Last week, we touched on consumer data collection that can be used to group users into potential customers and uninterested customers for ad targeting purposes. While some companies try to get as much data on consumers, whether ethically or not, others are more concerned with user privacy. As a result, they tend to use primarily user-provided data. With this data, companies can use algorithms to carry out all sorts of analytics to predict consumer behavior. One type of such analysis is binary predictive modeling. For example, whether a customer will donate a specific amount to a future re-targeting campaign or not. This is called supervised machine learning because the machine creates an algorithm based on the data given in order to group the data in categories pre-specified by the analyst.
This type of analysis helps both the company and the consumer. The goal of this type of targeting is to reduce the cost of communication with consumers while increasing the percentage of respondents by only sending a mail to those consumers that seem interested. In order to determine which consumers might be interested, analysts derive various types of variables on each of the consumers, for example, the length of the relationship and the engagement of the consumer, based on data that the consumer has provided. Then, they test these variables against a similar campaign in the past in order to determine which type of demographics responded to the campaign. In this way, a target demographic is created. This logistic regression model is then applied to the current user base to determine the likelihood that a given consumer will respond to the future campaign. Finally, the organization decides how much of the user base it wants to contact based on the model in order to make the highest profit.
This method of targeting is beneficial to both parties as companies get a higher return on investment due to the higher respondent rate and consumers get less promotional material for campaigns that they are unlikely to respond to. Today’s consumer engagement era is in stark contrast to the mass marketing era of the marketing evolution as companies no longer have to target massive amounts of consumers in order to get the return that they want. While consumers are right to try to protect their privacy, they need to also realize that this era of machine learning allows for more personalization of marketing material and reduces the number of irrelevant marketing materials. The important thing is to find a balance between data sharing and personalization.
Where have you noticed personalized ads? What were they marketing for?
If you’d like more information on the topic, feel free to explore a retargeting project in relation to this machine learning topic that I recently completed with some colleagues.