📈 economics

From Data-Driven Personalization to Culturally Adaptive Merchandising: A Systematic Review and the CAMI-DM Framework

This paper addresses the gap in culturally sensitive personalization by conducting a systematic review of existing literature and proposing the CAMI-DM framework, an integrative architecture that combines behavioral data with cultural context and spatial optimization to enable adaptive, culturally aware merchandising across physical, digital, and phygital retail environments.

Original authors: CLAUDE ISSA NOMBRE, KOUADIO PROSPER KIMOU, TANON KODJO

Published 2026-10-21

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: CLAUDE ISSA NOMBRE, KOUADIO PROSPER KIMOU, TANON KODJO

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the modern world of shopping, retailers have become masters of tracking what we do. They know what we click, what we linger over, what we put in our carts, and what we leave behind. Using this stream of information, they build systems designed to guess our next move, offering us products we are likely to want before we even know we want them. This is the promise of data-driven personalization: a store that knows you better than you know yourself. However, there is a quiet flaw in this approach. A system that only watches what you do might miss who you are. It might not understand that a color you love in one neighborhood is offensive in another, or that a product arrangement that feels efficient to one group feels chaotic to another. Culture—the shared values, beliefs, and habits that shape how people see the world—is often treated as a simple label, like a country code, rather than a living, breathing context that changes how a message is received.

This gap between knowing what a customer does and understanding the cultural meaning behind those actions is the focus of a new study by researchers from the National Polytechnic Institute in Côte d'Ivoire. They asked a fundamental question: can we build a retail system that doesn't just react to data, but adapts to the cultural context of the person holding the data? To find out, they did not simply invent a new gadget or claim a new discovery. Instead, they looked at the entire landscape of recent research to see what was already known and where the pieces were missing. They gathered 129 scientific articles from a major database, carefully reading titles and abstracts to find the most relevant work. After a rigorous process of checking and re-checking, they narrowed this down to a final set of 10 studies that truly connected artificial intelligence, retail behavior, and cultural context.

What they found was a collection of brilliant, isolated islands. Some researchers had shown how artificial intelligence could predict demand for modest fashion using big data. Others had demonstrated that the background color of a product image changes how people from different cultures perceive it. Some studies proved that chatbots work differently for customers in Brazil than for those in England. Each study offered a powerful piece of the puzzle, but none of them had put the pieces together into a single, working machine. The researchers realized that while we have the tools to track behavior and the tools to understand culture, we lack a unified system that uses both at the same time to make real-time decisions about what products to show, where to place them, and how to present them.

To solve this, the team proposed a new framework called CAMI-DM. Think of it as a blueprint for a smarter kind of store manager. This system does not just look at a customer's history; it constantly updates its understanding of the situation. It takes in signals about what a customer is doing, but it also layers in a rich, multidimensional map of their cultural context—things like their values, their language, their religious practices, and their local traditions. It then combines this with the immediate situation, such as the time of day, the stock available, or the weather. The system uses this combined picture to make a decision. It might rearrange a shelf, change the order of products on a screen, or adjust the language of a sign, not just because a customer bought something similar before, but because that specific arrangement makes sense for who they are right now.

The researchers also addressed the physical side of this problem by proposing a specific mathematical layer to solve the spatial puzzle. They suggested using a Quadratic Assignment Problem (QAP) to determine where products should be placed. This logic ensures that products are arranged in a way that minimizes the effort a customer has to make to find them, while also respecting cultural preferences about how things should be grouped. For example, it might ensure that items used together in a specific cultural ritual are placed close to each other, while keeping items that might be considered inappropriate neighbors far apart. This is not just about guessing what a customer wants; it is about designing the environment so that the right choices feel natural and respectful.

Crucially, the study emphasizes that this system must be built with trust and ethics in mind. The researchers argue that adapting to culture should not mean stereotyping people or invading their privacy. The system should be transparent, allowing customers to understand why they are seeing certain recommendations, and it must always give them the power to ignore or override those suggestions. The goal is to create a shopping experience that feels helpful and relevant, not controlling or intrusive. The team acknowledges that this framework is currently a proposal, a set of ideas and a structure waiting to be tested in the real world. They have laid out a clear plan for how to test it, suggesting experiments in physical stores, online shops, and hybrid environments to see if this culturally adaptive approach actually leads to better experiences and more successful sales.

The work suggests that the future of retail is not just about having more data, but about having better context. It moves beyond the idea that a customer is a set of numbers to the idea that a customer is a person navigating a complex cultural world. By linking what people do with who they are and where they are, this new approach offers a path toward a retail environment that is not only smart but also wise. The researchers conclude that while the technology to personalize exists, the true challenge lies in making that personalization culturally sensitive. Their framework provides the first comprehensive map for how to build that bridge, turning a collection of separate insights into a single, coherent strategy for the future of shopping.

Technical Summary: From Data-Driven Personalization to Culturally Adaptive Merchandising

Problem Statement
The paper addresses a critical fragmentation in retail research and practice. While Artificial Intelligence (AI), behavioral analytics, and recommender systems have significantly advanced the capacity for data-driven personalization, these systems often fail to account for the multidimensional nature of cultural context. Current approaches frequently rely on observed behavioral signals (e.g., clicks, purchase history) or static demographic proxies (e.g., country codes) to infer preferences. However, consumer behavior is deeply shaped by local practices, symbolic meanings, language, religious beliefs, and social norms that vary even within national borders. The authors argue that existing literature treats AI personalization, cultural consumer behavior, and merchandising optimization as separate streams. Consequently, there is no unified, end-to-end framework that operationalizes the joint sequence of: behavioral data $\rightarrow$ multidimensional cultural context $\rightarrow$ intelligent adaptation $\rightarrow$ merchandising decisions $\rightarrow$ observed consumer response $\rightarrow$ feedback-driven updating.

