SmartCost
AI-Driven POS & Cost Analytics System
for Small & Medium-Sized Enterprises (SMEs)
SmartCost is a mobile-first, cloud-based business management platform designed to help small and medium-sized businesses manage their daily operations, understand business performance, and make informed decisions through configurable business tools and data-driven research. Designed for SMEs such as retail stores, restaurants, cafés, bakeries, and salons, it delivers configurable modules across POS, inventory management, cost control, customer management, analytics, and reporting, supported by 4 targeted research novelties.
Novelties
Industries
First
Backend
Cost
Addressing SME Operational Challenges
Many small and medium-sized businesses rely on manual processes, disconnected tools, or software that may not fully meet their needs for affordability, mobility, flexibility, and data-driven decision-making. This research investigates how a mobile-first, cloud-based, configurable business management platform can support SMEs while exploring intelligent approaches to demand prediction, customer insights, product recommendations, and sales trend analysis.
Read Full Problem Statement →Configurable Platform with Research-Driven Intelligence
SmartCost combines configurable business management tools (POS, inventory, cost management, customer management, analytics, and reporting) with research into predictive analytics, customer insights, intelligent product recommendations, and sales forecasting. The existing system features provide day-to-day operational support, while the proposed research components explore data-driven enhancements for business decision-making.
⚡ Explore The SmartCost Platform Architecture →Four Core Research Contributions
Product Demand Prediction
Investigating machine learning techniques to predict future product demand using historical sales data.
02Customer Behavior Analysis & Sales Insights
Analyzing transaction data to identify purchasing patterns, popular products, and peak sales periods.
03AI-Based Smart Order Recommendation
Investigating intelligent recommendations based on current cart and historical transaction patterns.
04Sales Trend Analysis & Forecasting
Investigating historical sales patterns to identify trends and support business planning decisions.
The SmartCost Platform
A unified, modular platform connecting diverse business types with intelligent feature modules.
One unified architecture. Six commercial presets.
Instant offline POS execution with continuous background machine learning.
Retail Stores
POS, Barcode & Stock sync
Restaurants
Table billing & Menu tracking
Cafés
Fast-order & Loyalty points
Bakeries
Batch costing & Freshness
Salons
Service catalog & Appointments
Service Businesses
Custom jobs & Expense control
Configurable Feature Modules
Each business can enable only the tools they need while keeping operations unified.
POS Billing
Fast & intuitive checkout
Inventory Management
Live stock tracking
Cost Management
Expense & margin analysis
Real-time Analytics
Live sales performance
KPI Dashboard
Executive business metrics
Customer CRM
Purchase history & loyalty
AI Insights
Demand & order recommendations
Marketing & SMS
Automated customer outreach
Research Domain & Academic Foundation
Detailed academic domain exploration structuring literature findings, research gaps, formal problem framing, approved research objectives, methodology, novelties, and technologies used.
Literature Survey & Comparative Analysis
Survey of existing retail management systems, published academic methodologies, and comparative evaluation against traditional commercial POS tools.
| Evaluation Criteria | Manual / Spreadsheet | Enterprise Cloud POS | SmartCost (Proposed) |
|---|---|---|---|
| Hardware Cost & Accessibility | Paper ledgers & spreadsheets (High human error, no upfront hardware) | Proprietary heavy hardware terminals (High upfront cost + maintenance) | Mobile-first, runs on any existing smartphone or tablet ($0 hardware investment) |
| Predictive Intelligence | Gut-feeling guessing; high stockout and spoilage rates | Basic retrospective descriptive charts; no on-device proactive ML | Lightweight AI models for demand prediction & smart recommendations |
| Configuration & Multi-Industry Fit | Completely manual bookkeeping; difficult to scale or adapt | Rigid industry silos (separate software for café vs salon vs retail) | Modular configurable architecture adapting to Retail, Restaurants, Bakeries, Salons |
| Network Dependency & Offline Execution | 100% manual offline, zero cloud aggregation | Strict continuous cloud connectivity required; fails on outages | Edge-local offline capability with automatic background cloud sync |
Research Papers & Academic Publications
Peer-reviewed research papers produced by the SmartCost research cohort, detailing our architectural design, machine learning models, and empirical benchmark evaluations.
