Saumil Patel, Rogith Naini (2025). Human-Centered Design of Integrated Food Service Management Systems: Reducing Cognitive Load in Resource-Constrained Kitchen Operations. In: Human Factors in Design, Engineering, and Computing. AHFE Open Access, vol. 199, pp. 84–92. https://doi.org/10.54941/ahfe1006820
Small-scale food service operations face significant cognitive challenges in managing fragmented digital systems. This field study evaluated an integrated management platform designed using human-centered principles at a single restaurant over five weeks. Six staff members (tested two at a time across morning and evening shifts) served 612 customers, generating 967 transactions. The study compared two weeks of baseline operations using fragmented systems (Square POS, manual inventory, separate communication tools) against three weeks using the integrated platform. NASA-TLX measurements showed cognitive workload decreased from baseline (M=58.7, SD=10.2) to post-implementation (M=44.3, SD=7.8), a 24.5% reduction. System Usability Scale scored 85.8 (SD=7.4), indicating excellent usability. Operational improvements included 35% faster order processing (8.2 to 5.3 minutes), 65% faster payment completion (12.4 to 4.3 seconds), and 72% error reduction (12.1% to 3.4%). Customer satisfaction averaged 9.3/10 (n=147, 24% response rate). The system architecture demonstrates enterprise-grade capabilities (real-time communication, AI decision support, comprehensive monitoring) at small-business cost (~$240/month). Results suggest integrated human-centered systems can significantly reduce cognitive burden while improving operational efficiency in resource-constrained environments.
The Cognitive Overload Challenge Small-scale food service operations with limited staff face escalating cognitive demands that threaten operational viability and worker well- being. Industry data reveals 78% of small restaurants operate with fewer than three staff during peak periods while coordinating increasingly complex workflows across multiple disconnected digital platforms (National
Human-Centered Design of Integrated Food Service Management Systems 85
Restaurant Association, 2024). This operational context creates cognitive overload where information processing demands exceed working memory capacity, resulting in 40–60% increased error rates during peak service periods (Wickens, 2002).
Time-motion studies document over 42 attention switches per hour between disconnected systems for order management, payment processing, inventory tracking, and customer communication, with each context switch incurring cognitive costs of 400–1,000 milliseconds (Monsell, 2003). Restaurant kitchen workers experience cognitive workload comparable to air traffic controllers during peak service periods (Lundberg & Starfelt, 2019), contributing to 73% annual turnover rates and 30–40% business productivity losses.
Research Gap and Motivation Current academic literature addresses cognitive load primarily in educational contexts (Sweller, 1988; Paas et al., 2003) or examines individual interface design (Norman, 1986) but lacks comprehensive frameworks with practical implementation details for resource-constrained operational environments. Existing food service management solutions designed for large operations with specialized roles create cognitive fragmentation when one or two individuals must simultaneously manage all functions (Hutchins, 1995). More critically, while human-centered design principles are well established in theory (Norman, 1986; Shneiderman, 2020), there is limited documentation of complete technical architectures that small business practitioners can implement, limiting the practical impact of HCD research.
Commercial point-of-sale systems (Square, Toast, Clover) provide transaction processing but require separate applications for inventory management, customer communication, and business analytics. This fragmentation perpetuates the cognitive burden through forced context switching. Enterprise solutions (Oracle MICROS, NCR Aloha) offer integration but at cost points ($15,000–50,000 implementation plus $500– 2,000 monthly) prohibitive for small operations. Critically, neither category explicitly addresses cognitive load as a fundamental design constraint, nor do they provide sufficient customization for specialized operational requirements.
Research Objectives This field study investigates whether an integrated food service management system, designed using human-centered principles to reduce cognitive load, can improve staff workload, operational efficiency, and customer satisfaction in a real-world deployment. We address: (RQ1) Can an integrated system reduce staff cognitive workload compared to fragmented baseline systems? (RQ2) Does system integration improve operational efficiency metrics? (RQ3) What technical architecture patterns enable enterprise-grade capabilities at accessible cost points?
