Visual Thinking Assignment USA TV Network Renewal Decision
Transcription
Visual Thinking Assignment USA TV Network Renewal Decision
Visual Thinking Assignment USA TV Network Renewal Decision Visualisation Anna Thalassinos Walter Khumalo thlann002@myuct.ac.za khmwal002@myuct.ac.za Contents 1. Introduction.................................................................................................................................... 3 2. Overview and analysis of related work ..................................................................................... 6 3. Visual Queries............................................................................................................................. 10 4. Design .......................................................................................................................................... 10 5. Strengths and Weaknesses of Design .................................................................................... 24 6. Conclusions ................................................................................................................................. 24 7. Future work ................................................................................................................................. 24 8. Work Allocation ........................................................................................................................... 25 2 1. Introduction 1.1. Problem Description In the United States of America, television is a massive industry, with 99% of households owning a TV set [1]. There are a number of broadcast networks in the USA which purchase television shows from production companies and air them for public viewing. Towards the end of each television season, networks need to decide whether to renew or cancel a particular series depending on the amount of money a show generates for the network. This value is primarily due to the advertising revenue that the series brings in. Advertisers are willing to spend large sums of money to advertise to specific demographics who watch a series. There is a wealth of data that needs to be analysed in order for a renewal decision to be made regarding a series. This report proposes a visualisation of all of this data in order to allow rapid processing of all of the relevant information to facilitate accurate TV series renewal decisions. For the purpose of this report, the term “series” refers to a television program and the term “season” refers to an order of episodes of a series. The terms “series” and “show” are used interchangeably in this document. Series usually air from the Fall – Spring (henceforth referred to as the Winter television season), or Spring – Fall (henceforth referred to as the Summer television season). A season may have any number of episodes, generally between 13 and 24. The network decides on the size of the episode order based on the series’ performance. A particular series generally airs during the same season each year if renewed, although the network may move it to a different season if it is not performing well in its current season. The network may also decide to change the day a series airs or the time slot in which it airs. The visualisation needs to support visual queries which correspond to the questions mentioned above. 3 1.2. Data Currently, the most important consideration for networks is the ratings a show receives, as measured by USA information and measurement company, Nielsen. Nielsen, however only monitors a small subset of the population’s viewing, via set top boxes [2]. The viewership numbers reported by these Nielsen top boxes is therefore not representative of the true popularity of the show and in past years have resulted in shows with large cult followings being cancelled because of bad numbers. It has since been acknowledged that the true value of a show should not be assessed merely in terms of how many “eyeballs” the show has, but how the audience engages with the show. That is, is the television just on, or is someone actually watching the show and engaging with it. Nielsen and TV networks have begun to realise the importance of social networking as a better indication of a show’s value [3]. This social activity surrounding a show is termed “social buzz”. This visualisation therefore incorporates the series’ social buzz to represent the show’s total popularity. “Syndication” is an important term in the television industry. When a show has aired between eighty eight and one hundred episodes, the network may sell the rights of the series so that it may air on any channel (five or more times per week)[4]. This is referred to as syndication. Syndication is desirable because it allows a show to continue making money even once production on the show has ended (if the show is cancelled). Networks like to get their shows to syndication, therefore proximity to syndication is something that they take into consideration when making renewal decisions. For example, if a show is performing poorly but has 75 episodes, then giving the show a 13 episode order will get the show into syndication. This visualisation will include the following data: • Series name • Network name • Day aired • Timeslot aired • Genre • Season aired (Summer/Winter) • Nielsen number of viewers (total for the current season – all episodes added up, regardless of age group) • Social Buzz (Twitter and Facebook