SmartCost: A Hybrid AI-Driven POS and Cost Analytics System for Micro-SMEs
Presents the integrated SmartCost mobile analytics platform uniting point-of-sale operations, cost control, demand forecasting, business performance prediction, customer analytics, and real-time checkout recommendations. Evaluated on over 15,000 transactions across 128 products, achieving 91% demand forecasting accuracy, 82% recommendation confidence, and 18% inventory waste reduction.
SmartCost: Product Demand Prediction System for Micro-SMEs
Evaluates ARIMA and Prophet forecasting models on 2-year daily sales records for 50 products. Prophet achieved an 8.7% MAPE (vs 12.4% for ARIMA) and reduced RMSE by 18%, yielding a 14% reduction in inventory waste.
SmartCost: Customer Behavior Analysis & Sales Insights System for Micro-SMEs
Applies transactional data mining, RFM segmentation, K-Means clustering, and Mann-Kendall trend analysis on 15,420 transactions. Achieved 91% trend accuracy and 82% association rule confidence, boosting sales by 11%.
SmartCost: Smart Order Recommendation System for Micro-SMEs
Real-time POS checkout product recommendation engine using Apriori association rule mining. Achieved 84% precision, 79% recall, and 142 ms response latency, enhancing cross-selling effectiveness by 18%.
SmartCost: Business Performance Prediction System for Micro-SMEs
Couples demand forecasting with ingredient-level cost modeling to predict COGS, profit, and margins. Achieved 6.4% margin prediction error and 89% precision for low-margin risk alert generation.
Identified Research Gaps
Limitations identified during domain inquiry. These are provisional findings that will be refined as the literature review progresses.
Enterprise-Biased AI Model Complexity
State-of-the-art predictive algorithms assume high-performance server clusters and vast enterprise data warehouses, limiting their applicability to resource-constrained SMEs.
Investigating lightweight ML models capable of practical inference on standard mobile devices for SME use.
Prohibitive Cost & Hardware Requirements
Modern cloud POS solutions charge steep per-terminal monthly fees and require proprietary hardware lock-in, making them inaccessible for micro-merchants.
Delivering a mobile-first cloud system running on merchants' existing smartphones without recurring hardware costs.
Lack of Mobile-First SME Business Management
Most business management platforms are designed for desktop environments, offering limited or secondary mobile experiences unsuitable for on-the-go SME operations.
Investigating a mobile-first architecture approach where the primary interaction model is designed for smartphones and tablets.
Limited Data-Driven Decision Support for SMEs
Small businesses typically lack access to analytical tools that transform their transactional data into actionable business intelligence and forward-looking insights.
Exploring demand prediction, customer behavior analysis, smart recommendations, and sales trend forecasting specifically for SME data scales.
Research Problem Statement
Academic framing of the operational and economic challenges faced by small and medium enterprises.
Many small and medium-sized businesses rely on manual processes, disconnected tools, or software that may not fully meet their needs for affordability, mobility, flexibility, and data-driven decision-making. This research investigates how a mobile-first, cloud-based, configurable business management platform can support SMEs while exploring intelligent approaches to demand prediction, customer insights, product recommendations, and sales trend analysis.
Research Objectives
Main overarching project objective and four specific research goals corresponding to project novelties. These are draft objectives pending final approval.
To design, develop, and evaluate a mobile-first, cloud-based, configurable, multi-industry business management and Point of Sale (POS) platform supported by data-driven research components for small and medium-sized enterprises.
The platform aims to unify POS, inventory tracking, cost management, customer insight mining, and reporting into an accessible interface deployable on everyday mobile hardware.
Product Demand Prediction
To investigate an ML-based demand forecasting approach using historical sales records to predict upcoming product demand and support inventory management decisions.