86 Patel and Naini
LITERATURE REVIEW AND THEORETICAL FOUNDATIONS Cognitive Load Theory
Cognitive Load Theory (CLT), originally developed for instructional design (Sweller, 1988), provides essential theoretical grounding for understanding information processing limitations in operational environments. CLT distinguishes three types of cognitive load: intrinsic load from inherent task complexity, extraneous load from poor information presentation, and germane load supporting schema construction and learning. Working memory capacity limits to approximately four chunks of information (Cowan, 2001), and exceeding this capacity degrades performance across all task types.
Task switching, endemic in multi-system operational environments, imposes cognitive costs of 400–1,000 milliseconds per switch (Monsell, 2003), with performance degradation persisting beyond the switch itself. In food service contexts where 40+ switches per hour are documented, this translates to 16–40 seconds of cognitive overhead per hour, or roughly 4–11% of working time lost purely to context management. More critically, switch costs compound: performance decreases proportionally to switching frequency, creating multiplicative rather than additive degradation.
Distributed Cognition and System Design Distributed cognition theory (Hutchins, 1995) emphasizes that cognitive processes extend beyond individual minds to encompass interactions among individuals, artifacts, and environments. This framework is particularly relevant for understanding how system design can either support or impede cognitive work. When systems fragment information across disconnected interfaces, they force cognitive processes that could be distributed across well-designed artifacts back onto individual working memory, creating unnecessary cognitive burden.
In operational environments, effective distributed cognition requires what we term unified context consolidation – bringing related information into single interfaces that support rather than fragment cognitive work. This contrasts with traditional systems that distribute information across multiple applications, each requiring separate logins, navigation patterns, and mental models.
Human-Centered Design Principles Norman’s (1986) foundational work on user-centered design emphasizes natural mappings, immediate feedback, and error prevention through well- designed constraints. Shneiderman’s (2020) recent perspective on human- centered AI highlights maintaining human control, supporting reliable operation, and ensuring system transparency. These principles guided our system development through five specific design commitments.
Unified Context Consolidation: Eliminate cognitive fragmentation by integrating the operational context into single, coherent interfaces rather than
Human-Centered Design of Integrated Food Service Management Systems 87
requiring navigation across multiple applications. This reduces extraneous cognitive load from context switching and enables focus on task completion.
Automated Background Coordination: Offload routine coordination tasks to automated systems while alerting humans only for exceptions requiring judgment (Parasuraman et al., 2000). This preserves cognitive resources for high-value activities while maintaining appropriate human oversight.
Perceptual-Optimized Visual Design: Leverage preattentive visual processing for parallel information intake through strategic use of spatial arrangement, color coding, and Gestalt principles (Koffka, 1935). This reduces working memory load by supporting pattern recognition rather than requiring sequential interpretation.
Contextual Progressive Disclosure: Present exactly the needed information at appropriate cognitive load levels with one-click access to details (Miller, 1956). Default views show essential information; comprehensive details remain immediately accessible without requiring additional applications or complex navigation.
Transparent Collaborative Intelligence: Implement AI assistance that maintains human oversight through transparent reasoning (Ribeiro et al., 2016), controllable overrides (Horvitz, 1999), and collaborative learning from human decisions (Amershi et al., 2014). AI augments rather than replaces human judgment.
Table 1 maps these design principles to specific system features, demonstrating how abstract HCD principles translate to concrete implementation decisions.
Table 1: Design principles mapped to system implementation features.
Table available in the original PDF paper.
Design Principle System Implementation Unified Context Consolidation Single dashboard integrating orders, payments, inventory, analytics, and customer communication. Real-time updates via SSE/WebSockets eliminate the need for manual refreshing or switching applications.
Automated Background Coordination
Automated inventory tracking, payment processing, order routing, and restock alerts. Staff are notified only for exceptions (e.g., low inventory, failed payments, unusual orders) requiring human judgment.
Perceptual- Optimized Visual Design
Color-coded order status (red = urgent, yellow = preparing, green = ready), spatial grouping of related items, and high-contrast alerts for exceptions. Visual scanning identifies status without reading text.
Order cards show essential information (customer, items, status, time). A single click expands to full details (special instructions, payment status, customer history) when needed.
Contextual Progressive Disclosure Transparent Collaborative Intelligence AI predictions are shown with confidence levels and reasoning. Staff can accept, modify, or reject suggestions. The system learns from human decisions to improve future recommendations.
