Guide to Interactive TV (cont.)
How Does iTV Collect and Use Data

How Does iTV Collect and Use Data?

If you are curious about, or writing about, interactive television, you will want to become familiar with the following buzzwords:

Three Types of Interactive TV Service

  1. Cable
  2. Satellite
  3. Personal Video Recorder

(which can be joined to either of the above or stand-alone.)


Three Methods of Sending Content

  1. Broadband, sending a “carousel” of different content for the viewer.  As with ordinary TV, all viewers receive all options to their TV sets at once.  Unlike ordinary TV, they now have a computer program, perhaps two, making decisions about what pictures to show.
  2. Request – response, which works more like the internet.  Sometimes it is little more than the internet, providing content to a TV set.  When a user clicks a button, he or she is requesting a file of content.  The set top box waits for the file and then knows how best to use what it gets.
  3. PVR – The personal video recorder has been celebrated as a way to avoid commercials.  But not long after it was brought to market, new ways to send advertising were found literally hiding in the box.  TiVo, for instance, runs a service called TiVo Takes, which automatically tunes into commercial content running late at night.  The box then makes decisions about when and how to show that content.

Three Methods of Collecting Viewing Data

Real Time – Raw or aggregated data sent immediately to storage database.

Store and Forward – Data saved on set top box and sent to storage database at night.

Stay on the Box – Storage database is kept on the set top box.  Data is kept either in raw form, aggregated form or as statistics – the probability that a viewer or household does in fact match certain predefined profiles.

Seven Types of Data to Watch

  1. “Click stream” – the raw viewing data
  2. Viewing data aggregated over people
  3. Viewing data aggregated over time
  4. “Anonymous” identifier – cookie or “silhouette”
  5. Set top box IP address
  6. Personally Identifiable Information (PII)
  7. Birthday and ZIP code


Any viewer of interactive television will have filled out a subscription form of some kind.  The form will contain bits of information such as name, address and credit card number, that are called “personally identifiable information” (PII).  Most laws on privacy have dealt with this type of data.

But interactive television is a perfect example of the way that new technology is challenging the whole concept of PII.  For instance, given just a birthday and a ZIP code, direct marketers claim they can match members of the public to existing databases of consumer information with 97% accuracy.

And, more fundamentally, does it matter?  If you watch four hours of TV every day, if this one screen provides you with most of what you see and hear about the outside world every day, what does it matter whether or not someone knows your name?  Your IP address is more than enough.

Another practice to look out for is the “merge and purge”.  Two companies want to share data, but neither of them can sell their list to the other.  Perhaps they have told their subscribers that they would never give viewer information to “third parties”.  However, by merging the two lists into one, sending off whatever mail shot or programming is desired, and then destroying the combined data, marketing can be done as if one company owned the whole list.

Okay, so your mailbox is full of junk.  But your PII is safe!

10 Types of Click stream

On-Off Times: When?  How often?

Shows Watched: Which were seen in part, which seen from start to finish?  What tiny bits of shows were seen as the viewer flicked through channels, and how long did the viewer remain?   It is possible to ad coding to a show as it runs, and answer such questions as what was happening on the screen just before the viewer turned the channel?

Electronic Program Guide (EPG): What pages and lists of programs were viewed?  How is the guide used? 

Commercials Watched: Which viewed to completion, which viewed in part, time spent

Commercials Clicked: Whenever the viewer hit a "tell me more" icon, time spent reading or watching the further information.  Other clicks to other information screens.

Commercial Purchases: Which ads persuaded a viewer to click on an order form, which ads lead all the way to an online sale.

Services Used: Which services, such as email, diary, chat, weather, stock market quotes, shopping, viewing planner, child viewing filter etc. How often used, every keystroke typed in to any of the services.

Online Shopping:  Where viewer browsed, where viewer almost bought or requested more information, when and what was bought.

Forms and Surveys: Viewer requests for information, pop-quizzes, viewer opinion polls, electronic program guide preferences, program suggestion forms, game or activity entry forms.

