Forecasting in Media

October 2, 2006

This is just an emerging though based on a question asked by a friend.  What kind of forecasting do media houses do and whether there is a role of software in that.

Lets look at print, TV and internet seperately, as these are driven by different constraint..

In case of TV, the constraint is airtime. Increasing airtime means increasing channels and the decision of doing that is so big that it is taken at board level not at “editorial” level. There is a demand- supply economics which drives the available budget. There is a set of audience a subset of which the particular channel tries to capture. Unlike other material products, in case of media, the size of audience is variable. There are the youth who spend their time on online gaming but will consider TV if there is a relevant programme. There are sport fans who will follow the sport, regardless of the channel it is on. Then there are breaking news and big event coverage which drive the viewerships. The content teams judge their success based on the eyeballs they are able to get, and the organization judges its success based on the revenue it generates ( Subscriber and Advertisements) – which apart from viewership also depends on Demographics. Content type and “volume” of a given type are hence determined to drive viewership – from the targeted demographics.

Now coming to forecasting. One of the principles of forecasting is that you know the demand based on a set of known parameters. These known parameters in case of media may be

Judgemental Methods

  • Prediction based on known set of parameters – like our knowledge of most watched programmes during Planned Events ( Like World Cup)
  • Neilsen viewership metrics – Knowledge of demographics and type of content people are watching today ( like 17-21 year old guys  like to watch reviews of latest and greatest gadgets, and 40 year old men like to watch review of expensive cars and that 30 year old women like to watch business related reality shows, and that married men secretly like soaps though they show that they are forced to watch it because their wives wouldnt let them watch football instead)
  • Surveys: Viewers are frequently surveyed to find out what they like to watch, what they do, what would they like to watch etc. However surveys are always hampered by the sample set, and is hampered by peoples imagination for forecasted events.
  • By concensus : like the brainstorming sessions where people have lots of data, survey, research reports and their own experience – and they try to come up with a concensus of what they need.

The mathematical events – like time series – doesnot seem to be prevelant in these markets, however people tend to look at extrapolation and trends at judgemental level. What is more interesting is that the advertising side wants to run more reports and analyse this data from many different angles. They are the ones who use sophesticated software. For some strage reason, the content side doesnt play with this data. Maybe, I am talking about the wrong sampleset here.

Also – If a TV channel builds loyal audiences by catching them as teenagers, It may as well have them hooked as they grow up.

Breaking events: any big news has people wanting to know more, from all sources. It makes sense to cover them.

While these kinds of forecasting may give an idea on the kind of programming required, the quality and attitude of the programs themselves are important. What will click is not very easy to predict.

However by far the most popular and most effective means is experimentation. For each program, there is a pilot, which select audience watch. Based on that they decide if the show will go on air. The show is allocated a short time window in which  it has to prove its promise. The existing programs experiment by increasing the amounts of certain content and gauging the reaction by the viewership reports and surveys. For instance, increased lifestyle content (at the cost of “real” content), or increased take-away value may contribute to increased or decreased viewers interest.

Then there is a factor of excluisivity and your niche. The more exclusive the content is, the greater is its value.

When it comes to print, the constraint is the column inches. Adding more pages is usually not a very easy decision to make as it has huge financial impact. Will the benefit it give offset the cost of extra newsprint? Can the current set of printing machines print more pages and still manage to have newspapers on peoples doorstep at 6 AM? 

The print facilities take ages to set up and hence increasing capacity requirement must be forecasted atleast an year in advance.

A more effective forecast could mean reduced redundant capacity and hence significant cost savings.

Given that constraint, again the question is usually about relative proiority to different sections.  Most of the forecasting methods in either case is similar. Readership surveys – with similar demographics is available. The only difference is that the Time angle is taken away.

The internet media however is not so badly effected by such constraint

The servers ( for internet based media) should also be pro-actively sized , rather than reacting to overloads – or over-reacting to certain abnormal strikes or hacker attacks. And scaling server infrastructure has a lead time of weeks, hence being reactive is possible.

TV and Newspapers have a number of impression/viewership based advertising, not click/action based. This means that if the forecasting is not effective, the advanced sales may be too cheap if you are growing.

Now coming to what role can software play here?

Well, there is little role of Content management here. The prime role of software is in analyising the readership/viewership data.  On the web, it also means a good amount of tagging for the web-analytics system to pick up and analyze against.

Extrapolation of these numbers using a software helps. Assigning values to these numbers for determining ad prices is something which is already in place at most media houses.

In fact – the advertising side is so strong that you could go and  buy lets say an ad for 1 million teenagers and the companies will determind where and how many times to place these advertisements to get the given viewership.

As a concluding note, I would say that forecasting and use of software looks to be more prevelant for advertising and infrastructure support, whereas surveys and experimentation seem to be more relevant to editorial teams today.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: