Data Needed in Digital Advertising

 Data in Advertising

Digital Ritesh

Data needed in the Advertising

Retailers, banks, governments, social networks, credit reference agencies, and telecom companies, among others, hold vast amounts of information about customers. They know where they live, how much they spend, their lifestyles and opinions. Every year the amount of electronic information about people grows as they increasingly use internet services, social media, and smart devices.

Large and complex data sets have existed for decades, in that sense, Big Data is nothing new, however, by the early 2010s “Big Data” had become the popular phrase to describe databases that is not just large, but enormous and complex. There isn’t a universally agreed definition of “Big Data”, but the features of Big Data that are considered important are:

  • Volume: any database that is too large to be comfortably managed on an average PC, laptop or server can be considered Big Data. Big Data is generally taken to be on a database that contains more than a terabyte of data (1TB = 1000GB). Some Big Data sources contain petabytes of data (1PB = 1000TB).

  • Variety: Big data contains many different types of structured and unstructured data. Structured data is tidy and well defined and can usually be represented as numbers or categories: for example your income, age, gender, and marital status. Unstructured data is not well defined. It is often textual and difficult to categorize: e-mails, blogs, web pages, and transcripts of phone conversations.

  • Volatility: Some types of data are relatively static, such as someone’s place of birth, gender, social security number, and nationality. Other data changes occasionally, such as one’s address, employer, or the number of children that you have. On the other extreme some data is changing all the time: for example, what music you are listening to right now, the speed you are driving, and your heart rate. Big data is often volatile.

  • Multi-sourced: Some Big Data sources are generated entirely from an organization’s internal system. This means they have control over its structure and format. However, Big Data often includes external data such as credit reports, census information, GPS data, and web pages, and organizations have little control over how it’s supplied and formatted. This introduces additional issues around data quality, privacy, and security, over and above what is required from internally sourced data.

Companies proactively search for and obtain new data: they bring all data together and analyze it to produce insights about what people have done, what they are doing, and what they are likely to do in the future, that influence their decision making and what actions to take. 

A Big Data philosophy is about taking a holistic view of the data available and getting the best out of what a company has. If an organization is doing this, then it doesn’t matter if it has a few megabytes or many terabytes of data; if they are structured or not, and where it comes from. From a technology perspective, one seeks out IT solutions that deliver the required storage and analytical capability.

A myth about Big Data is that you need a huge amount of data to build a predictive model. A couple of thousand customers are more than enough, and many useful models have been built using less data than this. Obviously, the more data you have about people, the more predictive your models will be.

Big data has attracted so much interest in recent years because it goes beyond the traditional data sources that people have used in data mining and predictive analytics in the past.

In particular, Big Data very often are: 

  • Textual data. This comes from letters, phone transcripts, e-mails, web pages, tweets, and so on. This is unstructured data and therefore needs a lot of processing power to analyze it.
  • Machine-generated data. GPS data from people’s phones, weblogs that track internet usage, and devices fitted to cars. Machine-generated data is generally well structured and easy to analyze, but there is a lot of it.

  • Network data. That is information about people’s family, friends, and other associates. What is important is the structure of the network to which an individual belongs, how many people are in the network, who is at the center of the network, and so on. It used to be that the prime source of data for all sorts of predictive models was well structured internal data sources, possibly augmented by information from a credit reference agency or database marketing company, but these days Big Data that combines traditional data sources with these new types of data is seen as the frontier in terms of consumer information. 

The problem is that there is so much different and varied data around that it is becoming increasingly difficult to analyze it all. The IT and the analytical community are improving proportionally, the first is continuously developing their hardware ad software to obtain and store more diverse data, and the latter is trying to find better and more efficient ways to squeeze useful insights from all the data that IT solutions have gathered.

One feature of Big Data is that most of it has a very low information density, making it very difficult to extract useful customer insights from it, that because a huge proportion of the Big Data out there is absolutely useless when it comes to forecasting consumer behavior. Companies have to work hard in order to find out the useful data that will improve the accuracy of their predictive models: they need big computers with lots of storage, and clever algorithms, to find the important stuff amongst the chaff.

In that context, there is a lot of money to be made selling Big Data solutions, and whether the buyer actually gets any benefit from them is not the primary concern of the salespeople.

In some cases, the benefits of Big Data can be too small to justify the expense: in banking, for example, the potential for new Big Data sources to improve the predictive ability of credit scoring models is fairly small, over and above the data already available, that is because the key driver of credit risk is past behavior, and the banks have ready access to people's credit reports, plus other data supplied by Credit Reference Agencies such as Equifax, Experian, and TransUnion.

On the other end, the marketing teams that are trying to identify people who might be interested in their products but have no data to go on will search externally in Big Data company.

For those already using predictive analytics, Big Data is very much the icing on the cake once there are very good IT systems and good analytics in place, however, if internal data systems are inefficient, they won't store as much data as they could and if analytics culture is not well developed, Big Data solutions are not the next step.

Companies should concentrate on making better use of the easy-to-access data they already have before moving on to more complex solutions. Unless an IT and analytics systems are pretty slick, a company needs to spend on incremental improvements to their current systems, rather than implementing a whole new suite of dedicated hardware and software specifically for handling Big Data. 

In terms of the percentage uplift that Big Data provides, that's something of an open question and is very dependent upon the type of predictive models someone wants to build and how much data they already make use of. If there is already good data and analytics, and a company implements a Big Data strategy in the right way, they may see a 4-5% uplift in the performance of predictive models.

If they don’t currently have much customer data, and Big Data gives the ability to predict customer behavior where this wasn't an option before, then they could be looking at benefits of significantly more than 10%. Another perspective is that the biggest benefit of Big Data has little to do with enhancing existing models in well-run data-rich organizations.

The greatest opportunities for Big Data are where it is making new forms of customer prediction viable. Most existing healthcare systems are reactive: they treat you when you are already ill. Combining predictive analytics with Big Data makes it more viable to shift the emphasis to prevention.

It becomes possible to predict how likely each citizen is to develop certain conditions and intervene before the illness becomes apparent. This has the potential to add years to average life expectancy. Marketing is another area where Big Data is proving its worth. Combining information about people’s movements, gathered from their smartphones, with supermarket data about what type of food they like to buy, they can be targeted with promotional offers for restaurants in the city they are visiting before they get there.

Another marketing application is to use real-time information about electricity and gas usage to forecast when someone is likely to be at home, therefore a good time to contact them. These applications of predictive analytics are where the frontier of Big Data and predictive analytics currently lies.

This is the series on Digital Marketing.
will be coming with new aspects of Digital Marketing Every other day...
if you have any questions regarding this topic?
comment below... I will try to find out a solution for it...

Also Read,

Levers of Digital Marketing
Digital Marketing Vs Traditional Marketing
How Digital Marketing become Best
What is Digital Marketing?? Where To Start?
What is Web Hosting? Most affordable Web hosting sites.

Post a Comment