Retailer Achieves 15% Sales Uplift with Advanced Assortment Optimization Analytics

Misaligned product assortments cost retailers over $1 trillion annually in lost sales and excess inventory. A leading multi-category retailer faced this exact crisis, with shrinking margins and escalating carrying costs due to a disconnect between their product mix and local consumer demand. Their traditional, gut-feel-based merchandising could not keep pace with shifting preferences, leading to both stockouts on high-demand items and overstock of slow-moving products. This case study details how the application of granular assortment optimization analytics provided the data-driven clarity needed to overhaul their strategy. The engagement transformed their shelf space from a liability into a high-performing asset, culminating in a remarkable 15.3% increase in category sales and a significant reduction in operational waste.

Key Highlights

• Client Overview: A Retail Giant at a Crossroads

  A multi-billion-dollar retailer with over 1,000 stores across North America was struggling with declining profitability. Despite their market leadership, their one-size-fits-all assortment strategy was failing. Different regions and store formats exhibited unique demand patterns that their centralized planning process ignored. The primary objective was to transition from this outdated model to a localized, data-centric approach. They needed an assortment optimization framework that could tailor the product mix for specific store clusters, thereby maximizing sales opportunities, improving inventory turnover, and enhancing the overall customer experience by ensuring product relevance at the local level.

• The Challenge: Crippling Inefficiencies from Poor Assortment

  The core challenge was a profound lack of analytical capability to manage SKU complexity. The retailer was grappling with rampant product cannibalization, where new product introductions were simply stealing sales from existing ones without growing the overall category. Inaccurate demand forecasting led to a persistent cycle of stockouts on popular items, resulting in lost sales and customer frustration, and excessive overstock on others, which inflated holding costs. This operational chaos was a direct result of relying on historical sales data without the tools to predict future demand or understand the nuanced interplay between products on the shelf.

• Solution: Advanced Analytics for Merchandise Planning

  Quantzig deployed a comprehensive assortment optimization solution centered on predictive analytics and machine learning. The approach involved harmonizing disparate data sources—including POS data, inventory levels, promotional calendars, and local demographic data. We developed sophisticated demand forecasting models and a cannibalization analysis engine to predict the true sales potential of each SKU. This enabled the creation of an interactive assortment planning tool that allowed category managers to simulate the financial impact of different product mixes, optimizing for sales, margin, and inventory constraints before implementation.

• Impact: From Reactive to Predictive Merchandising

  Delivered a staggering 15.3% increase in sales for pilot categories and reduced inventory holding costs by 24%. The solution provided a clear, data-backed path to profitability. Forecast accuracy improved from a mere 65% to 88%, drastically cutting down on stockouts and overstock situations. More strategically, the engagement empowered the client's merchandising teams to shift from reactive, spreadsheet-based planning to proactive, strategic decision-making. They could now confidently tailor assortments to local tastes, leading to a 4% improvement in overall profit margins and a stronger competitive position.

Problem Statement

A premier North American retailer, despite its significant market presence, found itself navigating a complex and costly operational environment. The central issue stemmed from an outdated assortment strategy that failed to recognize and adapt to regional and local market dynamics. This resulted in a cascade of problems: critical stockouts of high-velocity items in certain regions while the same products gathered dust in others. The financial ramifications were severe, with an estimated 18% stockout rate on key products leading to millions in lost revenue and a tangible decline in customer loyalty. Furthermore, the inability to perform SKU rationalization effectively led to bloated inventories, with holding costs spiraling upwards by 30% year-over-year. The core of the problem was a fundamental gap in data visibility and analytical maturity. The retailer's merchandising decisions were driven by historical sales reports and intuition, lacking any predictive or prescriptive analytics. They were unable to answer critical questions: Which products are true substitutes? What is the sales lift from a new product introduction versus the cannibalization it causes? How should the assortment in a dense urban store differ from a suburban one? This lack of insight rendered their entire merchandise planning process inefficient and reactive.

• Inaccurate Demand Forecasting : The client relied on simple moving averages and historical sales, which failed to account for seasonality, promotions, or local trends. This resulted in forecasts that were, on average, only 65% accurate. This inaccuracy was the primary driver of both stockouts and overstocks, creating a constant state of inventory imbalance across the store network and undermining financial planning.
• Rampant Product Cannibalization : Without a method to measure the impact of new products on existing ones, the client's category managers inadvertently saturated the assortment with redundant items. New SKUs often just redistributed sales rather than generating incremental growth. This lack of a robust product assortment strategy led to increased complexity in the supply chain and confused customers without adding to the bottom line.
• Ineffective SKU Rationalization : The process for delisting underperforming products was subjective and inconsistent. This led to a bloated catalog of over 100,000 SKUs, many of which contributed less than 5% of category sales while consuming valuable shelf space, capital, and logistical resources. The inability to systematically prune the long tail of the assortment was a major drag on profitability.
• Lack of Localized Assortments : A one-size-fits-all approach was applied to all stores, regardless of local demographics, climate, or competitive landscape. An assortment that worked in a Miami store was replicated in a Seattle store, leading to missed sales opportunities and irrelevant product offerings. This failure to tailor the product mix was a key factor in losing market share to more agile, locally-attuned competitors.