88 Patel and Naini
METHODOLOGY
Study Design and Context This field study employed a single-site, within-subjects design comprising three phases over five weeks at a small fast-food restaurant in Southeast Texas. Phase 1 (weeks 1–2) established baseline measurements during normal operations with existing fragmented systems. Phase 2 (week 3) involved system deployment, staff training, and familiarization. Phase 3 (weeks 4–5) collected post-implementation measurements with the integrated system operational.
The restaurant operates with two workers per shift across morning (6 AM–2 PM) and evening (2 PM–10 PM) shifts, serving 30–50 customers daily with a mixed model of table service and takeout orders. The baseline system used a Square Terminal for payments (separate device), paper order slips for kitchen communication, Excel spreadsheet for inventory tracking, and personal mobile phones for customer communication. No integration existed between systems, requiring all coordination through manual staff effort.
Participants Six staff members participated (3 morning shift, 3 evening shift), ages 24–42 (M = 31.2, SD = 6.8), with 2–7 years food service experience (M = 4.2, SD = 1.9). All provided informed consent. Staff worked in pairs of two per shift, with different worker combinations tested across shifts to assess system performance across varied user pairings. Customer participants comprised 612 unique individuals generating 967 transactions (baseline: 289 customers, 458 transactions; post-implementation: 323 customers, 509 transactions). Post-implementation customer surveys achieved 24% response rate (n = 147/612).
Measurement Instruments NASA Task Load Index (NASA-TLX) assessed perceived workload across six subscales: Mental Demand, Physical Demand, Temporal Demand, Performance, Effort, and Frustration (Hart & Staveland, 1988). Staff completed NASA-TLX at shift end twice weekly (N = 60 assessments total, 30 baseline, 30 post-implementation). System Usability Scale (SUS) evaluated usability at study end (Brooke, 1996). Objective operational metrics included order processing time, payment completion time, and order error rates, automatically logged post-implementation and manually observed during baseline. Customer satisfaction surveys (5 items, 0–10 scale) were administered via email 24 hours post-service.
Data Collection Procedures Given the field study context and within-subjects design, we present descriptive statistics (means, standard deviations) and percentage changes between baseline and post-implementation periods. NASA-TLX scores were averaged across all assessments per period. Operational metrics were
Human-Centered Design of Integrated Food Service Management Systems 89
aggregated over each complete period. We acknowledge that without a control group, observed changes may reflect factors beyond the system intervention, including practice effects, seasonal variations, or measurement artifacts. Results should be interpreted as preliminary evidence suggesting potential benefits worthy of controlled validation.
SYSTEM ARCHITECTURE AND IMPLEMENTATION The integrated platform employs a modern full-stack architecture: Next.js 14 with React 18 frontend providing unified interfaces; Node.js backend with REST and real-time endpoints; PostgreSQL for structured transaction data and MongoDB for flexible analytics; Redis caching for sub-100ms responses. Real-time communication uses Server-Sent Events for status streaming and WebSockets for bidirectional collaboration, achieving 99.7% uptime and 47ms median latency. Authentication employs NextAuth with multiple providers (Google, GitHub, Email OTP, Firebase Phone), secured by Argon2 hashing and rate-limited JWT sessions. Stripe Payment Intents API enables atomic, PCI-compliant transactions with 99.3% success rate. AI modules provide ARIMA demand forecasting and collaborative-filtering recommendations, achieving 94% predictive accuracy with transparent confidence indicators and human override capability. System observability via Prometheus metrics, OpenTelemetry tracing, Sentry error monitoring, and Grafana dashboards ensures production reliability. Total operational cost remains under $240 monthly, demonstrating enterprise-grade capabilities at small-business scale.
RESULTS
Cognitive Workload Assessment (NASA-TLX) Overall cognitive workload significantly decreased from baseline (M = 58.7, SD = 10.2) to post-implementation (M = 44.3, SD = 7.8), t(29)=5.82, p<0.001, Cohen’s d = 1.58, representing a 24.5% reduction with a very large effect size. Table 3 presents subscale comparisons. The largest reductions occurred in Frustration (33.6%), Mental Demand (27.4%), and Effort (24.3%). These improvements demonstrate the integrated system successfully reduced cognitive burden through context consolidation and workflow automation.
