Online advertising is helping corporations monetize the Internet by transforming the way corporations are interacting with and marketing to their customers and prospects. The effectiveness of online advertising is so dramatic that corporations are constantly shifting their budgets to the digital channel, taking share away from other forms of advertising such as print and radio. Corporations are attracted by the improved targeting, ability to easily experiment with formats and messages, rapidly adapt to changing market conditions, transparency (through measurement and attribution) and cost-effectiveness offered by the various types of online advertising (search, display, social) and the channel through which it is delivered (desktop, mobile, digital radio). At the core of these attributes is big data management and analytics.
Marketing is about acquiring, retaining and growing the value of customers. Online advertising is proving very effective in achieving these goals. It is proving particularly effective for demand-generation through increasing awareness and interest. As a result, and as is shown below, the share of spending for online (or digital) advertising as a percent of total amount being spent on advertising has been increasing and is projected to continue to increase.
Like with every other form of marketing, for customers and prospects to want to engage through online advertising they expect to receive through appropriate channels contextually relevant messages that are specific to them. In a recent study conducted by Forrester, consumers showed that they notice online ads more than other forms of advertising even if they don’t click on them.
Big data is playing an increasingly important role, perhaps the central role, in the improving effectiveness of online advertising. This is because over the past couple of years online advertising has been moving to programmatic buying. Programmatic buying refers to the practice of automating the buying of online ads by using algorithms to drive the best possible price for each impression. This occurs in real time, on demand and on an impression-by-impression basis. Real-time bidding, RTB, refers to the impression-by-impression buying. As it is shown below, programmatic buying saw a significant influx of activity in 2012, growing over 100% in the US to $2.2bn, according to IDC, now represents close to 16% of display advertising, and is expected to grow to over 30% by 2016.
The share of programmatic buying is increasing because of:
- Premium ad inventory becoming available through exchanges developed by public and private companies, including the Facebook Exchange and Turn’s exchange. We expect additional platforms to enter the market such as the one being built by Amazon.
- The creation of private exchanges created by brand advertisers.
- Improved transparency in ad placement.
- The availability of video and mobile ad inventory, in exchanges such as Brightroll’s.
Programmatic buying in general and RTB in particular are generating big data. Interestingly, it is not the volume characteristic of big data that is important in this case. The challenge comes from the velocity and variety of the data that is being used in order to make decisions. In RTB decisions have to be made in milliseconds. To make a bid decision, the RTB system must not only use one of more predictive models that have been developed using machine learning techniques, but it must also combine and consider data that is produced, and often changes, at different speeds; some data changes very fast, other less so. Specialized Data Management Platforms, called DMPs, are starting to be used by programmatic buying systems to address the issues relating to volume, variety and velocity of the data. They integrate, manage, and analyze first-,second-, and third-party online and offline data that is used to significantly improve the targeting of online advertising, increase the ability to measure advertising effectiveness by performing more detailed attribution. As online marketing budgets are increasing, and the number of marketing channels is multiplying (for example, for online marketing alone we use email, search, display, social, mobile, video), the importance of attribution is increasing. Marketers are not longer satisfied with last-click attribution but they want to understand which marketing channels contributed to a customer’s decision and by how much. Marketing channel attribution analysis requires sophisticated big data analytics.
While online advertising can benefit significantly from the use of big data management and analytics technologies, digital marketers are facing significant issues applying these technologies effectively. There are two reasons for that.
- Big data technology is ahead of the use cases, even in online advertising which has been one of the first sectors that starting using these technologies.
- While we all want to believe that the world is moving from Mad Men to Math Men, the truth is that it has not moved there yet. Marketers today are asked to make decisions based on data and information presented to them via a multitude of dashboards and other increasingly sophisticated analytic solutions that are based on technologies such as big data, machine learning, real time analytics, etc. But they struggle with the proper use of these solutions often focusing on the wrong metrics, taking a short view of performance data, optimizing quickly on metrics such as clicks and “actions” but often ignoring more predictive metrics such as customer lifetime value (CLTV). Unfortunately, the ad agencies the marketers use, which are primarily staffed with creative people rather than data scientists, are not in much better position to help them.
To succeed in effectively using online advertising solutions and getting the best possible ROI particularly from programmatic buying and RTB, marketers but develop the right big data strategies. These strategies must begin with developing a proper understanding of the big data that is becoming available and being utilized by these increasingly sophisticated solutions, i.e., the “new big data,” rather than by trying to process the marketing data the organization may have stored and used in the past, the “old big data,” using modern big data analysis techniques. These strategies must provide the proper balance between under-utilization and over-reliance on the new big data. Finally, these strategies must provide the ability to leverage the new big data in a sustainable way to produce repeatable outcomes.