Imagine this: an online bookstore aims to deliver more personalized recommendations. Based on internal customer data–collected from registrations, purchases and other direct interactions—the bookstore knows it has 1,000 customers who are between 25 and 35 years old, female, and readers of science fiction novels. Relying only on this data, the most that can be done for this group is personalization based on age, gender, and reading preferences.
But what if the bookstore knew that 100 people in this group had recently spent several weekends on mountain bike trails? What if it knew that 300 had liked social media posts related to a musician that had just published a biography? The bookstore now has many new options available in how it tailors its products and services for these customers—but it needs the power of AI to hyper personalize for these micro audiences.
AI can use these new external data sources—collected, anonymized, and obtained with consent from sources such as mobile devices, social media, and geolocation services—to generate higher-value signals and precise insights. It can do this thanks to the power of machine learning, which can identify the degree of association between enormous collections of different data types and formats, helping us see the important patterns across, and connections between, all the features in the data.
Intelligent hyper personalization
Using more of this contextual or behavioral data to train AI models, the bookstore—or any business—can anticipate and deliver on customer interests for very small segments. As a complement to, rather than substitute for, traditional forms of customer data, these new data sources can help organizations become more sophisticated in their personalization tactics, leading to more engaged customers and better business results. Indeed, our research has found that two-thirds of consumers are more likely to shop at a retailer in-store or online if the retailer remembers their previous purchases.
It’s not only with retail that new data sources for AI models can lead to better outcomes. For example, we worked with a company in the financial services sector that wanted to use new data sources to acquire new customers rather than buying leads from an aggregator or pre-defined audiences from a demand-side platform (DSP), which is costly and crowded. In order to do so, we first identified the underlying need for why people would want to borrow money—for example home renovations, university tuition, or buying a car.
We then identified prospects with the above needs by analyzing external data, such as geolocation data on visits to car dealerships or browsing data from university websites. By applying AI to external data sources, we were able to narrow down the pool of potential customers. This made it possible for the organization to reach the right segment at the right time with the right offer for a consumer loan very early in the consumers’ decision-making process. It also enabled the business to cut its cost of new customer acquisition by almost 50 percent.
Using AI on internal data coupled with customer data from third party vendors is only the tip of the iceberg when it comes to what organizations can do to boost their business. We can help increase the value obtained from AI significantly by using internal and external ‘new’ data. Using AI on these new data sources can drive multiple business use cases such as increasing cross-sell and up-sell, increasing customer retention, enhancing customer journeys and experience, and optimizing marketing and media spend.
Preparing to expand the scope of your data analysis
Before you can develop actionable insights from these new data sources, there are three key considerations you need to make.
- Identify your sources. First, identify what new sources of relevant data are available to you. There are many types of data available, including data on what other shops your customers visit, what mobile apps they use, whether their credit rating is improving, and how the weather impacts their shopping behavior. The first step is to determine what is available and begin to explore which sources seem most likely to offer value and quality data. Organizations can access new data sources via partners, so there is no need to establish the AI data collection capability in-house.
- Allay fears. This needs to be accomplished while maintaining customer trust around data privacy and security. According to a recent Gartner survey, 65 percent of consumers believe that AI will harm, rather than improve, their privacy. To allay fears, customers and prospects need to be reassured that this level of personalization is based on anonymous categorization rather than their individual data. Companies enable segment and non-individual-level actions on select digital channels only to improve personalization in marketing and customer experience. Detailed legal and compliance reviews should also be in place to help companies try to diminish bias and stop the use of data that leads to exclusion or negative actions.
- Be Agile and Test for ROI. Agile build and practical testing must be rolled out that ensures return on investment (ROI) is measured effectively. Not only should businesses establish a wide range of ROI criteria, they need to regularly do A/B testing to identify AI strategies that are proven to work and offer clear and transparent routes to ROI. This will secure continued investment in their AI strategy now and in the future, while strengthening the company’s position within a fast-moving landscape.
With the right sources, governance and systematic approach, AI can isolate fascinating and valuable patterns between your internal data and the ever-increasing number of new datasets. It is a way to provide evidence-based answers to important strategic questions, as well as predictions that could add significant business value. As such, you should not limit your projects to internal data, because creative use of new datasets is now one of the most powerful ways to increase the value of your AI investments.