posts) • Number of episodes (proximity to syndication) • Renewal Status (cancelled/renewed/pending) 4 For the purposes of this design, data was manually collected. The data is dispersed across various sources, which makes manual gathering laborious. Therefore we have restricted the data to five networks in the USA (ABC, NBC, The CW, Fox and CBS) and our four favourite genres (Crime, Comedy, Sci-fi and Reality). Nielsen makes the ratings information available to networks, advertisers and media who in turn make some of this data available to the public via online articles. The validity of the data in these articles cannot be determined. Ideally, the data should be bought directly from the Nielsen company. Social buzz numbers are provided by Mashable [5]. The output is essentially the visualisation into which all of the data mentioned above is encoded graphically. The only processing of data is the calculation of the size of the shape, based on viewership numbers and social buzz. 1.3. Target User Profile The primary users of the visualisation are the network executives at television networks in the United States of America. We therefore assume familiarity of network terminology (syndication, season etcetera) as well as knowledge of who their network’s competitors are (that is, they know which network they should select from the drop-down boxes for comparison). We also assume that the executive is familiar with the shows on their own network as we use the logos of the shows in order to help the user identify what they are looking for. Although the visualisation proposed by this project is targeted at network executives, it might also be valuable to fans of television shows so that they can see ahead of time how their favourite show is doing in order to assess the likelihood of cancellation. People become consumed with television series and their characters and when the season comes to an end they are desperate to know whether their favourite shows have been renewed for another season, thus the visualisation can help set their minds at ease. 5 2. Overview and analysis of related work There is no indication of whether there is a visualisation that networks use to analyse and compare the performance of television series in order to make the renewal decisions. There are a number of informal visualisations relating to social buzz and ratings of TV shows online. These have been created by fans and infographic companies. Some examples are outlined below: (Source: http://pinterest.com/pin/72550243967440446/) The above graphic shows social buzz per network for a year. Although the main shows contributing to a networks total are mentioned, a network would be more interested in seeing exactly how each show on the network contributes to this social buzz total. The series names written alongside the chart makes the graphic quite cluttered. 6 (Source: http://www.throng.co.nz/files/u2/american-2009-season-ratings.gif) The graphic above uses red circles to represent the number of viewers of a series’ first episode of the season. The burgundy circle represents the average number of viewers for all episodes that have aired. It does not distinguish between networks and also does not include social buzz and other necessary data such as number of episodes and genre. The primary visual query above is “is the average number of viewers for a series for the season more than or less than the number of viewers for the premiere of the season?”. This is not really an indication of what should be renewed and what should be cancelled. (Source: http://adage.com/article/the-media-guy/inside-social-media-triumph-nbc-svoice/228534/) The graphic above shows the tweet volume for one episode of American Idol and one episode of The Voice. This is not a particularly interesting visualisation as it is simply a horizontal bar chart displaying the tweets relating to a show for one episode. It does not include any information about ratings or other important data need to make decisions. 7 (Source: http://mashable.com/2013/01/14/social-media-tv-chart-1-14/) The social media statistics as represented by Mashable above is more useful as it ranks shows according to their popularity on social networks, indicates which show the network belongs to and also shows whether the buzz was predominantly positive or not. It is not very visual though. Additionally it only shows the top ten shows according to social buzz whereas a network executive would like to see all shows on their network. It also does not show ratings. For fans to monitor which of their favourite shows are likely to be renewed, many TV blogging sites maintain score cards. Once such scorecard is represented below: http://tvline.com/2013/01/04/tv-shows-renewed-cancelled-2013-abc-nbc-cbs-fox-cw/ 8 It would be useful to have the information represented graphically so that fans are able to look at it and will immediately be able to see what has been renewed/cancelled without doing much reading. The infographic presented alongside is a far more detailed visualisation which provides information about social buzz relating to television shows. The last visualisation at the bottom of the image is particularly visually appealing. It even uses colours and symbols to distinguish between genres (see the key). It also varies the thickness of the lines according to how many social comments there were per episode. Once again, not all series are represented and networks of the series are not mentioned either. (Source: http://pinterest.com/pin/72550243968196151/) 9 This analysis of related work indicates that although there are visualisations relating to TV shows social buzz and ratings, these visualisations are dispersed. There is no visualisation that encodes all of the relevant information for decision making. 