Innovative strategies are crucial for maintaining a competitive edge, particularly in user engagement and activation. This post outlines our integration of Artificial Intelligence (AI) and Machine Learning (ML) to significantly impact user experience and business outcomes for Mashkor. Our primary focus was on optimizing the experience for both first-time and existing users.
Need
Our journey was driven by a clear goal: to elevate activation rates for first-time users while simultaneously enhancing engagement for our existing users of the application. The challenge extended beyond introducing users to our platform; we aimed to ensure they discovered immediate value, encouraging ongoing engagement and improve a deeper connection with our services. To meet these objectives, we turned to the capabilities of ML and AI, focussing to create a seamless, efficient, and valuable user journey from their initial interaction.
Initial Solutions
Our strategy was executed in phases, acknowledging the intricate nature of user behavior and the diverse spectrum of their requirements.
Phase One: Personalized Recommendations for First-Time and Existing Users
The first phase of our initiative involved harnessing ML to tailor recommendations for both first-time and existing users, using a variety of input parameters to fine-tune these suggestions.
For new users, we focused on their search terms and category interactions as a primary indicator of their immediate needs and interests. This allowed us to present a curated list of the most relevant stores, aiming to improve CTRs and, consequently, conversion rates. Google Vertex AI’s recommendation model was our tool of choice for this task, selected for its algorithms, compatibility with our tech stack and ability to scale efficiently.
In addition to addressing the needs of new users, we also refined our approach for existing users by analyzing their past behaviors. This analysis and inputs included various parameters that enriched the personalization of their recommendations. Given their history with our platform, we anticipated that the model would yield faster and more accurate results for these users, enhancing their overall experience and satisfaction.
We just initially decided to focus on batch recommendations and later evolve to real time recommendations.
Phase Two: Refining the Model
The second phase focused on refining our model to better achieve our goals of increasing activation for new users and bolstering engagement for returning ones. This stage involved iterative adjustments and enhancements, driven by continuous feedback and performance analysis. We also employed Generative AI in certain scenarios to create compelling copy, further personalizing the user experience.
Challenges
Adopting to ML came with its set of challenges. Initially, the limited amount of available data constrained our models’ predictive accuracy. Moreover, the success of these AI-driven solutions heavily depends on the quality of data, emphasizing the importance of sophisticated data collection and management strategies.
Benefits
Over time, the benefits of our AI integration became increasingly evident. For first-time users, the AI-powered recommendations facilitated a smoother discovery process, significantly improving their initial engagement with our platform. Existing users enjoyed enhanced personalization through the “Recommended for You” feature, which evolved to more accurately reflect their preferences and behavior patterns. These experiments helped us understand the impact of AI and ML on creating a user-centric, personalized experience. Integrating into our product was challenging, fun and rewarding.
Pavan's Blog 