Retail Viewpoint: Combining Machine Learning with GenAI to Multiply Business Impact
The role of GenAI in the retail industry is rapidly expanding and is here to stay. While recent technological advancements like cryptocurrency and NFTs may not have met the hype, GenAI is already driving tangible business implications that are set to expand even further. OpenAI’s recent developer conference showcased exciting updates in this field.
Many retailers have already adopted GenAI for content creation and automating routine processes, particularly in call centers. However, the true potential lies in combining GenAI with machine learning (ML) in innovative ways. ML focuses on developing algorithms that can analyze complex data, identify patterns, and inform predictions and decisions. Generative AI, on the other hand, takes ML techniques and uses them to generate new and original content, such as text, audio, and video, based on the teachings of its training data.
By blending these technologies into existing processes, retailers can achieve a level of hyper-personalization enabled by micro-segmentation and sequential recommenders. This combination creates a virtuous cycle that drives continuous improvement, resulting in a multiplied impact. When implemented within a broader marketing effectiveness program, these capabilities can augment customer count, basket size, total trips, and ultimately, generate a substantial increase in revenue.
One real-life example of this combined approach is a collaboration with a nationwide specialty retailer to develop custom marketing solutions using ML and GenAI. The goal was to enhance the effectiveness of promotional emails and provide customers with a personalized experience. The results were outstanding, with open and click rates increasing by 30-45%. The company directly linked the success of the campaign to revenue improvements.
The application potential of this combined approach is vast. To illustrate, let’s consider a common marketing challenge: generating additional revenue with limited resources.
Sales and marketing teams often face a long list of priorities with limited touchpoints and resources. Maximizing overall efficiency requires prioritizing the right customers, and AI advancements make this task much easier.
Customer lifetime value (CLV) can be a valuable metric in identifying high-potential customers. ML models can estimate CLV by forecasting customer churn and future spending patterns based on their past purchase experiences. Developing proprietary CLV models can assist marketing teams in prospecting, but it’s essential to define any potential constraints that may limit the effectiveness of these models.
ML algorithms can also help build response propensity models that identify high-CLV customers most likely to make additional purchases after contact. Beyond purchase intent, these models also reveal the preferred products and the segments that are more sensitive to promotions.
To fine-tune the strategy, it’s crucial to test the model’s quality and customer response. A company embracing this approach generated 40% more revenue per targeted customer and doubled the average order value by contacting high-CLV customers identified by its ML models compared to traditional targets.
However, knowing who to contact is only half of the challenge. This is where GenAI comes into play, creating and delivering personalized offers and messages that motivate customers. Unlike traditional recommendation systems, GenAI’s product recommendation engines generate content and suggestions based on each individual customer’s preferences and previous interactions. This level of personalization is a significant advantage for marketers and sales teams looking to engage customers along their personal journeys.
By utilizing the ML output, it’s possible to develop micro-segments that allow marketing teams to tailor messages to specific sub-groups. These micro-segments can then be fed into GenAI algorithms to create different subject lines and targeted messages. Testing all iterations helps uncover what resonates best.
When implementing this approach, ensuring alignment between teams is crucial. Smooth transitions from ML models to micro-segments to testing are dependent on transferring key data and understanding the why behind each step. A strong focus on quality control guarantees that all models rely on a shared understanding.
This approach enables marketers to optimize their most impactful messages quickly. Reinforcement learning has long been a challenge, as humans can only create a limited number of iterations and struggle to comprehend the reasons behind a message’s success. However, with ML, knowledge gained from testing can be utilized to take the top-performing AI-crafted messages, iterate on them, and continue testing and improving.
This creates a virtuous loop that gradually enhances message effectiveness over time. ML models, providing high-value CLV customers, continually feed pinpoint target segments into GenAI algorithms, resulting in increasingly effective messages tailored to individual customers. This is the holy grail for sales and marketing teams.
Looking ahead, there are countless other possibilities for applying GenAI and ML to various processes. For example, ML models can match customers to sales associates who best understand their goals, preferences, and personalities. Advanced marketing mix models and strategies can enhance the ratio of CLV-to-CAC (customer acquisition cost). Discount elasticity models can maximize margins by applying discounts only where and when they are most effective. The potential benefits are immense.
While GenAI and ML have strong use cases independently, their true power lies in their combination. Effectively blending these tools is a complex process that may require strengthening operational capabilities and boosting AI and ML talent. However, the benefits far outweigh the challenges.
By integrating these two technologies into existing processes, retailers can exponentially multiply their impact, often achieving results sooner and to a greater extent than expected. Now is the time to take advantage of these tools and witness the cumulative effect they can have.