Customer Behavior Analysis & Insights
To design an analytical approach that examines customer purchasing habits, identifies peak commercial windows, and surfaces actionable sales intelligence patterns.
AI-Based Smart Order Recommendation
To develop a recommendation approach that analyzes current cart contents against historical item affinities to suggest relevant supplementary items during checkout.
Sales Trend Analysis & Forecasting
To build an analytical module that examines longitudinal sales data to identify trends, support business planning, and provide viability guidance.
Research Methodology
Structured research process from literature review to system validation. Stages are marked as proposed, in progress, or completed.
Four Core Research Novelties
Detailed academic descriptions of the four foundational research contributions of the SmartCost project.
Product Demand Prediction
Investigate the use of historical sales data and machine learning techniques to predict future product demand, with the aim of reducing stock shortages and overstocking.
View Research Focus & Details
Research Focus Areas:
- Historical sales patterns and product demand trends
- Data preprocessing and feature selection
- Prediction model development
- Forecast evaluation using appropriate metrics
- Potential reduction of stock shortages and overstocking
- Models Evaluated: Prophet (Additive Time-Series) vs. ARIMA baseline
- Dataset: 24-month daily sales, 50 products, 24,000+ transaction records (80-150 daily orders)
- Evaluation Metrics: MAE, RMSE, MAPE (across 1-day & 7-day forecast horizons)
- Empirical Results: Prophet achieved 8.7% MAPE (vs 12.4% ARIMA), 18% RMSE reduction, 14% inventory waste reduction
- Research Paper: SmartCost: Product Demand Prediction System for Micro-SMEs (PDF)
Customer Behavior Analysis and Sales Insights
Investigate how historical transaction and customer data can be analyzed to identify useful patterns for business decision-making.
View Research Focus & Details
Research Focus Areas:
- Customer purchasing patterns and frequency
- Popular and slow-moving products identification
- Peak sales periods and time-based patterns
- Day-of-week and seasonal behavior analysis
- Sales performance insights generation
- Analytical Techniques: RFM Customer Segmentation, K-Means Clustering, Association Rule Mining, Mann-Kendall Trend Analysis
- Dataset: 15,420 transactions, 128 products, 12 months simulated bakery & café records
- Evaluation Metrics: Trend accuracy, clustering quality, association rule confidence, pattern frequency
- Empirical Results: 91% trend accuracy, 82% association rule confidence, 18% inventory optimization, 11% sales increase
- Research Paper: SmartCost: Customer Behavior Analysis & Sales Insights System for Micro-SMEs (PDF)
AI-Based Smart Order Recommendation
Investigate an intelligent recommendation approach that suggests relevant additional products based on the current shopping cart or order and historical transaction patterns.
View Research Focus & Details
Potential Research Areas:
- Neural Collaborative Filtering or another justified ML approach
- Historical transaction analysis for pattern mining
- Shop-specific product recommendations
- Top-K recommendation ranking
- Comparison against baselines (e.g., Apriori or FP-Growth)
- Evaluation using Precision@K, Recall@K, Hit Rate@K, NDCG@K
These are possible approaches, not claims that algorithms have already been implemented.
- Recommendation Engine: Apriori Association Rule Mining with Support, Confidence (>=60%), Lift ranking, and real-time Top-N ranking
- Dataset: 15,420 transactions, 128 products, 12 months, 3.8 average cart size
- Evaluation Metrics: Precision@K, Recall@K, Hit Rate@K, Average Recommendation Confidence, Latency
- Empirical Results: 84% precision, 79% recall, 81% hit rate, 82% confidence, 142 ms average response latency, 18% cross-selling boost
- Research Paper: SmartCost: Smart Order Recommendation System for Micro-SMEs (PDF)
Sales Trend Analysis and Forecasting
Investigate historical sales patterns to identify trends and support business planning, including analysis of growth and decline patterns and seasonal behavior.