The breaking point arrived during the post-mortem of the crucial back-to-school season. The analysis revealed a catastrophic failure: while the company's national marketing campaign for new backpacks was a success in terms of driving foot traffic, a staggering 40% of stores had stocked out of the advertised hero product within the first week. Simultaneously, warehouses were overflowing with last season's models. The financial impact was a double-edged sword—millions in lost potential sales from the stockouts, compounded by the steep discounts required to liquidate the old inventory. It was a stark, quantifiable demonstration that their current approach to assortment planning was not just inefficient; it was actively destroying value. The executive team realized they were flying blind, making multi-million-dollar inventory bets with flawed data and no analytical rigor. The status quo was no longer a viable option; they needed a fundamental shift toward a data-driven assortment optimization capability.

Objectives

To address these systemic challenges, Quantzig and the client co-defined a set of clear, measurable objectives for the engagement. The overarching goal was to transform the retailer’s merchandising function from a cost center into a strategic driver of growth and profitability through advanced analytics.

• Enhance Forecast Accuracy : The primary objective was to increase demand forecast accuracy from 65% to over 85%. Achieving this would directly address the core drivers of inventory inefficiency, enabling the client to significantly reduce both stockouts and overstock situations. This improvement in analytics capability would form the foundation for a more reliable and responsive supply chain.
• Enable Localized Assortments : A key goal was to develop a scalable framework for creating store-level assortments. This involved creating data-driven store clusters based on sales patterns, customer demographics, and other local factors. Success would mean moving beyond a monolithic national plan to a portfolio of tailored assortments that maximized relevance and sales potential for each unique cluster.
• Quantify Cannibalization Effects : The client needed to move beyond guessing the impact of new products. A critical objective was to build an analytical model to quantify product cannibalization and accurately predict the incremental sales lift of any new SKU. This would empower category managers to make informed decisions about product introductions and delistings, ensuring a healthier, more productive assortment.
• Improve Category Profitability : Ultimately, the engagement had to deliver a tangible financial return. The final objective was to increase overall category gross margin by at least 3 percentage points. This would be achieved by optimizing the mix of high- and low-margin products, reducing inventory costs, and maximizing sales from the most productive shelf space, directly linking analytics to bottom-line performance.

Solution Implemented

Quantzig's solution was an analytics-driven consulting engagement designed to embed a data-first culture within the client's merchandising team. We developed a holistic assortment optimization framework that combined advanced statistical modeling with a user-friendly decision support system. Our methodology focused on transforming raw data into actionable assortment strategies, moving the client from reactive reporting to predictive and prescriptive planning. The final deliverable was a comprehensive report and a set of dynamic analytical tools that empowered category managers to build and validate optimal assortments for every store cluster.

• Data Harmonization and Enrichment : Integrated and cleaned POS, inventory, and product attribute data from multiple legacy systems.
• Store and Customer Segmentation : Utilized clustering algorithms to group stores with similar sales patterns and customer profiles.
• Predictive Demand Modeling : Built machine learning models to forecast SKU-level demand, incorporating seasonality and promotions.
• Assortment Optimization Engine : Developed an engine to identify the ideal product mix based on constraints like shelf space and budget.
• What-If Scenario Analysis Tool : Delivered an interactive dashboard for simulating the revenue and margin impact of assortment changes.

Technologies Used

• Data Processing and Warehousing with Python and SQL : We leveraged Python's Pandas library for large-scale data manipulation and cleansing of terabytes of historical sales data. SQL was used to query and join data from the client's disparate databases, creating a unified analytical data mart. This foundational layer was crucial for ensuring data quality and consistency, enabling all subsequent modeling efforts. It allowed us to create a 'single source of truth' for all merchandising analytics, eliminating data silos that previously hampered decision-making.
• Predictive Modeling with Scikit-learn and XGBoost : For demand forecasting, we employed a combination of time-series models (like ARIMA) and machine learning models. XGBoost was particularly effective in capturing complex relationships between sales and external factors like promotions, holidays, and local events. For the cannibalization analysis, we used regression models within Scikit-learn to quantify the substitution effects between products, providing a clear metric for the net impact of adding a new SKU to the assortment.
• Optimization using SciPy and PuLP : The core of the assortment optimization engine was built using linear and integer programming techniques. We utilized Python libraries like SciPy and PuLP to formulate the optimization problem: maximize total margin subject to constraints such as limited shelf space, inventory budget, and category-specific rules (e.g., must carry certain 'destination' items). This allowed us to generate mathematically optimal assortment recommendations, a significant leap from manual, trial-and-error methods.
• Visualization and Reporting in Power BI : While the heavy lifting was done in Python, the final insights and tools were delivered via an interactive Power BI dashboard. This served as the primary interface for category managers. It allowed them to visualize sales trends, explore store clusters, and most importantly, use the 'What-If Scenario Analysis' tool. They could drag and drop products into a virtual assortment, and the dashboard would, in real-time, display the projected impact on sales, margin, and cannibalization, making complex analytics accessible and actionable.