3. Visual Queries The main objective of the visualisation is to allow the user to look at it and be able to decide relatively quickly which shows are worth renewing for another season and which shows have run their course. This is done by comparing a show with other shows on the same network or with its competitor shows on a different network. There are a number of small visual queries which, combined, allow for the answering of the most important query which is: • Which shows should be cancelled and which shows should be renewed? This decision cannot be made without considering a number of variables, therefore there need to be visual queries to support comparison of these variables. An effective visualisation would therefore allow for comparison between shows on different networks, different seasons and days. Some smaller queries that aid overall renewal decision are: • If I move Show A to a different night/timeslot/season, what will its competition be? • How popular is Show A compared to other shows in a genre on other networks? • How many episodes should the show be renewed for, if renewed? Just enough to get it into syndication maybe? 4. Design There were a number of iterations of design, as well as two crit sessions with feedback. Although the networks and series used in the designs are real, their viewership numbers and details are fictitious. The rough sketches of our initial designs as well as detailed descriptions of the two final designs, are discussed in this section. 10 4.1. Initial Designs After a meeting, during which we discussed our understanding of the problem and shared links to resources, we each went off and came up with our own design sketches. Between the two of us we came up with a number of designs. We then came together and shared our ideas. No design was perfect. We each took time to draw our designs up on the whiteboard in the lab while explaining the design as we went. The watching member then offered suggestions and pointed out omissions in the design that needed to be present. We also evaluated whether the design answered all of the important visual queries (as listed above). If a query was unanswerable, we discussed how the design could adapted and made to address the query. Once all designs had been discussed, we observed that the shortcomings of one design could be resolved by borrowing from another design. As such, we began to merge designs, discussing, and occasionally butting heads over clashing opinions. In hindsight, the clashes lead to a more fleshed out design. Our initial design sketches as well as initial Photoshop designs are depicted below: 11 12 From all of the initial design sketches above we ended up with Design 1, which follows and which was presented to the class during the first crit session. 13 4.2. Design 1 The slides presented at the first presentation are available for download at the following URL: http://people.cs.uct.ac.za/~athalassinos/visWebsite/US%20TV%20Series%20Renewal.pptx http://people.cs.uct.ac.za/~athalassinos/visWebsite/US%20TV%20Series%20Renewal.pptx High-level view Detailed View 14 Explanation All USA television networks branch out from the central node of the tree. From each network’s node, there are two branches; one for winter and one for summer. From each of the winter and summer branches, the genres of that network branch out. From each of the genres, the days of the week branch out and from there, the shows that air on that day are represented. In the high-level high level tree view, the shows are represented as circles with not much detail, however, when the user clicks on a particular show, a more detailed view is displayed with more information. This is explained below. Design Rationalisation • The logos of the networks are used because they are symbols that an executive would be familiar with. It would be easy for the executive to look at the tree and find their network. • The branches representing winter and summer are colour coded. Winter is blue, which is a cool colour, while, summer is a warm orange colour. • Circles represent series, series, with the size of the circle representing the popularity of the show (the ratings of the current season plus the social buzz on Twitter and Facebook). • The colour of the circle represents the number of episodes of that series that has h aired. The colour corresponds to a range of episodes, and therefore also indicates the proximity of a show to syndication. Circle colour range o Episodes < 64 (red) o 64 <= Episodes < 88 (green) ( o 88 <= Episodes (already in syndication) (blue) ( The colour heat intensity is used to convey meaning. (http://donnayoung.org/art/color http://donnayoung.org/art/color-theory.htm) 15 Red is a very hot colour, and may indicate danger. If a series has less than 64 episodes, renewing it for another season of 24 episodes will not get it in to syndication, therefore we thought red an appropriate colour. Green is cooler than red, thus representing a series with between 64 and 87 episodes – another season will get it into syndication. A show that