Games and activities: Every click to move characters in games of hand-eye co-ordination, scores, preferences and all choices or actions in games of fantasy and strategy.


Analyzing the Data

It's all about choice.  All the choices a viewer makes, from solemn decisions to fleeting impulses, reveal things about him that can be used, if you know how to look at them the right way.  The first kind of analysis then is viewing analysis, what Neal Muranyi of the Database Group calls "telegraphics". 

It will be easy for broadcasters to create the following reports about viewers of interactive television:

·        Viewing hours over week

·        Channel choice over week

·        Loyalty to shows

·        Who watches a certain show or shows, in order by wealth of neighborhood

·        Advertisements missed or seen

·        Viewer restlessness by type of program

·        Viewer restlessness by type of advertisement

·        Response to big events


These reports are simple but powerful.  They describe behaviors in each household that broadcasters and advertisers have long dreamt of knowing. How many people are watching a car program, for how long, and how many of them live in wealthy neighborhoods?  Who are they?  What are their addresses?  It has been shown that the more involved viewers are in what they watch, the more likely they are to accept advertising messages.  Interactive TV has been shown to increase involvement, but can also be used to measure it in clicks and keystrokes. 

Telegraphics is vital in any strategy to keep you in front of the screen.   How do you and your family use television?  What content on TV, and situations outside, combine to make you watch more?  When do you turn it off altogether?  How can you be stopped from doing this?

But beyond selling airtime, television providers are now in the business of selling data or the use of data.  And their data customers, the Watchers, will be asking other questions which spring from the types of analysis described below.  To them, viewing and online buying are behaviors to study, and telegraphics will put that viewing data into a form they can use.  What they get will be more varied and meaningful than just changing channels.  Telegraphic analysis will also tell advertisers which viewers clicked on what ads or took a peek where, how long they peeked, who played a game, how they played etc. 


5 Types of Profiling


1. Demographic

How much you make, married,

2. Geodemographic

As above, but linked to sophisticated profiles of your neighborhood

3. Lifestyle

The things you buy and do linked to things you might buy and do in the future

4. Market Segment

The things you buy linked to a theory about where you fit in the culture

5. Psychographic

The things you buy linked to a theory about what they meant to you


Number five is worth a close look.  People have long dreamed of, or dreaded, the day when psychology could be used to push up sales.  However, advertisers have always been hampered by one thing: a lack of knowledge about individual consumers.  Vance Packard, in his 1957 classic “The Hidden Persuaders” was able to find the word “sex” in bubbles of soft drink on billboards.  But to move beyond such crude tricks would require that advertisers to live with, and observe, individual consumers for hours every day.  That has never been possible until now.


Where the Data Goes

Follow the data, it leads everywhere.  Some companies plan to gather huge amounts of data and use it themselves for direct marketing.  Others hope to gather huge amounts of data and pass it on to others,  charging  money just for the infrastructure.  Some companies are saving large amounts of data, and have no idea what to do with it.  They just think it is good business to start hoarding what they can.

What follows are some of the different ways data is gathered, held and sold to the highest bidder.