Table 2: NASA-TLX subscale scores: baseline vs. post-implementation.
Operational Efficiency Operational metrics showed substantial improvements across all measured dimensions. Order processing time significantly decreased 35% from baseline (M = 8.2 min, SD = 1.9) to post-implementation (M = 5.3 min, SD = 1.2), t(965)=28.4, p<0.001, Cohen’s d = 1.78. Payment completion time significantly reduced 65% from baseline (M = 12.4 sec, SD = 2.7) to post-implementation (M = 4.3 sec, SD = 1.1), t(965) = 54.7, p<0.001, Cohen’s d = 3.96. Order error rates significantly decreased 72% from baseline (12.1%, 55/458 transactions) to post-implementation (3.4%, 17/509 transactions), χ2(1)=52.8, p<0.001, ϕ=0.23. Context switches per hour declined approximately 91% from baseline (∼42 switches) to post- implementation (∼4 switches), based on observational time-sampling data.
System Usability and Customer Satisfaction System Usability Scale assessment yielded a mean score of 85.8 (SD = 7.4, range = 75–95), indicating excellent usability. All six staff members rated the system above 75, with four exceeding 85 (top 10% of systems). Staff particularly appreciated the unified interface (M = 4.8/5), ease of use (M = 4.7/5), and system integration (M = 4.8/5). Post-implementation customer satisfaction averaged 9.3/10 (SD = 1.1, n = 147), with high ratings across all dimensions: order accuracy (9.4/10), service speed (9.2/10), communication clarity (9.3/10), and payment ease (9.4/10). The 24% response rate suggests potential self-selection bias toward satisfied customers.
DISCUSSION
Mechanisms of Cognitive Load Reduction Five mechanisms explain observed cognitive load reduction: (1) Context switch elimination reduced overhead from ∼42 to 4 switches/hour, freeing 11–17% of working time; (2) Working memory offloading through unified interfaces eliminated mental tracking across systems; (3) Routine automation reduced daily exception events from 18–24 to 3–4; (4) Perceptual optimization enabled preattentive processing through color-coded status indicators; (5) Progressive disclosure presented information at appropriate detail levels, reducing visual clutter while maintaining drill-down access. These mechanisms align with cognitive load theory predictions and demonstrate practical application of human-centered design principles.
Practical Implications This study demonstrates that enterprise-grade capabilities (comprehensive monitoring, AI decision support, robust authentication, real-time communication) can be deployed at small-business cost points (∼$240/month). The system architecture—Next.js unified platform, hybrid database approach, modular authentication, Redis caching—provides replicable patterns for practitioners. The 35% reduction in order processing time and 72% reduction in error rates indicate substantial operational
Human-Centered Design of Integrated Food Service Management Systems 91
value. For small restaurants operating on thin margins, these improvements translate directly to enhanced customer experience and staff retention.
Study Limitations This field study has several important limitations. The single-site design limits external validity; results may reflect site-specific factors. Without a control group, we cannot definitively attribute improvements to the system versus alternative explanations (practice effects, seasonal variations, Hawthorne effects). The five-week duration may capture initial enthusiasm rather than sustained benefits. Different measurement methods (manual observation baseline vs. automated logging post-implementation) may create systematic differences. The 24% customer survey response rate raises concerns about non-response bias. Testing six workers (vs. two) across different shift combinations strengthens confidence in usability findings but does not address single-site limitations. Future research should employ multi-site randomized controlled designs with extended observation periods.
CONCLUSION
This field study provides preliminary evidence that integrated food service management systems, designed using human-centered principles to explicitly address cognitive load, can reduce staff cognitive burden (24.5% workload reduction), improve operational efficiency (35% faster processing, 72% fewer errors), and enhance usability (SUS = 85.8). The demonstration that enterprise-grade capabilities are deployable at small-business scale (∼$240/month) challenges assumptions about technological accessibility. Testing with six workers across different shift pairings provides reasonable confidence in system usability and reliability.
While the single-site design and lack of control group require cautious interpretation, the observed improvements across multiple independent measures (subjective workload, operational efficiency, usability, customer satisfaction) suggest genuine system benefits. The cognitive overload facing small food service operations represents both a human welfare concern (73% staff turnover) and business sustainability issue (30–40% productivity losses). This work demonstrates that systematic application of human- centered design principles to integrated architectures offers a promising approach worthy of rigorous controlled validation.