View Research Focus & Details
Research Focus Areas:
- Historical sales analysis and pattern recognition
- Sales growth and decline pattern identification
- Seasonal or periodic behavior, where supported by data
- Trend visualization and interpretation
- Forecasting approaches, if included in the approved research
- Evaluation and interpretation of findings
- Integrated Models: Prophet & ARIMA Demand Forecasting + Ingredient Cost Modeling (COGS, Revenue, Profit Margin) + Alert Mechanism (15% margin threshold)
- Dataset: 24-month dataset containing 24,000+ sales and ingredient-cost records across 50 products
- Evaluation Metrics: MAE, RMSE, Profit Margin Prediction Error, Alert Precision & Recall
- Empirical Results: 6.4% profit-margin prediction error (vs 16.8% naive baseline), 89% alert precision, 14% waste reduction, 17% stock-out reduction
- Research Paper: SmartCost: Business Performance Prediction System for Micro-SMEs (PDF)
Technologies Used
Technologies used in the SmartCost research system (note: this showcase website itself uses only HTML and CSS as required).
Frontend & Mobile
- React Native
- Next.js & React
- CSS
Backend & APIs
- Node.js
- Express.js REST APIs
Database & Storage
- MongoDB Atlas
- SQLite
Machine Learning
- Python
- Scikit-Learn
- Data analysis libraries
Mobile Development
- React Native (Cross-platform)
- Expo
Deployment & Infrastructure
- Cloud Hosting
- Git Version Control
Research Project Milestones
University assessment roadmap. Select an assessment below to view its details, marks allocation, and required deliverables.
Project Proposal
Scope & Overview
Proposal report and presentation covering research novelty identification, literature survey foundation, technical feasibility, and supervisor endorsement.
Key Deliverables
Progress Presentation 1
Scope & Overview
Present progress of the individual and group components. First milestone evaluation reviewing formulation of the 4 novelties, initial dataset curation, and foundational architecture.
Key Deliverables
Progress Presentation 2
Scope & Overview
Present progress of the individual and group components. Second milestone evaluation reviewing implementation progress, working prototype, and benchmark results.
Key Deliverables
Final Assessment
Scope & Overview
Individual and Group reports, Final presentation, and website. Comprehensive evaluation of the completed SmartCost research platform, system validation, and submission of the final thesis.
Key Deliverables
Viva
Scope & Overview
Individual viva examination before the academic faculty panel covering individual novelty contributions, algorithmic rigor, and research findings.
Key Deliverables
Total Assessment: 100 Marks
Research Documents
Access Topic Assessment Form (TAF), project charters, research proposals, checklists, research papers, and final academic reports.
Topic Assessment Form (TAF)
Official SLIIT CDAP Topic Assessment Form approving and registering the research topic, project objectives, novelty breakdown, and supervisor endorsement for Group R26-IT-026.
Project Charter Document
Formal project charter outlining research scope, stakeholder commitments, initial timeline, and supervisor endorsements.
Individual Proposal: Sales Trend Forecasting
Individual research proposal by K.A.C.H Kodithuwakku (IT22544872) covering novelty formulation, data methodology, and forecast metrics.
Individual Proposal: Customer Behavior Analysis
Individual research proposal by Perera N.K.M (IT22151810) detailing customer clustering algorithms and basket association mining.
Individual Proposal: Product Demand Prediction
Individual research proposal by D.D.M. Jayasingha (IT22162250) detailing time-series demand models and inventory optimization.
Individual Proposal: Smart Order Recommendation
Individual research proposal by M. P. J. R. Dasanayaka (IT22152664) detailing collaborative filtering and real-time checkout suggestions.
SmartCost: A Hybrid AI-Driven POS and Cost Analytics System for Micro-SMEs
Overarching IEEE conference paper detailing the 4-module architecture (Demand, Profit, Behavior, Recommendation) achieving 91% forecasting accuracy and 18% inventory waste reduction.
SmartCost: Product Demand Prediction System for Micro-SMEs
Evaluates ARIMA and Prophet models on 2-year daily sales records across 50 products. Prophet achieved 8.7% MAPE (vs 12.4% ARIMA) and reduced RMSE by 18%.