is coloured in blue is already in syndication. • The colour of the circle’s outline represents whether the series has already been renewed, whether it has been cancelled or whether the decision is still pending. Outline colour o Cancelled – red outline o Renewed – green outline o Decision pending – black outline In terms of renewal, we represent cancelled shows in red, once again drawing on red’s reputation as representing danger, while we use green, the opposite of red on the colour wheel to indicate the opposite of cancellation - renewal. Additionally, green is considered to be a colour that represents something good as well as growth and renewal, which is appropriate for our context. We decided to use black for the outline of series pending decision because these are the shows that the network will be looking at because they need to make a decision regarding these shows. Making the outline black (a colour that has not been used to colour the inside of the circle), makes the series stand out. • The number within circle indicates the timeslot in which the series airs. The number is transparent and overlaid on the circle so as not to detract from the other aspects, especially in the detailed view. • The series circle is divided into a pie chart, with the pie slices representing the following (in this order in a clockwise fashion), Twitter, Facebook and viewership. These slices are differentiated by texture. The textures are very distinct to prevent the slices from blending in with each other. Crosses are easy to differentiate from horizontal lines, as are dots. Additionally, the outline colour of the circle has been used to define the slices of the pie (that is, if the show is pending decision, the outline of the slices is black, just like the outline of the circle). 16 Example The image above represents two shows, show , A and B. B is much bigger than A, indicating that is a far more popular show.. Show A, A which is far from syndication, airs at 8pm and has been cancelled, while show B,, which is close to syndication, airs at 10pm and a decision regarding its renewal has not yet been made. Viewership accounts for a larger portion show B’s popularity than social buzz, while show A has almost equal amounts of social buzz and viewership. Feedback A number of valid comments were wer made in the first crit session which had been overlooked in the initial design sessions. Often it is easy to overlook things in your own design but an impartial observer is able to spot issues and make suggestions by looking at it. For this reason, we found the session very useful. The main problem was that too much was being represented and the circles representing the shows were too small for comparison. It was suggested we look in to using blur, other shapes and symbols somehow. It was also pointed out that the outline of the shapes was not easily visible because it was sometimes the same colour as the shading of the shape. We were not completely happy with our first design and we felt it necessary to revisit the brainstorming process. We realised that although there is a wealth of data that we need to represent, we should make use of an interface in order to allow the user to select some desired parameters, which would limit the amount of information we needed ed to encode in the visual at any one time. We made a number of changes to the initial design, that we felt addressed ed the concerns raised and also incorporated some of the suggestions made in order to produce a more 17 effective visualisation. Additionally, drawing on inspiration from classmates’ designs and the feedback that they received, helped us avoid things that would not work. Our second iteration was therefore quite different from the design first presented. Another important point that emerged in the crit session is that the design seemed to be constrained by the fact that we had selected a tool to use. Had we not had a tool in mind, we could have been more creative. The next time we embark on a visualisation design, we will know not to select a tool first. 4.3. Design 2 The slides presented at the second presentation are available for download at the following URL: http://people.cs.uct.ac.za/~athalassinos/visWebsite/VIS%20Pres%202.pptx This is the design that was presented at the second session, which is an effective redesign of the overall visualisation, keeping some of the details of the first design presented. The inspiration of this design was a simple doughnut chart, that evolved into an interactive doughnut with multiple concentric circles, drawing inspiration from the Sunburst example presented at this URL http://philogb.github.io/jit/static/v20/Jit/Examples/Sunburst/example2. This design is more refined because we adhered to the “detail on demand” principle which means that we visually encoded only the most relevant data, making the design less “busy“ and putting the spotlight on the information that a user would typically want to see. To this end, we now allow filters so that the user is in control of what he/she sees. Whereas the hyperbolic tree design attempted to display all networks and its shows for each season simultaneously, design two allows a maximum selection of two networks to compare and a minimum of one network. 