  1. PII and log info go back to a central computer.   Viewing data may or may not be separated from PII.  Aggregated viewing data used to improve collaborative filtering algorithms. Individual PVR machine uses downloaded algorithms to build profiles of users.  Content for all profiles is sent out over broadband connection.  Individual PVR machines know which programs to show, based on the profiles they contain.
  2. Similar to above, but profile is not created by software on the individual set top box, but by software running at a regional server, or “head end”.    
  3. The head office may have access to software like XTV AdManager from NDS.  Marketing departments attach tags to TV programs, parts of programs and commercials.   Software on the set top box, or the head end, attaches tags to households and users.  From there, advertising executives can play the software as they would a video game –  instructing the system to show the right commercials to the right families or family members at just the right time.
  4. Some companies are buying in third party data from consumer research companies.  SpotOn, for instance, is digging around in the lives of the residents of Aurora, CO, trying to answer such questions as “when will this person’s auto lease run out?”.  When that data is collected in one database, SpotOn will use it to send different commercials to each household.
  5. By analyzing the click stream that comes back from every television set, and combining it with third party information, it is then possible for a service provider to create a new, combined set of data.  Targeted advertising to the same community could then gain in value over time. 
  6. More adventurous service providers may wish to do a deal with the consumer data companies, offering to upgrade either their anonymous or personally identifiable information with analysis done on click stream data.  Everybody except the viewer gets something.
  7. If the service provider does not want to sell viewer data to third parties, the viewer data might be used for  “merge and purge” campaigns – combining what is collected with what other companies have to offer for the sole purpose of doing a mail shot or a targeted TV ad campaign.  After the campaign, the combined data is destroyed.  This allows a company to say “We do not give information to third parties,” while effectively handing them anything they want.
  8. Artificial intelligence algorithms are being used on set top boxes.  A box can start with, say, age, household size, marital status and number of children.  It could then be given family income and type of car or magazines taken.  None of that need come from the viewers themselves, although MSO’s are being encouraged to standardize their service subscription forms, so that the same types of information are available to advertisers across the nation.

But then, the software can take over.  It can ask questions, it can record and analyze each viewer’s click stream.  It can compare what the users do to a map of what people in the general public are doing.

This map is based on something called “collaborative filtering”.  Raw data from many users is attacked with artificial intelligence software to find patterns in the lifestyles and viewing habits of subscribers.  You can see this kind of software at work on, when you pull up the book 1984 by George Orwell and are told that people who read that  book also enjoyed Fahrenheit 451 by Ray Bradbury.

Collaborative filtering can be used in this straightforward way, simply offering items to people who have shown similarities to other people who purchased those items in the past.  But it can also be used to create and test theories about people.  Who are these Amazon customers who read 1984 and Fahrenheit 451?  Are they just people who like numbers in the titles of their books?  Probably not.  Probably they are members of the same market segment that could be described in detail and targeted across a wide range of products.  Software running on interactive televisions will be used to describe such groups of people.

How to Psychoanalyze Viewers

Some people don’t worry about electronic privacy, because they figure “What could anyone figure out from what I watch on TV anyway?  They’re welcome to it.”  But a technique like collaborative filtering is extraordinarily powerful because of its simplicity.  Just as you can find two books a person might like, you can just as easily find two people who share the same hopes or fears.

Pat Lalonde Dade is a director of Synergy Consulting, a database consultancy working with interactive broadcasters.  He has split up the populations of countries like Britain and the United States into a wide number of “value groups” – people who conform to certain cultural types within their country, and who are driven by the same psychological needs. 

This work is based on the theories of Maslow, who theorized a hierarchy of needs.  Some people are motivated primarily by need for food and shelter.  Some by a need for security.  Some by a desire to belong.  Others by a desire to succeed etc.

Dade has refined these groups, and found large numbers of people willing to answer his in depth questionnaires and sit in his focus groups.  For 25 years he has been building up data on these groups of like-motivated individuals.

If he could now find some easily identifiable behavior in the general public, that could be used to link the many people Dade has never met to his “value groups” whom he knows so well, then he could work backwards and point advertisers or marketers to people who would buy their products.  In fact, Dade hopes to go further.  He wants to be able to pick out consumers who will identify with his clients products.  He wants to help consumers ‘find themselves’ in frozen dinners and bottles of detergent.

The easily identifiable behavior?  Interactive television.  Dade plans to get people in his value groups watching television and then hopes to find patterns in their viewing.  These patterns don’t have to mean anything at all.  They just have to act like a fingerprint, linking you and me to one of his psychographic profiles.

"That's when the behavior starts coming,” says Dade, “Now that we know who this person is, we can tell from their interaction with the television that they use QVC or they use Discovery, more importantly during the breaks they're flipping around, they're not watching the ads at all, but they tend to look at the Fiat Uno ad every time it comes on.  I wonder why that is?  We can do that in real time."

