What is programmatic advertising? All you need to know

Programmatic advertising is revolutionizing ad buy-sell process

Programmatic advertising is revolutionizing ad buy-sell process


Programmatic buying is a buzzword these days. Just as many other fields, machine learning is fast expanding its influence in the digital advertising market. And most of it is happening via the programmatic advertising route. Various estimates suggest that in the long term, programmatic advertising spending will eclipse all other digital advertising mediums.

What is programmatic advertising?

It is a hard truth that several misconceptions exist in the market on what programmatic buying is. While some believe that it is another name for RTB (Real Time Bidding), another few think it is a completely automated ad buying process that requires no human intervention. Both these interpretations are incorrect. It is with the intent of clearing up these misinterpretations, along with many others, that I have compiled this post.

Simply put, programmatic advertising involves buying digital advertising impressions via a trade platform auction, one impression at a time, using automated mechanisms. The buyers from these automated systems are called Real Time Bidders and the systems that run the auction-buying mechanism are called Exchanges.

What are the types of programmatic advertising?

Broadly, programmatic ad buying employs either the Price-Based Trading route or the Value-Based Trading route. In the former, the focus is on cost optimization by buying cheap impressions. In the latter, the system learns from ad responses and values the next impression based on data analysis, employing factors of probability of responses and their values. Value-Based Trading, by virtue of its learning mechanism being able to decipher efficacy of ad impressions, is more adept at maximizing returns on investments.

As indicated earlier, programmatic advertising is not only RTB. Below, I have highlighted key processes within the programmatic ad portfolio.

  1. Real Time Bidding – Real Time Bidding is also known as Open Auction as inventory prices are decided in real time through an auction. The system is so designed that buyers and suppliers aren’t in a dependent relationship. Publishers, if they want, can block advertiser types and also set floor price for their inventory. Even the publisher’s identity isn’t always revealed to the advertiser.
  2. Invitation Only Auction – Invitation Only Auction is a type of RTB where publisher have restriction rights on advertisers for the auction. Publishers run such auctions and only invited advertisers can participate and bid.
  3. Unreserved Fixed Rate – These are also known as Fixed Price Auctions. In these, publishers prioritize deals based on the advertiser’s backend ranking or the pre-negotiated price. In most implementations, neither the ranking nor higher-priced auctions are exposed at bid time, defeating the “first look” purpose of the programmatic deal and obscuring the execution workflow.
  4. Automated Guaranteed – In this process, automated mechanism of RTB is employed for traditional direct sale advertisement buying. Buying process happens directly between the buyer and seller, with guaranteed inventory and pricing. Also, the process runs at no priority disadvantage over other direct deals in the ad server.

How does it actually works?

Typically, publishing ads on a web page means you are running JavaScript tags from ad networks on that page. Ad networks serve ad media using the tags. What it also implies is that the ad network has access to the website’s traffic or “inventory” which it then bundles and sells off to advertisers or programmatic technology platforms. Most ad networks, including Google AdSense, work in this manner. What happens next is that technology platforms or Demand Side Platforms (DSPs) plug into these ad networks and buy remnant, or unused, inventory from them at bid rates. It is here that the DSPs or the programmatic ad exchanges use their data and analytics to target user specific ad impressions.

How does the tech stack of a programmatic ad exchange looks like?

Broadly, it has six elements. A Campaign Management UI/APIs, State Update Engine, Real-Time Ad Auction (ad decisioning algorithms), Real-Time Data Collection, Distributed Data Processing/ Optimization (Hadoop, Python) and Analytics Capabilities (distributed column cluster stores).

Programmatic advertising companies specialize in building decision making algorithms. Their ads are mostly hosted on CDNs (Content Delivery Networks) such as Akamai, AWS, etc. CDNs best suit the requirements of local, efficient and cost effective deliveries of images and videos. A related terms to programmatic advertising is programmatic stack, which refers to the resource pool set employed by agencies and advertisers for marketing campaigns at each step of the relationship (say, advertising, on-site automation, e-mailing, etc.) with their customers.

Are publishers happy?

Despite its merits, members of the publishing community still considers the programmatic ad buying a mixed bag. While it definitely helps the community monetize the remnant inventory, the monetization rates are nearly always much lesser than the direct buying processes. With maximum selling at maximum prices being the objective, their wish-list is a little different. And that is the reason why most publishers usually maintain a private marketplace on top of their stacks and structure their ad exchanges in a priority list to bid on remnant inventory.

Whatever these members might think, programmatic ad buying is on a growth spree. The best testimonial of this trend is the mushrooming of programmatic ad companies in the market. What is undeniable is that in this market, just like many others, innovation will hold the key. And the player(s) that focus on unifying the user experience across mediums and devices, offline and online, will stand to gain.

Anubhav, a data scientist, writes about new developments and future trends in the machine learning and data analytics domain.
He can be reached at anubhav@thinkbigdata.in
Follow him on Twitter at: https://twitter.com/think_bigdata

Leave a Reply