Request a demo

Results and Impact

The implementation of Quantzig's assortment optimization framework delivered transformative results, fundamentally reshaping the client's merchandising operations and financial performance. By replacing intuition-based decisions with a rigorous, data-driven methodology, the retailer was able to achieve a level of precision and efficiency that was previously unattainable. The impact was felt across the organization, from the supply chain to the store shelf, and ultimately, on the bottom line. Our analytics solutions provided not just a one-time lift but an enduring capability for continuous improvement, enabling the client to definitively resolve their core challenges of inventory mismanagement and stagnant growth. The engagement established a new benchmark for performance and created a clear competitive advantage.

Qualitative Impact

• Operational Transformation: Data-Driven Daily Decisions : The most immediate impact was on the daily workflow of the category management team. The manual, time-consuming process of analyzing hundreds of spreadsheets was replaced by interacting with the Power BI dashboard. Weekly assortment review meetings, which were previously contentious and subjective, became productive, data-driven sessions. Instead of debating opinions, the team now analyzes scenarios generated by the optimization engine. This has freed up hundreds of hours per month, allowing experienced merchants to focus on strategic vendor negotiations and promotional planning rather than manual data crunching. The result is a faster, smarter, and more agile merchandising operation.
• Strategic Enablement: Confident Market Expansion and Localization : Strategically, the assortment optimization framework unlocked new possibilities for growth. The retailer is now able to model the potential of a new store opening with a high degree of accuracy, using the store clustering model to predict which assortment will perform best. This has de-risked their expansion strategy. Furthermore, decisions that were once impossible, such as determining the optimal balance between national brands and private labels for a specific neighborhood, are now standard practice. The analytics provide the confidence to make bold strategic moves, backed by a clear understanding of the likely financial outcomes.
• Cultural Shift: Fostering Trust in Data : Perhaps the most profound change was cultural. Initially, there was skepticism from veteran merchants who had built careers on their 'gut feel.' However, after the pilot program's results conclusively outperformed the control group's traditional methods, a shift occurred. The analytics tools were no longer seen as a 'black box' or a threat, but as a powerful aid that augmented their own expertise. This success fostered a newfound trust in data across the organization. It sparked a broader demand for analytics in other departments, leading to a company-wide initiative to build a center of excellence for data science and analytics.
• Future Trajectory: Building on a Foundation of Analytics : With a robust assortment optimization capability in place, the client is now positioned for the next phase of retail innovation. The rich demand forecasting models are being repurposed to inform dynamic pricing and promotion strategies. The insights from the market basket analysis are being used to redesign store layouts and planograms to increase transaction size. The engagement has not just solved a problem; it has provided a foundational analytics platform upon which the client can build future competitive advantages, ensuring they remain a leader in the evolving retail landscape.

How Quantzig Can Help

Quantzig's success in this retail assortment optimization engagement is a direct reflection of our deep-seated expertise in the retail analytics domain. With nearly two decades of experience partnering with global retailers, we have cultivated a profound understanding of the intricate challenges and opportunities within the sector. Our approach is not merely technical; it is a synthesis of advanced data science, business acumen, and a practical understanding of retail operations. We don't just build models; we build solutions that integrate seamlessly into business workflows and drive measurable financial impact. This case study exemplifies our ability to dissect a complex business problem, like SKU rationalization and localized assortment planning, and apply a tailored analytical framework to it. Our team of retail experts, data scientists, and analytics consultants worked in concert to translate vast, messy data into a clear, strategic roadmap for profitability. The 15.3% sales uplift and 24% reduction in inventory costs are not just numbers; they are the outcome of a meticulously executed strategy grounded in years of focused experience. Quantzig's capability lies in turning analytical potential into tangible performance, empowering our clients to navigate market complexities with data-driven confidence and precision.

Our Expertise in Retail Analytics

• Demand and Sales Forecasting : We specialize in developing sophisticated forecasting models that go beyond historical data, incorporating dozens of variables to predict demand with unparalleled accuracy, forming the bedrock of any robust retail strategy.
• Merchandising and Assortment Analytics : Our core competency lies in transforming merchandising from an art to a science. We provide analytics solutions for assortment optimization, SKU rationalization, and category management that directly enhance profitability.
• Supply Chain and Inventory Management : We help clients optimize their supply chains by using analytics to improve inventory allocation, reduce holding costs, and minimize stockouts, ensuring product availability at the lowest possible cost.