What iTV Can Do With The Data

Gathering and analyzing data is only half of what interactive television does.  The payoff for all this diligent observation is being able to send targeted content to individual households, and even viewers within those households.

How an Interactive TV Can Respond to Data

Like a web page, interactive TV will offer different services and entertainments all over the screen, all the time.  What you watch, what buttons appear, and what happens when you click one, can change in the following ways:

Programs offered - Everyone will see their own trailers, teasers and pop-up reminders.  When they tune in to a show, they may even see their own version of it.  This techniques will be especially effective where product placement is used.

News - Any program or service that offers information can be customized into special issues, local editions and individualized bulletins.  Headlines, pictures, text and video can all be changed from minute to minute, person to person.

Commercials - Anything that sells something will be targeted to people who might buy - whether streets, schools, households or individuals. Infomercials can be tailored to provide different information to each viewer, depending on what is likely to influence their buying decisions.  Expect commercials to pop up in games, on-line services, programs and anywhere you don't seem to mind seeing them.

Branding - Likewise, product logos and product related images will be placed on your screen when and wherever they are most effective.  Advertisers are especially keen to exploit online services and children's games.  Look for the use of "virtual billboards" where, for instance, your local dry cleaner's name might be digitally inserted on to a building behind an international sporting event.

Tools - A wide variety of online services will be offered by interactive TV.  Email, personal diaries, financial planners, chat rooms and so on.  Expect them to be products in their own right, with special versions for boys, girls, working women, teenagers, retired men, etc.  Which are offered, how they operate, and how they are tied to other products will all be open for targeting.

Games - Aside from children's video games, many services and entertainments will feature 3-D environments where viewers can pretend to walk, run, jump or lie around.  What games are offered and what viewers encounter in these virtual spaces can all be customized and changed on the fly.

Forums and Chat - Email list services, message boards and virtual chat rooms can all be offered to different viewers.  The subjects for discussion, the participants and what they are allowed to submit can all be controlled.  In some chat rooms, viewers will be offered a choice of on-screen personality to inhabit.

Clubs and Organizations -- As with the internet, groups of people will be able to meet for fun or to conduct business.  But the entire communication environment will be under much tighter control.  What groups exist, how you find out about them, and what they can do online can all be decided by your service provider.  You will be informed about some, as if they were TV shows.  For instance, expect a great deal of promotion for a chance to chat with your favorite TV star.  Others you will hear nothing about, or may be prohibited from operating.

Online Transactions - Shopping and banking through your television will be entirely customized to (or targeted at) you.  What you are offered, how much it costs and how much you pay will all change from person to person. You will be offered exclusive discounts and special offers.

Offline - Service providers and their marketing partners are not limited to responding on your television screen.  In response to what you do with your television, you may receive email, pop-up messages, mail by post or telephone calls.  How people in the real world, outside, respond to your viewing is limited only by their imaginations and the law.  Most times you won't know it is happening.


Four Strategies for Responding to Data

Simple Fit

As described above, it is possible to dig up data on viewers from third party sources and aim programming at them.

Simple Response

In this case, the television responds to data that it has gathered itself.

"Let's say you watch every nature program," says Virginia McMullan of NTL "and it becomes clear you're a nature freak.  Well, we can say "Here, buy this organic food or environmentally friendly washing powder."

Response with Collaborative Filtering

The television responds according to the marketing theories of its software programmers.  But it also responds in an intelligent way, using more of what has been shown to work in a particular situation, regardless of why.

Data analysts explain in Spy TV how a television can decide to begin bargaining with a viewer, changing price and playing on viewers’ emotions.

Evolving Response

If you do something once or twice and it works, you do it again.  If it fails a couple of times, you might try something new.  These simple, everyday principles of trial and error become hugely important when a television company is able to put them to use on millions of people, four hours a day, every single day of their lives.