Complete technical documentation enables practitioners to replicate and adapt this implementation. Future research should extend this work through multi-site randomized controlled trials with longer observation periods (12+ weeks), diverse operational contexts, and comprehensive cost- benefit analyses. This foundational field study establishes feasibility and provides preliminary evidence supporting the hypothesis that integrated human-centered systems can significantly improve outcomes for resource- constrained food service operations.
92 Patel and Naini
REFERENCES
Amershi, S., Weld, D., Vorvoreanu, M., Fourney, A., Nushi, B., & Collisson, P.
(2014). Power to the people: The role of humans in interactive machine learning. AI Magazine, 35(4), 105–120. Box, G. E. P., & Jenkins, G. M. (1976). Time series analysis: Forecasting and control.
San Francisco: Holden-Day. Brooke, J. (1996). SUS: A “quick and dirty” usability scale. In P. W. Jordan et al.
(Eds.), Usability Evaluation in Industry (pp. 189–194). London: Taylor & Francis. Bureau of Labor Statistics (2023). Job openings and labor turnover survey.
Washington, DC: U.S. Department of Labor. Cowan, N. (2001). The magical number 4 in short-term memory. Behavioral and Brain Sciences, 24(1), 87–114. Gold, R. L. (1958). Roles in sociological field observations. Social Forces, 36(3),
217–223. Hart, S. G., & Staveland, L. E. (1988). Development of NASA-TLX: Results of empirical and theoretical research. In P. A. Hancock & N. Meshkati (Eds.), Human Mental Workload (pp. 139–183). Amsterdam: North-Holland. Horvitz, E. (1999). Principles of mixed-initiative user interfaces. In Proceedings of
the SIGCHI Conference on Human Factors in Computing Systems (pp. 159–166). New York: ACM. Hutchins, E. (1995). Cognition in the wild. Cambridge, MA: MIT Press. Koffka, K. (1935). Principles of Gestalt psychology. New York: Harcourt, Brace & World. Koren, Y., Bell, R., & Volinsky, C. (2009). Matrix factorization techniques for
recommender systems. Computer, 42(8), 30–37. Lundberg, J., & Starfelt, M. (2019). Cognitive work analysis in the kitchen. In Proceedings of the Human Factors and Ergonomics Society Annual Meeting, 63(1), 1234–1238. Miller, G. A. (1956). The magical number seven, plus or minus two. Psychological
Review, 63(2), 81–97. Monsell, S. (2003). Task switching. Trends in Cognitive Sciences, 7(3), 134–140. National Restaurant Association (2023). State of the restaurant industry report.
Washington, DC: National Restaurant Association. Norman, D. A. (1986). The psychology of everyday things. New York: Basic Books. Paas, F., Renkl, A., & Sweller, J. (2003). Cognitive load theory and instructional design. Educational Psychologist, 38(1), 1–4. Parasuraman, R., Sheridan, T. B., & Wickens, C. D. (2000). A model for types and
levels of human interaction with automation. IEEE Transactions on Systems, Man, and Cybernetics – Part A, 30(3), 286–297. Ribeiro, M. T., Singh, S., & Guestrin, C. (2016). Why should I trust you?
Explaining the predictions of any classifier. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 1135–1144). New York: ACM. Shneiderman, B. (2020). Human-centered artificial intelligence: Reliable, safe & trustworthy. International Journal of Human–Computer Interaction, 36(6), 495–504. Sweller, J. (1988). Cognitive load during problem solving: Effects on learning.
Cognitive Science, 12(2), 257–285. Wickens, C. D. (2002). Multiple resources and performance prediction. Theoretical Issues in Ergonomics Science, 3(2), 159–177.
LISTED IN THE AHFE DIGITAL BOOKSTORE CATALOGUE
Explore the complete digital volume
Human Factors in Design, Engineering, and Computing
Read this individual research paper free through the original publisher, or explore the full digital proceedings book.
ISBN 978-1-964867-75-5
Explore this ISBN in the AHFE Bookstore ↗Listed in the July 2026 bookstore catalogue; search by ISBN to confirm current availability · Digital book purchase optional