Retailers Turn to Predictive AI Models to Deliver Accurate Delivery Promises at Checkout
As online shopping becomes second nature, customer tolerance for vague or missed delivery dates is wearing thin. A broken delivery promise now costs more than just a lost sale; it erodes trust. Today’s shoppers expect fast, reliable, and transparent delivery timelines, and retailers are scrambling to meet these demands without breaking the bank. The growing complexity of global supply chains, coupled with unpredictable disruptions such as weather, labor shortages, and logistic delays, has made accurate delivery forecasting one of the retail industry’s most urgent challenges.
That’s where predictive AI models are stepping in, and few know this space better than Arulmozhi Kasthurirengan. A key contributor in the Delivery & Fulfillment Channel of a leading supply chain network, she has played a critical role in building AI-driven systems that optimize delivery forecasts in real-time. Her work doesn’t just solve a technical puzzle; it changes how retail organizations operate, cutting costs and elevating customer experience in equal measure.
When Arulmozhi first entered the retail IT space, she quickly noticed a recurring issue: vague or inconsistent delivery messaging was causing customers to abandon carts and flood customer support lines. “Shipping is incredibly complex. You have so many moving parts: carriers, distributors, manufacturers, and the customer. When one link falters, the whole chain feels it,” she explained.
Instead of patching over the problem, she dived into the core logic of predictive modeling using AI/ML. She studied the intricate algorithms that powered the systems and helped lay the foundation of a predictive model now used across her organization. Her grasp of the algorithm’s depth allowed her to align cross-functional teams and engineer a data-driven model responsive to real-world challenges.
The model’s rollout delivered measurable impact almost immediately. By dynamically detecting patterns and adjusting to shifts in real-time, the model introduced flexibility through built-in buffers and continuous learning. “After implementation, we saw our Add to Bag and Bag to Order Rate rise by 0.7%, while customer service calls dropped by 35%. That’s a significant improvement in both user engagement and operational efficiency,” she said.
But these numbers tell only part of the story. Internally, the model also cut costs associated with upgrades and late deliveries, freeing up resources that could be reinvested in customer experience and innovation.
To ensure these insights weren’t locked away in backend systems, Arulmozhi developed a KPI dashboard that brings clarity to the chaos. “It gives daily and weekly visibility into how predicted and actual delivery days compare across brands, fulfillment types, and shipping methods,” she said. The dashboard didn’t just improve monitoring; it shaped smarter decision-making across departments.
It is this transparency commitment that differentiates Arulmozhi. She espouses democratizing data inside the organization so that teams throughout logistics, operations, and customer service can agree on what’s going on the ground.
Among the most significant advances were those that addressed problems long thought insoluble. Mismatches in delivery date as a result of differences between calendar and business days were one of them. She led efforts to sync these variables by collaborating with delivery partners and adjusting for contract limitations like the inability to deliver on weekends.
She also proposed a novel probabilistic model to show customers a delivery window instead of a fixed date, increasing accuracy and managing expectations. “We worked closely with carriers to bring in real-time feed and developed what we now call Conditional PDD (Promised Delivery Date), which adjusts based on the mode of delivery,” she added.
Two of the most defining projects in her journey—Returns and Registry—gave Arulmozhi an opportunity to embed these models into customer-facing systems. Each initiative required navigating technical and operational hurdles, but the results strengthened the backbone of her organization’s supply chain.
By focusing on delivery messaging and aligning that across all points in the shopper journey, Arulmozhi’s work reduced checkout friction and restored trust at a critical touchpoint: the checkout page.
With an eye on what’s next, she remains grounded about both the potential and the limitations of predictive AI in retail. “AI/ML will continue to dominate the future of delivery systems. But we need to stay vigilant about model drift, ensure transparency in predictions, and never lose sight of customer experience,” she warned.
She emphasizes the importance of continuously training models with fresh data and being prepared for external shocks—weather, holidays, policy changes—that can upend even the smartest systems. “Accuracy is not a one-time win. It’s a moving target, and you need the infrastructure and talent to keep hitting it.”
For Arulmozhi Kasthurirengan, the work is far from done. But her message to retailers is clear: don’t just promise better delivery, predict it, track it, and own it. The technology is ready. The question is, are you?