Over and over, if you visit the websites of interactive software developers, you will see “the loop”, some shiny graphic depicting the way their software works on viewers over time.

When talking about their paying customers, the men and women interviewed for Spy TV and the new report by the Center for Digital Democracy repeatedly slipped into the language of lab rats and salivating dogs.

"Direct marketers have been doing that for a long time anyway," says Howard Hughes of NTL.  They will start with a set package to mail or message to say, a set group of consumers to address or telephone, and their method of consumer response.  After a while, they work out a normal response rate to use as a benchmark. "Once in a while," says Hughes, "they will test a different creative with a small group, and if it turns out to be a lot bigger, a lot better result than what their standard is, then they'll run with that creative across the whole group."

Direct marketing has always been like the board game Cluedo (called Clue in America). The perfect junk mail campaign is hidden, and to find it the marketers stand in different rooms on the board, calling out what they think it might be.  "I think it's the yellow pamphlet mailed to lower middle-class families, asking them to call our toll-free number".   They could move systematically from one room to another, eliminating any mailings that didn't work.  But, as in the board game, that strategy might alert their competitors and customers.  So they have to disguise their intentions.

The real problem is time.  Each turn in this game takes weeks or months.  For each turn, a company might have to design new material, do another mailing,  await the results, enter the results, analyze what can be changed, run surveys or focus groups asking people about the new and old designs, and then do more mailings.  Telephone calls are more flexible, but cost more per person reached.

"Even that has a very slow cycle," says Howard Hughes, "and it's very expensive.  On the electronic medium it's not.  The turnaround time is going to be a couple of days when it used to be six months."

Interactive television will make entire direct marketing campaigns as easy to do as buying a classified ad in a newspaper.  It means more variables can be tested, more groups can be targeted for less money, and more people who don't respond to one message can be shown something else. As the cycles of trial and error turn more quickly, advertisers can break their messages down into smaller and smaller segments, approaching the ideal segment of one.  Marketers can stop thinking in terms of individual campaigns and treat the entire system as a single engine.

Control of that engine will increasingly be automated.  Once again, as computing power is hooked into every home by interactive television, data will find data.  Adaptive systems can be used to control the entire process of elimination described above.  It is learning process, ideally suited to the systems already used to gather data.

Here Robin Melvyn of NTL speaks with the new vocabulary of home entertainment:

"You have to create some control group testing, in effect throw people some placebos.  So if we're trying to increase their spend, or increase their usage or increase their customer satisfaction scores, we'll take one group and split it down the middle and expose it to two separate batches of data presentation." 

A good example of how such systems work is a robot created at the University of Sussex. Looking like a toy car with an insect's head carrying a microchip, the robot was placed in a large square box and went crashing into walls.  Its software was told to avoid the walls, but not told how. 

So it tried different things, observed the results, looked favorably on anything that worked and began to distrust anything which led to a crash against the walls. To speed the learning process, its microchip was put into a PC and a virtual robot did weeks' worth of driving around a virtual box in just a few hours.  When the chip was put back in the robot, it drove in a neat square, hitting nothing.

Now, who programmed the robot?  Nobody.  It learned how to drive on its own.  How did it do that?  It just kept trying things, always measuring its progress and keeping a note of which actions worked and which didn't.  That is how any learning or adaptive system works, and that is how software based on such systems will run interactive television.

It is a matter of pride and excitement to people like Howard Hughes that a selection of commercials and groups of viewers can be fed into this system and, after a few days, out come the sales!

Our Home

This is where the next generation of human beings is going to live: inside this cycle of persuasion, observation, refinement, and new persuasion.  It may be pleasant for them; it may be hellish.  But whatever it is, it will no longer be democracy.

None of the techniques listed in this article are limited to the work of selling soap or toothpaste.  Every one of them has already been applied in various countries to the work of selling politicians, worldviews and political ideals.  The only thing that has been missing is the automation, the two-way pipe in and out of individual households and the computers to make the decisions.