The Impact on the Marketing of Being Able to Measure the Reliability of User Responses
My father was a real-life “Mad Man”. In case you are not familiar with my reference, “Mad Men” was an American period drama television series created by Matthew Weiner. The show was set in the 1960s and its plotlines focused on the day-to-day business of advertising agencies, as well as the personal lives of its characters. My dad worked on Madison Avenue in New York City for nearly 25 years. His expertise in the advertising industry was media buying. His clients relied on him to decide on the optimal way to convey their product’s message to their target consumer.
Media decisions back then were straightforward, you either bought time on television or radio-based on sampled ratings or you hung up your client’s message on a big static billboard on the side of a building or along an interstate highway. The data available to help my dad make the correct decisions was crudely limited and extremely expensive to obtain.
The Internet, of course, changed everything. In addition to creating a whole new medium to deliver product messaging, it completely changed the nature of how businesses deal with data. Data used to be stored in siloed centralized databases. Today the average individual spends hours each day connected to social media. The data set this connectivity leaves behind is harvested and used to create targeted product messaging delivered right back to their laptop or handheld device. Eerily, this whole process sometimes takes only seconds. Our news is filled with headlines about how third-party companies such as Cambridge Analytica have been able to “weaponize” our information.
Welcome to the era of big data marketing. The explosion of this field has happened so fast that it has left regulators in the dust. Governments are only now just starting to discuss who should own our personal data and who should monetarily benefit from that data. In 2016, the United Nations General Assembly adopted a new resolution on the right to privacy in the digital age The UN’s response is a good start but individual nations, especially the United States are responding at a much slower pace.
My company, ImagineBC supports completely the idea that it is only “We” the individuals who are entitled to benefit monetarily from our personal information, intellectual property, creativity and time. We are in the process of launching a new ecosystem that will make it easy and efficient for an individual to regain control over their personal information and help them leverage that controls to receive fair compensation for the value of their information. ImagineBC believes that AI/ML technologies can be used as a “weapon” of good and we are lucky to have partnered with WeR.ai, one of the leading technology companies in the field of data science to help us make this belief become a reality.
Companies have been looking for the most effective and efficient way to deliver their product message to as many interested “eyeballs” as possible for as long as business to consumer commerce has existed. Together ImagineBC and WeR.ai are working to employ AI/ML technologies to create a superior new way for companies to accomplish this goal, while simultaneously benefiting the very consumer the advertiser is trying to reach.
ImagineBC believes that rather than surreptitiously collecting data from individual’s social media and devices and using that data to create intrusive online advertising campaigns, companies should instead simply ask the individual directly to provide them with the information they are interested in and compensate the individual fairly for the time it takes them to respond. In return, companies are entitled to understand the reliability of the information provided by the individual.
The data scientists at WeR.ai, led by my friend and co-author Man Chan, are working on developing the technology to calculate what we call a RELIABILITY score. This game-changing a new piece of information will predict and communicate to any company running a marketing or advertising campaign within the ImagineBC ecosystem, the accuracy of an ImagineBC’s members responses to a survey or advertisement campaign.
ImagineBC selected to partner with WeR.ai to develop this new concept of a RELIABILITY score because WeR’s existing AI as a Service (AIaaS) model already offers a powerful tool to help enhance decision making while simultaneously controlling costs. WeR’s rapid AI commercialization technology makes it possible to leverage a large commercial library of industrial proven AI solutions to help ImagineBC address the challenges of improving both their client’s and member’s experience within the ecosystem. However, WeR is still very aware that with such power comes the need to exercise caution and responsibility.
Let’s visualize this concept or a RELIABILITY score with an example. Imagine a pet food company wants to start selling a new type of organic dog food. Certainly, the company would like to promote their new product to as many dog owners as possible. Using the tools provided by WeR.AI, the company will be able to quickly determine who among the self-identified pet owners within the ImagineBC community are genuine dog owners.
In addition to the survey response provided by the ImagineBC community members, the pet food company will receive a RELIABILITY score that predicts the likelihood a member’s survey response is accurate. If data truly is the most valuable asset on the planet today, imagine how valuable it will be for companies to receive an associative score that predicts the level of accuracy and trustworthiness of the data they purchase.
Since the scale of the data exchange between a company and consumer is ever increasing, we need to employ AI to mine and guide us through this new frontier. At WeR.AI we strive to create a medium between algorithmic and business model challenges while being conscious of the ethical challenge in AI applications. As we start to understand this new concept of RELIABILITY scoring, the one thing we must tread lightly on OR be wary of is to avoid the introduction of human bias into the exchange of opinions between the company and consumer.