18 View Explanation The user selects their desired criteria from the drop-down boxes on the left hand side. The network filter needs a minimum of one selection and allows a maximum of two networks to be selected for comparison (a drop-down is provided for each of these two options). All other filtering fields are not mandatory. A large doughnut chart represents each of the selected networks with the relevant data encoded within it as outlined in the rationalisation below. Only once the user clicks the “view” button does the visualisation appear. This is done to avoid having to render different parts of the visualisation each time a filter is changed. That is, the user should select all filters and then click “View”. 19 Design Rationalisation • Different shapes represent different genres. Circles represent Sci-Fi series, squares represent Crime series, triangles represent Comedy series and stars represent Reality series. The selection of stars to represent Reality series plays on the fact that the people on Reality series want to be stars. • The size of the shape represents the popularity of the show. This includes the number of viewers plus the social buzz (Facebook and Twitter posts about the show). • The colour of the shapes represents the number of episodes of that series that has aired. The colour corresponds to a range of episodes, and therefore also indicates the proximity of a show to syndication. Circle colour range o Episodes < 64 (red) o 64 <= Episodes < 88 (green) o 88 <= Episodes (already in syndication) (blue) This is the same as in design 1. • The logos of the networks are used for familiarity and are placed at the focal point of the doughnut chart which represents a specific network. • Shapes are outlined in black so that stand out from the colour background of the circle’s rings. The black outline makes the shape “pop”. • Each of the concentric rings of the circle represents a television season. The blue ring represents winter because blue is a cooler colour than orange which represents the summer television season. • The inner and outer rings can swap in and out so that the season selected by the user in the filter is the outer ring. This makes it easier for the user to see the shows that they are interested in seeing. For example, if the user selects winter, the blue ring will be the outer ring of both network circles select (as illustrated in the image above). If the user would like to compare the winter shows of one network to the summer shows of another network in order to determine whether a show should be moved to a different season, the user can click on the inner ring of the second circle in order to swap it with the outer one. Having the two seasons being compared on the outer ring makes comparison between shows on different networks easier because the distance between them is reduced and there are no irrelevant shapes in the path that the eye follows from one network to another. 20 • The segments (renewed/pending/cancelled) of each network aligns with that of the other network. If the user selects “pending” from the filter, the pending segment of the first network will be aligned with the pending of the second network. By clicking on the second network and dragging in a clockwise direction, the second network can be rotated to align a different renewal status with that of the first circle. This will allow a user to compare the pending shows on their network with those shows already renewed by another network so that they can see what the competition will be like next year for that season. • Series not corresponding to the filters selected by the user are blurred out. The blur is very effective in placing emphasis on the things that are relevant and making them stand out more while not completely eliminating all other data. • If filters are selected and the user clicks “View”, the shapes of the shows corresponding to those filters will flash once to grab the user’s attention and focus it on those shows. This, along with the blur, uses two channels to emphasis the important information and drowns out the rest. • In the key, coloured paintbrush symbols are used to represent the colours used for episodes ranges. • Text (renewed/pending/cancelled) is always readable – they are never upside down even when the circle is being rotated. Query Walkthrough The following is a query that I would need answered if I were network executive at ABC. Query: Should I cancel Castle? I select the following from the drop-downs: Networks: ABC and NBC Season: Winter Time slot: 10pm Day: Monday Genre: Crime Status: Pending I click view and I see two network circles as in the image on page 19. The blue ring (winter) is on the outside of each circle and the pending of the circles is lined up. I see that Castle is very popular because it is a very big square. Although blurred, I can see that it is bigger than shows in other genres in winter on ABC. It is clearly one of the best 21 shows for ABC. It is already in syndication. Because it is so popular, it would be a wise decision to renew it, but to make sure, I want to see what it will be up against next season. To do this, I click on NBC’s circle and rotate it clockwise until the “renewed” of NBC lines up with ABC’s “pending”. Now I can see that there are no crime shows that have been renewed yet, so it will not have competition in the genre. I should definitely renew Castle. Feedback Once again, a number of useful constructive criticisms and suggestions emerged. • One of the best suggestions we received is something that really cuts to the core of