SmartCost: Customer Behavior Analysis & Sales Insights System for Micro-SMEs
Applies transactional data mining, RFM segmentation, K-Means clustering, and Mann-Kendall trend detection on 15,420 transactions with 91% trend accuracy and 82% rule confidence.
SmartCost: Smart Order Recommendation System for Micro-SMEs
Real-time POS checkout recommendation engine using Apriori association rule mining. Achieved 84% precision, 79% recall, and 142 ms response latency.
SmartCost: Business Performance Prediction System for Micro-SMEs
Couples demand forecasting with ingredient-level cost modeling to predict COGS, profit, and margins. Achieved 6.4% margin prediction error and 89% alert precision.
Final Research Thesis (Main Group)
Comprehensive group final dissertation covering system design, architecture, integrated evaluations, and joint conclusions.
Individual Thesis: Sales Trend Forecasting
K.A.C.H Kodithuwakku (IT22544872) — Individual final thesis covering time-series models, metrics, and experimental results.
Individual Thesis: Customer Behavior Analysis
Perera N.K.M (IT22151810) — Individual final thesis covering customer segmentation, transaction mining, and behavior patterns.
Individual Thesis: Product Demand Prediction
D.D.M. Jayasingha (IT22162250) — Individual final thesis covering demand prediction algorithms, evaluations, and inventory impacts.
Individual Thesis: Smart Order Recommendation
M. P. J. R. Dasanayaka (IT22152664) — Individual final thesis covering recommendation engine architectures and checkout validation.
Proposal Assessment Checklist
SLIIT research proposal checklist confirming topic clearance, supervisor approvals, and novelty criteria.
Final Thesis Submission Checklist
Faculty compliance checklist covering CDAP formatting rules, originality reports, and code artifact requirements.
Presentations & Slide Repository
Access slides used for academic reviews, milestone assessments, and final project defense.
Proposal Presentation Slides
Initial academic proposal presentation defending project motivation, SME challenges, and novelty identification.
Progress Presentation 1 Slides
Progress review deck detailing literature survey findings, formulation of the 4 research novelties, and dataset preparation.
Progress Presentation 2 Slides
Second progress review covering implementation status, working prototype, evaluation benchmarks, and revised research timelines.
Final Assessment Presentation
Comprehensive research presentation covering the end-to-end SmartCost platform, research findings, and evaluation results.
The Minds Behind SmartCost
Guided by academic supervisors and engineered by student researchers at the Sri Lanka Institute of Information Technology (SLIIT).
Project Leadership & Supervision
Mr. Deemantha Siriwardana
SUPERVISORAcademic Project Supervisor
Guiding research formulation, algorithmic feasibility, and overall academic rigor.
Miss. Ayesha Wijesooriya
CO-SUPERVISORAcademic Project Co-Supervisor
Supervising system architecture, empirical evaluation metrics, and quality compliance.
Research Team & Novelty Ownership
K.A.C.H Kodithuwakku
IT22544872Sales Trend Analysis & Forecasting
Responsible for trend modeling, sales pattern analysis, and forecasting approaches for SME viability metrics.
Perera N.K.M
IT22151810Customer Behavior Analysis
Responsible for transaction clustering, customer segmentation, and peak-hour sales pattern mining.
D.D.M. Jayasingha
IT22162250Product Demand Prediction
Responsible for time-series demand forecasting, model formulation, and inventory optimization research.
M. P. J. R. Dasanayaka
IT22152664Smart Order Recommendation
Responsible for recommendation model research, basket analysis, and real-time checkout suggestions.
Contact Project Team
Connect with the SmartCost student researchers and academic supervisors.
Sri Lanka Institute of Information Technology
Faculty of Computing • Department of Computer Science & Software Engineering
📍 Campus: Malabe Campus, New Kandy Road, Malabe, Sri Lanka
✉️ Project Email: smartcost.research@gmail.com
📞 Phone: +94 11 754 4801 (SLIIT General Line)
Send an Inquiry
This form opens your default email client. No server-side processing is used.
Or email directly: smartcost.research@gmail.com