Methodology
The study employs a systematic literature review guided by the PRISMA 2020 framework to synthesize evidence at the intersection of AI, data-driven retail, and cultural adaptation.

Data Source: A single bibliographic database, Scopus, was used to ensure reproducibility.
Search Strategy: An expanded query covering "smart technologies/data," "retail/merchandising," and "culture/cultural adaptation" was executed on September 22, 2026, covering the period 2019–2026.
Selection Process:
- Initial retrieval: 129 records.
- Filtering to Articles and Reviews: 60 records.
- Title/Abstract screening: 34 records retained.
- Preliminary eligibility: 22 reports advanced to retrieval.
- Retrieval status: 3 reports could not be retrieved; 19 reports were assessed for eligibility.
- Final Synthesis Set: 10 studies confirmed (9 excluded from the 19 assessed).
Appraisal: A structured domain-based appraisal was conducted across six domains (research objective clarity, data transparency, analytical appropriateness, cultural construct explicitness, link to retail decisions, and limitations) to assess methodological quality and relevance to the proposed framework, rather than applying a uniform risk-of-bias instrument.

Key Contributions
The paper makes four primary contributions:

Reproducible Synthesis: It replaces non-reproducible estimates of literature volume with an explicit, auditable Scopus search and screening process, identifying a focused corpus of 10 high-relevance studies.
Conceptual Distinction: It explicitly separates "technological personalization" (predicting individual fit based on history) from "cultural adaptation" (ensuring the intervention is appropriate within the specific cultural and situational context), arguing they are complementary rather than synonymous.
The CAMI-DM Framework: It proposes the Culturally Adaptive, AI- and Data-driven Merchandising (CAMI-DM) architecture. This framework integrates behavioral signals, multidimensional cultural context, situational conditions, and adaptive intelligence into a single decision system.
Mathematical Formulation: It introduces a spatial optimization layer using the Quadratic Assignment Problem (QAP). This layer mathematically links behavioral flows and cultural utility to physical shelf/slot assignment, subject to distance constraints that preserve metric consistency.

Results and Findings
The synthesis of the 10 included studies (published 2022–2026) reveals:

Data vs. Meaning: Behavioral data (e.g., transaction logs, eye-tracking, reviews) effectively identify what consumers attend to, but culture determines the meaning of those actions. For instance, the same visual presentation or recommendation can yield different responses depending on holistic vs. analytic thinking or specific cultural values like mianzi (face) or Buddhist reflective cognition.
Context-Dependent AI: AI personalization is not culturally neutral. Studies show that chatbot recommendations and automated retail acceptance vary significantly across cultural contexts (e.g., Brazil vs. England, Thailand's specific religious context).
Fragmented Architecture: While individual studies excel in specific areas (e.g., big-data analytics for modest fashion, Transformer-based sentiment analysis for wine reviews, phygital luxury retail storytelling), none operationalize a closed-loop system that updates merchandising decisions based on real-time cultural and behavioral feedback.
Gap Identification: The literature lacks a unified model that connects cultural representation to spatial optimization (planograms) and dynamic feedback loops.

The CAMI-DM Framework
The proposed framework consists of five layers:

State Representation ( $S_t$ ): A dynamic state comprising behavioral signals ( $B_t$ ), multidimensional cultural context ( $C_t$ ), situational/operational conditions ( $X_t$ ), and prior responses ( $R_{t-1}$ ).
Adaptive Cultural Intelligence ( $ACI_t$ ): A computational layer that fuses behavioral and contextual inputs using learned weights to infer actionable context.
Context-Aware Adaptation Engine ( $CAE_t$ ): Translates $ACI_t$ into a candidate set of merchandising actions (e.g., adjacency, signage, digital ordering) while applying ethical and operational constraints.
Spatial Optimization (QAP): A mathematical module that solves for optimal product placement ( $M^*_t$ ) by minimizing flow-weighted movement costs and maximizing culturally/behaviorally relevant utility, subject to capacity and distance constraints.
Response and Feedback: Measures customer experience (CX), trust, and behavioral response (BR) to update the state for the next cycle.

Significance and Claims
The paper positions CAMI-DM not as a claim that cultural adaptation is universally beneficial, but as a research architecture designed to address the integration problem in intelligent merchandising.

Theoretical Shift: It moves the conceptualization of culture from a static descriptive moderator (e.g., a country code) to an actionable, multidimensional input within a sequential decision architecture.
Operational Bridge: It bridges the gap between static segmentation models and dynamic learning systems, offering a pathway to test how cultural relevance influences trust, experience, and commercial performance.
Empirical Agenda: The authors propose a set of 11 testable hypotheses and an empirical validation protocol (including PLS-SEM and controlled experiments) to validate the framework in physical, digital, and phygital settings.
Modesty: The authors acknowledge that the framework is a proposal requiring empirical validation. They explicitly state that the study does not yet test the hypotheses and that the QAP formulation is a mathematical layer for decision-making, not a guarantee of commercial superiority. The core objective is to enable context-sensitive decision-making while preserving measurement rigor, transparency, and consumer autonomy.

Technical Summary: From Data-Driven Personalization to Culturally Adaptive Merchandising

More like this