the problem this visualisation endeavours to address, that is, comparison of shows between networks. It was suggested that we add an “all other networks” option to the networks drop-down list that would allow an executive to compare their network to an aggregated view of all other network’s shows. • Also mentioned was the fact that the shapes representing the genres are not related to the genres and as such, it is an unnecessary cognitive load to remember which shape represents which genre. It was suggested to consider using symbols as an outline with a show logo transparency over that (or to hover over shape to see genre). Ideally for this presentation we would have liked to have an interactive version of the design in order to fully demonstrate the user interaction with the design and how the interactivity component facilitates visual queries. For example, allowing the user to align renewal statuses and switching the seasons in and out so that the relevant season is on the outer ring. These design aspects are not effectively conveyed with static images. 4.4. Final Design The final design incorporates the feedback from the second crit and it is only slightly different to design 2. 22 View The changes that were made are as follows: • Symbols are used to represent genres. Although we kept the shapes the same as in design 2, an alien symbol is overlaid on the sci-fi shapes, a gun on crime shows and theatre masks on comedy shows. Reality does not have an overlay because the star shape already conveys what the genre is. • We have removed the show logo transparency from each shape as it has been replaced with the genre symbol. Having both the logo transparency and the symbol overlaid would have been messy. The show name is written next to the corresponding shape and the show logo can be seen if the user hovers over the shape. • By clicking on a specific show, the details of the show are shown on the left hand side of the visualisation. 23 5. Strengths and Weaknesses of Design We consider the following to be the strengths and weaknesses of our final design. • If there are many shows in a season, the ring will need to be very big. Therefore, the width of the ring, as well as the placement of the shapes (which represent series), will be determined ad-hoc. The width of the inner and outer rings therefore does not need to be equal – it is dependent on the number of series and the size of the shapes which represent them. • Shows in different genres are difficult to compare because one cannot easily compare different shapes. • The popularity of a show can be misrepresented because of the inclusion of social buzz in the size calculation of the shape. This is because buzz is not necessarily good, but a lot of buzz (albeit bad buzz) will make the shape big creating the illusion that it is very popular. o Solution: formula for calculating shape size: #viewers accounts for 80% ; social buzz 20% o details view shows ☺ and to represent good and bad buzz percentages 6. Conclusions To our mind, our final visualisation is refined and effective. It allows for the answering of the most important visual queries in order to allow rapid decision making for executives at television networks. We believe that this fleshed out idea can be handed to a development team as a brief for development. We think that our biggest success was abandoning our first design despite the fact that one can be attached one’s first design. In the end our second design was far better. The presentation sessions were incredibly helpful for the design process because many ideas were picked up and feedback specific to our group helped refine the visualisation. 7. Future work Although we would have liked to have a fully interactive implementation of the design, badly documented JavaScript visualisation libraries prevented us from doing so. We tried to adapt the Sunburst visualisation 24 (http://philogb.github.io/jit/static/v20/Jit/Examples/Sunburst/example2), however it required us to define custom nodes that allowed shapes, rather than text to be rendered but the documentation was not comprehensive enough to allow us to do this. We did however implement the parts of the visualisation that we could with conventional HTML and JavaScript. This implementation is available on the project website (people.cs.uct.ac.za/~athalassinos/visWebsite) under the “Prototype” menu item. An implementation with real data would be useful in order to determine whether the network circle could be scaled appropriately to fit all series’ shapes and whether it is visually appealing, or too cluttered with all of the shows in it. 8. Work Allocation Anna Thalassinos • Report (content and website) • Modelling of Design 1 using ThinkBuzan iMindMap (for tree) and Photoshop (for piechart) • Presentation slides Walter Khumalo • Tool discovery • Modelling of Design 1 using ThinkBuzan iMindMap and Photoshop (for pie-chart) • Design 1 Implementation using InfoVis (hyperbolic tree design) • Design 2 Implementation using HTML and JavaScript Reference List [1] http://transition.fcc.gov/Bureaus/Mass_Media/Factsheets/factvchip.html Accessed 20 May 2012 [2] http://www.howstuffworks.com/question433.htm [3] http://www.nielsen.com/us/en/press-room/2012/nielsen-and-twitter-establish-social-tvrating.html [4] http://variety.com/2003/tv/news/cablers-raise-syndie-stakes-1117892381/ [5] http://mashable.com/2013/01/14/social-media-tv-chart-1-14/) 25