In a rapidly-moving world, targeted marketing campaigns need to work with customer sentiment in real-time. Marketers need to work with incoming data and adjust their messages accordingly. Data science and machine learning can help marketers make more targeted decisions.
“From building and deploying AI models in near real-time frameworks, to doing something as complex as A/B testing with a number of different AI models, there are a number of ways to have feedback based on the successes and failures of specific modeling campaigns.” Dr. Kenneth Sanford, U.S. lead analytics architect, Dataiku, said.
The thing about testing is that there must be room for (controlled) failure. Sanford explained the basis of an A/B test is setting out a choice, including “something that probably won’t work.”
You must allow for making mistakes, albeit in a very guarded way, to discover which direction is the better one to follow. That discovery is expedited by using AI models or prediction models.
Related: We Think You Think You Will Buy That
One example of A/B testing would be launching two highly-personalized model ads, which are close but have a slight difference. As a result of very targeted predictive models, it is possible to have very similar options presented to the same person, Sanford said. The goal would be to see which ad leads to the intended outcome, whether it is an actual purchase, brand awareness, website visits, or engagement on social media accounts.
Behind fine-tuning ads to achieve this high level of personalization is data — lots of data. First, there is first-party behavioral data tracked through the brand’s site.
“Every individual has unique footprint that is identifiable from the way they click through the website,” Sanford explained. That behavioral pattern can be used as the basis of predictive models.
Partnering for more data
For marketers who want even more data to build a complete view of a customer, increasing insight into individual customers fall on third-party data partnerships.
Companies like , for example, form partnerships with digital-based businesses like Lyft and Airbnb, to gain a much more comprehensive view of customer movements and spending habits, to form that “360-degree view,” Sanford explained.
“Portals to other sellers incentivized by data sharing and access to new markets,” Sanford said.
Sometimes partnerships appear to make strange bedfellows, as in the case of . Walmart’s positioning offers not just the advantage of increased traffic, but of valuable consumer insight.
“Walmart can micro-segment and target in ways very similar to Amazon,” Sanford said.
Amazon, of course, is the ultimate model of building on customer data to drive personalized marketing even with new customers. According to Sanford, the advantage of working of data models is that you can synthesize and apply to the general population.
With enough data to form a good sample model, you can target individuals even without their personal data.