Read More: Don’t Procrastinate on Benefits Renewal
“Unfortunately, we have biases that live in our data, and if we don’t acknowledge that and if we don’t take specific actions to address it then we’re just going to continue to perpetuate them or even make them worse.” — Kathy Baxter, Ethical AI Practice Architect, Salesforce
We need to be aware of the effects of marketing promotions and other factors can have on the exchange and avoid the possibility that the AI is designed to favor one demographic consumer group over other consumer groups. As a responsible AI practitioner and citizen, we need to be sure to maintain a deliberate, meticulous, and mindful balance when creating such measures.
Calculating the RELIABILITY score begins with understanding the nature and class of data points available. From the consumer, we can gain access to or harness two major classes of information or data points: Demographic and Behavioral information. Demographic information typically refers to systemic information such as gender, profession, age, incomes, survey opinions, geographic details. This information remains rather static or stable across time and is monitored by ImagineBC to ensure that members are not manipulating their profiles to increase their earning potential.
The second class of data is behavioral information. How would individual members react differently to a targeted advertisement for the new 2020 Toyota Prius? How often a member utilizes the Job Search feature in ImagineBC? Behavior information is how a member, directly and indirectly, interacts with products and brands. Surprisingly, for the purposes of determining a RELIABILITY score, we consider a member’s behavioral footprint to be largely more important than their demographic information. If our model focuses too much on demographics information, it will create a systematic bias. However, if we could find an optimal balance between a member’s demographics and behavioral information, we would be able to construct more accurate representations of the entire member or community pool.
Combining both demographics and behavioral information is by no means a straightforward or trivial process. It is not an easy feat. There are several approaches to creating a weighted scoring system. One common approach is to assign each of the quantifiable contributing factors a corresponding weight as described in this formula: SCORE = w0 +w1*x1+ w2*x2 ….. The weights essentially provide a relative strength for each of the factors in the measurement.
RELIABILITY score is very unique because it is not a straightforward function simply based on member demographics and behavior. RELIABILITY score of a member is dependent on the survey subject and how the questions are asked. And there is no one-size-fits-all scoring for a member. One can be reliable in providing opinion on certain product group while they can be more or less reliable for another product. RELIABILITY scoring is a dynamic system which uses statistical modeling to capture the different facades of a member.
Besides the algorithmic challenges, it also implies a different business model. Before the age of AI and Machine Learning, marketers would have to run their campaigns based on assumptions to simplify the business model; considering a member and product as homogenous. With the advances of AI, a more granular and precise definition of multiple facades becomes possible and allow room for targeted personalization. Marketers will have more powerful tools to carry out granular market research and precision advertising, e.g. market segmentation helps marketers to understand the constituent of their target market and their existing user base.
One commonly difficult problem in marketing survey and polling is finding a representative sample. Essentially, it is the premise of many statistical techniques. The most notable pitfall is the 2016 election polling of the U.S. presidential election. The poll prediction and margin of errors could be statistically sound, however, it failed to concatenate reality.
“AI will do analytical thinking, while humans will wrap that analysis in warmth and compassion.” ― Kai-Fu Lee, AI Superpowers: China, Silicon Valley, and the New World Order
Biases are ubiquitous. In the age of AI, RELIABILITY score is not a simple Machine Learning function to estimate one single score of a member; however, it is a dynamic system by taking both objective and subjective inputs from the member. In most behavioral targeting systems, similar scores are taken purely from the data a member passively generates e.g. web pages they are browsing. It is an indirect measure.
“Humane technology starts with an honest appraisal of human nature. We need to do the uncomfortable thing of looking more closely at ourselves.” —Tristan Harris, Co-Founder & Executive Director, Center for Humane Technology
By combining all these benefits; advances in ML algorithms, ability to collect both demographic and behavioral information, more dynamic interactions between members and products, and willingness of ImagineBC members, this creates a unique opportunity for developing a RELIABILITY scoring system. RELIABILITY scoring enables marketers to “ask the right person the right questions”. It provides a secure and rewarding environment for members to provide their honest opinions, cultivating the creation of a healthy ecosystem of reliability. In order to maintain the healthiness of this reliability, RELIABILITY scoring puts much emphasis on AI ethics to make sure everyone has a fair share of offering their opinions so that what gets reflected in the survey can be a reliable representation of the community.
After reading WeR.Ai’s approach to developing this new concept of a RELIABILITY score, I think you may agree with me that things have certainly changed since my father’s time. ImagineBC looks forward to creating a vibrant direct market between advertisers and our members. A market that has the potential to dramatically redistribute the 100s of billions of dollars of wealth currently retained by specific third-party intermediaries back to the very consumer the advertiser is attempting to reach.