How to Win with AI and Machine Learning in Retail
Artificial Intelligence (AI) and Machine Learning (ML) are increasingly important tools to connect with consumers seamlessly as they traverse channels along their shopper journeys. Break the vicious cycle of launching and abandoning ideas by using them to learn, iterate and ultimately identify high-performing retail solutions based on smart analytics, not just standout creative.
AI and Machine Learning in retail are rapidly evolving. AI is the broad concept that machines can carry out tasks in a way we consider “smart.” Turn to AI to solve concrete issues with procurement or logistics or more abstract ones, like defining which channel, creative, message and offer/pricing will most effectively reach specific consumers or consumer segments. Answering the right questions is a key imperative of delivering success.
Machine Learning is an application of AI that gives machines data and lets them learn for themselves. Supervised ML is data with a clearly defined output that can be used to predict outcomes. Unsupervised ML automatically recognizes and understands data patterns and structures, but an outcome is not given.
The benefits and impact of AI and Machine Learning include:
- Better understanding consumer motivations
- Creating new opportunities for product discovery
- Hyper-personalization and unique 1:1 consumer experience
- Real-time decision-making that drives predictive behavior
Finding insights from unstructured data
Increasingly sophisticated algorithms can improve the shopper experiences with insights into everything from store layouts to turbo-charged personalization to media channel orchestration. Affordable elastic computing makes it easy to expand or decrease the computer processing, memory, and storage needed to test ideas.
Industry estimates say 80 percent of the data we use to understand consumers from the newest media and content channels is unstructured, coming from streaming audio and video, social media, clickstream, as well as sensor and log reports. The challenge with unstructured data is that it’s not arranged or delivered in a pre-set data model or schema, therefore making it incompatible with a traditional relational database.
AI and Machine Learning can efficiently clean and deliver this critical information. You can then use it to balance breakthrough creative ideas with data-driven strategies that optimize results. You can more effectively reach and influence any shopper, anytime, anywhere and understand the why and who behind every buy. Most importantly, you can measure real-time evolving shopper behavior – everywhere they shop - versus relying on models.
What problem do you want to solve?
Think of AI and Machine Learning as the superglue between data analytics, marketing and technology. Use it to align on goals and metrics, balance quick wins with fundamental process redesign, and hire the right talent and re-train existing employees to meet your needs.
Start by clearly defining the problems you want to solve. Ask yourself, “What do I want to unlock with data? What patterns do I want to see?” With that settled, you can then determine the scale of data input and be more precise in your discovery.
AI and Machine Learning in retail is evolving into four types of data:
Descriptive – Explains what happened. Catalina uses business intelligence reports to provide a better understanding of what happened with a particular audience or campaign. Post-Campaign Reports
Diagnostic – Explains why it happened. Catalina uses measurement, multi-touch attribution and consumer-mix modeling tools to proactively track in-market impact of programs and identify ways to optimize measurement input and results in real-time. In-Flight Measurement
Predictive – Forecasts what might happen. Catalina uses contextual targeting based on site, time of day and device to identify what is most likely to lead to incremental sales. Its new product recommender model constructs shopper and offer scores based on sales lift and its new Shapley Value attribution model looks at the relative importance of predictor variables. Predictive Audiences
Prescriptive – What should happen. Creates recommendations based on forecasts that incorporate in-store results. For example, Catalina can advise on media buys, such as optimal bidding for digital ads. Media & Activation
The rapid evolution of AI and Machine Learning has been the driving force behind Catalina’s new Digital Circular Personalizer (DigitalCircP), which uses AI and Machine Learning to leverage targeted media across mobile and in-app platforms to deliver personalized recommendations to shoppers.
DigitalCircP reaches shoppers with personalized weekly circular ads on their mobile phones in the apps and websites they frequent. It gives retailers greater reach and frequency with low-cost media. Catalina data analysts use deterministic data (known shoppers) with competitive conquesting (unknown shoppers) to deliver 1:1 personalized ad units in real-time by linking digital IDs to frequent shopper cards across channels. On average, participating retailers have seen a $20:1 return on ad spend (ROAS) and a 2% sales lift in measured store sales.
Learn more about Catalina’s AI and Machine Learning-driven solutions by visiting www.catalina.com/solutions.