Here is a fun fact: An average urban wakes up to be involuntarily exposed to an average of 2000 brands per day in different forms of advertisements. It is quite obvious why advertisements are necessary at the right time and only for a specific consumer subset. Brands love targetted native programmatic advertising. Brands love to understand their market audience and still are looking for specificity in clientship which has become a necessary bottom in recent campaigns. ADskore, a programmatic advertising platform, which once heavily relied on managing direct marketing strategies, understood the growing importance to migrate from internet to mobile advertising, especially in growing mobile and app based services market. The challenge was a difficult one — to work with mobile publishers and ad-exchanges in an ecosystem that demanded processing of bids in a span of few milliseconds.
Automated trading helps brands, especially mobile advertisers and publishers who can harness RTB to their advantages using variety of inventory creatives that requires monetization. The mobile-based advertising business with its monopoly at play, has long faced an unlevel playing ground — inventory demand-supply managed by the same corporations. It’s quite fair to acknowledge both the buy-side and sell-side providers try to outweigh competition in an already convinced market. The ecosystem over the years has become tad complex, especially involving several exchanges within. The playing fields change significantly when the ad-inventory trading becomes more programmatic. With the help of app developers, the platforms segment all inventories, determine what traffic exchange partners actually need. With better data platforms, there is lot of confidence that an “impression” (consumers meeting specified criterias) is a true positive by analyzing certain outliers.
How real time bidding works with an AD? Real-time Bidding (RTB) is a server to server integration option for network buyers using the Ad Exchange to evaluate and bid on each available impression. The process begins when you visit a publisher’s web page or perhaps play a game. The publishers are those who generate content such as news, music, information, and sports; provide advertising slots for ad-exchanges. The publisher’s ad-server that has built in “business logic” to choose what happens next.
The framework identifies potential prospects based on rules like — Are there any premium buyers? What do I know about the consumer? Which advertiser is the best fit? And, if the advertising slots are not reserved to any specific brand or exchanges, the publisher ad server makes it a point to open-auction for a good deal. The publisher connects with the SSP (supply-side provider) server which pragmatically analyses the consumer using an existing cookies lookup data provider which is in-house managed at SSP. The information is exchanged with the AD-exchange, whilst the exchange keeps itself busy toggling between many DSPs (demand-side platforms) which have potential clients, or even other exchanges.
Consider a scenario, when a brand is interested to sell its fashion accessories to urban females living in South Asia, with a probable age above 25, and it can at its fair budget, spend $1 per 1000 ads that are served. This is known as a pre-cached bid. If a pre-cached bid is full-filled it speeds up the process of programmatic bidding. Well, if to consider a situation where there are no pre-cached bids? The exchange now plays “the house”, and open-auctions the impressions (consumers meeting the specific criterias) and requests bids to be placed in less than 10 milliseconds. To put in context, it takes 300 milliseconds for an average person to blink his eye. [IAB] Despite this seemingly unrealistic requirement of the exchange, the buyers respond in that time and the exchange selects the winning bid. The winning DSP passes instructions to the Ad-exchange for retrieving an ad creative for native programmatic advertising. The exchange passes the instructions to the SSP, and the SSP sends the request to the publisher’s Ad server. And, finally the publisher responds to the still open HTTP connection, redirecting the ad-slot to “consume” the advert from the agency AD server. Apparently, the agency Ad-server tracks the performance for adverts, and records this request as an impression. Now, the web server can finally render the advertisement appropriately with the web page content, resulting in an advertisement most appropriately matched for your requirement.
Lets take a quick intro into bidding before being “real-time”. For example A bids $10 per slot, and B bids $8. Assume B raises the bid to $15, and A raises this by $20. By so, A has been betting on a bigger eCPM. Now, consider the wait situation. If the gap between the bids are small, competitors are right behind you. There are other situations, where you could strike gold, waiting for the perfect times to bid high & lower CPM. In an volatile market, you can either expect a perfect win-win or a blood bath. In the Game Theory, there is phrase called “winner’s curse” which suits the scenario. Now for another good instance when A bets $20, while B bets a mere $15 for the same impression, A wins the auction, and has to pay a fair eCPM of $15. But, raising the stakes could be an expensive affair, far more like gambling, especially for retargeted impressions. The thumb-rule for the win-rate would be find the percent of impressions per bid. The win-rate is a factor which decides what percentage of bids you would be winning per open auction.
ADskore required a faster, more efficient way to sift through their supplier-side platform’s data provider. The high-velocity spatial data from the ad-exchange needs a comprehensive analysis with real-time “impressions”. It required was a more faster processing of specific queries, to keep the engagement with consumers and exchanges and — bringing them together in real time. Why HANA, was a big question for a small-medium company. The licensing in 2013, was a little shy of $5 per hour, per instance without the data charges, which was available on Amazon web services. Not a cheap option, but at inflection times as this, the need demanded an “in-memory” product like SAP HANA to outweigh competition. “To be successful in the business of real-time programmatic advertising, the need demands to make the data platform quite efficient, with an emphasis on velocity over variety or volume, enough by a factor at least 50% more than your biggest competitor.”
After replacing few home-grown applications, the project went underway in late 2013. A few months of working with SAP’s big data product, gave significantly lesser time to generate the suppliers, lookup app-publishers. The core applications — are used to generate approx 80 million impressions per day. The platform served ads with minimal processing times and no latency was faced in weekends. The best number served, clocked nearly 1.2 billion. In the end, the ad revenues is all that matters over CPMs (cost-per-thousand-impressions).
Processing Real time bidding using SAP HANA - With the HANA DB, the ADskore platform could offer agencies an easier opportunity for cross-market and cross-publisher targeting campaigns. It enabled their ad-partners to deliver mobile advertising that is both target group-specific and highly-effective in measuring branding. Its platform now gives marketers the opportunity to tap into new advertising budgets as well as potential new sales. The products make it possible to use data from external sources to reach very specific target groups with highest level of accuracy.
