries, stores, Kwds, Apps, etc. for which we have repeated observations. • An additional panel structure here is that we have 3 observations for each Time of Day. • Question: Do we have more conversions when we spend more money? © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 51

Conversions 60 Regression line of the form: �� = �� + ���� 50 Conversion = 4. 812 + 0. 00243*Cost 40 30 20 10 0 $0 $4, 000 $8, 000 $12, 000 $16, 000 © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 52

Implementing Simple Panel Data Analysis in Excel © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 53

Example Analysis with Dummy Variable Coding Estimated Regression Coefficients t-statistics 6. 14543 2. 133705 Cost (X) 0. 00279 4. 960883 KD 1 1. 46423 0. 317322 KD 2 -9. 30951 -1. 96203 Interpretation of “Fixed Effects” of keywords 10 8 6. 145 7. 61 6 4 2 0 -2 -4 f 1 f 2 f 3 -3. 164 © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 54

Example Results from Simple Panel Regression Plot of Cost versus Conversions Kwd 1 Kwd 2 Kwd 3 60 50 40 30 20 10 0 $0 $2, 000 $4, 000 $6, 000 $8, 000 $10, 000 $12, 000 $14, 000 $16, 000 Conversion = 6. 1454 + 0. 00279*Cost +1. 454 (for Kwd 2) - 9. 3095 (Kwd 3) © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 55

Hour-of-Day Fixed Effects Conversions By Hour 40 35 30 Base is Conversion at Midnight = 19. 667 25 20 15 10 5 0 H 1 H 2 H 3 H 4 H 5 H 6 © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. H 7 56

Panel Data with Simple “Fixed Effects” 5 Conversions 4 A panel of Keywords 3 2 1 0 Observation 1 Observation 2 Observation 3 © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 57

More Complex Panel Data Example 4 Conversions 3 2 1 0 Observation 1 Observation 2 Observation 3 © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 58

Implementing Fixed Effects Panel Data Analysis in Excel • Fixed Effects panel data analysis can be done in Excel if the number of fixed effects is small (less than 16). Use dummy variable coding (as I did in the example shown). • To model N fixed effects you need N-1 dummy variables – example, for 3 fixed effects, you need 2 dummy variables. © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 59

General Approaches to Modeling Panel Data • If there is a large number of panel members whose effects we need to account for, we have two options: 1. Implement “fixed effects” model in standard statistical software such as SPSS, SAS, STATA, or R. This works best if there is a sufficient number of observations for each fixed effect. © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 60

General Approaches to Modeling Panel Data • © Copyright Decision. Pro, Inc. 2018. Commercial distribution and any publication is not authorized. Copying and distribution for non-commercial educational purposes only is authorized if this notice appears on every copy. 61