Opening up the “Black Box” of Digital Marketing and the Role of Programmatic Advertising
By: Jeff Chester | Jun 12 2015
There is a growing and much needed debate on the role that algorithms and machine-driven decision making play in our lives. The use of “Black Box” assessments of individuals to determine what kind of financial product to offer is raising legitimate concerns about discrimination and unfair practices. Why are some people targeted for a high-cost payday or higher interest credit card, for example, while others receive better terms and conditions?
The answer to such questions can be found inside the “Black Box,” in which there are very clear objectives from businesses designed to effectively use all the new power they now have due to the merging of “Big Data” technologies with our always-connected (and data-generating) online way of life. Corporations have armed themselves with the latest tools to harvest and analyze the ever-growing flow of information available. Through “Data Management Platforms,” alliances with giant data brokers, and through effective use of the powerful digital marketing apparatus created by Google, Facebook and many others, financial, retail, grocery, health, education and nearly every other sector can make decisions about an individual in lightening speed. Companies can now reach us with offers in milliseconds, regardless of whether we are using a mobile phone, personal computer or other connected devices.
Over the next few months, we will explore the business models driving the “Black Box.” But today we will examine one crucial element that enables companies to so easily take advantage of our information to assess and influence our behavior. So-called programmatic advertising is a data-driven system that allows companies to “buy and sell” individuals in real time when they are online or using their mobile device. In today’s digital marketplace, our “profiles”—the information gleaned about who we are and what we do—is traded as a commodity. Pioneered in the U.S. by Yahoo, Google, Rubicon Project, Appnexus and others, this little known system is quickly dominating how online marketing and advertising operates throughout the world. Programmatic advertising relies on superfast computers that keep tabs on where we are online, so we can either be sold to the highest or special-interest bidder or labeled as someone not worth targeting at all (in ad terms, these people are classified as “waste”).
Regulators in the U.S. and EU have not done a good job addressing the privacy and consumer-protection concerns raised by programmatic marketing. It’s a key area in which CDD has played a role advocating for a more responsible regulatory approach. But for now, we want to highlight some of the features of programmatic marketing by excerpting from the new “Adgorithms” IPO (which just went public in the UK). We think the excerpts provide an opportunity for the public to peer inside an important part of the “Black Box” machine that is increasingly dominating our lives. Take a look and we will return soon with a discussion. For more information on programmatic advertising, see adexchanger.com and exchangewire.com.
Excerpt: Business Overview
The Company's software, Albert, is a proprietary artificial intelligence based programmatic platform, which plans, identifies, prices and delivers relevant advertisements in multiple fields of online advertising.
Using complex algorithms, historical data and artificial intelligence, Albert seeks to predict user intent and deliver advertisements that are likely to engage that particular user and result in higher engagement for the brand. It analyses the available advertising opportunities on the advertising exchanges, decides which one of them is most relevant and ultimately determines the right price to pay for a specific impression. The advert is then displayed on the screen of the user. This whole process occurs in under a second.
The accuracy of Albert improves with every online advertisement it delivers, as it incorporates new data whilst continuing to learn from previous data. This ability of Albert enables it to adapt to changes in the marketplace in order to capitalize on opportunities and to minimize purchasing of non-effective inventory, including fraudulent advertising activity.
Understanding of consumer behavior patterns
Albert has powerful and actionable insights into consumer behavioral patterns and web properties that it can leverage, such that the Directors believe it is able to target the most relevant audience for a particular advertising campaign more effectively, and achieve KPIs set by brands quicker and more cost effectively, than its peers. Adgorithms offers clients targeted online advertising via demographic, geographic location, time of day and behavioral characteristics.
Adgorithms works directly with companies who wish to advertise their goods and services, and also with media agencies working on their behalf to optimize advertising campaigns. At the outset of an engagement, the Company is supplied with creative materials, such as a banner or video advertisement, a pre-defined KPI to launch an advertising campaign and an advertising budget. Examples of KPIs include a number of user click-through on a banner or a user watching a video advertisement for a specified period of time. The creative materials and KPIs are inputted into and processed by Albert, following which it bids for impressions in real time based on observed or predicted user intent. Albert will then optimize the performance of the campaign until the KPI is reached. Albert does this by using its own data and also proprietary data that its clients provide (including data in relation to which users have the highest value to the client), contributing to the feedback loop.
By using Albert to determine the right price to pay for a particular impression, the Company has a proven ability to maximize ROI from a client's advertising budget and reduce customer CPA. The automation of the campaign management by Albert also minimises the need for human intervention, creating efficiencies and reducing labour costs for the advertiser, and particularly for media agencies (which often manage many campaigns concurrently).
The Company's technology solution enables online advertisers to efficiently and effectively engage and convert customers. Its solution is comprised of the Adgorithms software, called Albert, data assets, the feedback loop and access to display, video, mobile and social advertising inventory through the online advertising exchanges.
On a daily basis, the Company is presented with billions of opportunities to deliver an advertisement to users when advertising impressions become available through the various advertising exchanges. For each impression that becomes available, the Company has realtime software systems that recommend an advertiser's specific creatives (e.g. banners or videos) based on a prediction of the likelihood of a user engaging with an advertisement. Albert is designed to determine the most appropriate advertisement to show to the user and determines what price to pay for the advertising impression.
The core of the Company's solution involves:
· determining a user's engagement with display advertisements, which is a relatively rare event that requires a large sample size of relevant data to accurately predict;
· obtaining a large sample size of relevant data, which is difficult, in particular where the most relevant data points are also the most sparse e.g. very recent data on specific product interest; and
· building powerful, scalable and flexible systems that operate both accurately and quickly, between the time a user navigates to a page and an advertisement is delivered.
Albert is designed to continuously download data from advertising exchanges, analysing and storing it in its internal database. It then re-calibrates its prediction models so that the prediction and bidding are constantly up to date with new media sources available through the exchanges and with the ever-changing pricing and quality landscape of existing media. As those internal predictions are updated in Albert, they are propagated to the various exchanges so that customer's campaigns running in the exchanges can bid more aggressively for opportunities that are considered positive for that advertiser, and less aggressively or not at all for opportunities the software now considers less favourable.
To achieve those performance goals, Albert acquires hundreds of gigabytes of data daily, containing information on impressions, engagements and conversions, and performs tens of thousands of updates every hour. It collects and analyses information on millions of online websites and mobile applications that are available for it to advertise in, evaluating the performance of each of the campaigns it manages in those websites and based on that generates prediction for future performance of advertising campaigns in relevant media spots. Using this feedback loop, Albert is able to choose from the tens of billions of available opportunities daily, the hundreds of millions of impressions it predicts would be optimal for its customers.