RTB Explained: How Ad Tech Learned From Big Finance

The computerization of the finance industry began in the 1970s when they began to use computers to efficiently route orders through the exchanges. Realizing the utility of these machines, they added tasks and responsibilities to these automated systems in order to speed up processes that were traditionally completed by humans.

As these technologies were developed, people felt more comfortable allowing machines to complete these tasks. Once a sufficient level of trust was reached, the institutions realized that simple trading could also be automatically completed by the systems, in addition to the routing-type operations that were already automated. The systems were empowered to make micro-transactions at much higher frequencies than were possible with human-conducted trades. Because of the high frequency, each trade was done with a very small margin of profit; however, the sum total of all of these small trade margins add up to substantial profits over time (a familiar sounding strategy, even used as the main plot point in some films).

As the financial institutions realized that they could just let these machines run - making money without supervision or interaction with the human traders - they gave more trades to the systems to handle.

This cycle continued without much interruption for many years, until 2012, when the Knight Capital Group lost $440,000,000 when they had an issue with their automated trading systems. This event was called a flash crash a situation where errors made by the machine were compounded by each transaction, resulting in much more damage than might have occurred if this bug had not been present.

The end result of the compounding errors was a crash in the market. Luckily, it was corrected fairly quickly, mostly due to the fact that the errors did not reflect the true state of the market, and the market was able to adjust back to pre-flash crash levels. Yet, companies still need to be diligent to avoid similar situations and demonstrate stability and consistency,  in order for investors and partners to have any faith in the system. No one wants to invest in technology that has the ability to crash without rhyme or reason.

The advertising industry is currently going through a similar transition - allowing machines to handle high volumes of small transactions - and there are lessons that can be learned from the financial sector. As ad tech embraces programmatic techniques, there may be similar challenges to overcome.

Are there risks that can affect advertising technology in the same way that they've affected the financial industry in the past?

The good news is no, advertising does not have the same risks as finance. Because ad technology is mostly about selling one impression at a time - impressions are not traded back and forth in the same way that stocks and commodities are - the industry is able to avoid most of the issues that arise with automated trading (like compounding errors on mispriced trades, or a limited ability to realize that a transaction is incorrect). Since all of the transactions are effectively one-time events, it is difficult to get into a downward spiral where the end result is a flash crash. The worst that can happen in most cases is that the best price for each impression is not captured, or that the impression is not monetized at all.

The fact that the effect is reduced doesn't mean that faulty systems can't have an impact on the digital advertising space. Mistakes could cost the industry billions. The key to being successful with automation in the ad industry is finding the balance between filling impressions effectively and mitigating risk.

How does all of this relate to the financial industry's adoption of automated systems?

The mistakes made in the development of the automated systems, and the consequences of those mistakes, are lessons we can use to ensure that our design decisions today do not cause issues tomorrow.

Automation can accomplish tasks that were previously handled entirely by people, usually with increased efficiency, but great care must also be taken to ensure that the same quality of service is provided by those systems. One of the benefits of having humans handle transactions is that they are able to use their best  judgment at every step of the process. Moving to an automated system to accomplish these same tasks eliminates the opportunity for a human to make decisions, and opens up the possibility for the systems to be misused. While it's not a mistake to let machines handle high-volume transactions, the removal of human judgment from the process means that the system needs to be designed to mitigate potential issues.

One current example of this human-system transition is the relative ease with which fraudulent ads can be sold through the networks. Where financial institutions had to worry about the wrong trades being made, advertising has to worry about serving fraudulent ad impressions. The mechanics are different, but the outcomes - potentially large amounts of lost revenue - can be the same. Automated systems have to be designed in such a way that advertisers and publishers are protected from undesirable consequences. As the advertising technology matures, the risk of fraudulent behavior should decrease, but it may take years to reach that point (much like the pains that financial institutions went through to get to a point where their systems were reliable and consistent.

The use of automated systems to produce revenue from the financial market enabled the companies to concentrate human effort on higher impact strategies. Programmatic advertising tools should be used to help achieve maximal revenue for publishers with minimal effort so that they can then concentrate on the content and experience for visitors of their web properties. Better advertising leads to a better experience for users of the site and supports a healthier ecosystem overall.

Did you find this article interesting?

Sign-up to get more from the Sortable blog as we continue to explore the ad tech industry, tackling subjects like: Google's position in the market, how AdX rules affect CTR, and question whether above the fold ad placement really matters.