Domain

Artificial Intelligence in Retail

Artificial Intelligence (AI) applications in retail is the domain, with a

specific focus on dynamic pricing and inventory optimization. The
retail domain, especially at the small and medium enterprise (SME) level,
is undergoing a digital transformation driven by the need for data-
informed decisions. AI technologies such as machine learning, Bayesian
optimization, and rule-based automation empower retailers to respond
to market fluctuations, customer demand, and competitive pressure in real-
time. Addressing the pressing need for accessible, intelligent pricing
systems in retail environments with limited technical resources, bridging
the gap between traditional pricing methods and modern AI-driven
solutions is the objective.

02
Abstract

The dynamic nature of retail demands agile pricing strategies, yet

small-scale retailers often lack the infrastructure and data capabilities to adopt
sophisticated models. This project proposes a hybrid dynamic pricing
framework, combining rule-based logic with Bayesian Optimization, tailored
for low-resource retail environments. By leveraging minimal data inputs such
as inventory levels and sales trends, the system enables real-time price
adjustments without requiring advanced infrastructure or large datasets. The
proposed model, named Cogni Price, improves revenue, reduces overstocking,
and enhances competitive positioning. Simulations demonstrate its
effectiveness and practicality, making AI-driven pricing accessible to small
businesses.

03
Introduction
Overview of the Project :

Introduces an AI-powered hybrid dynamic pricing strategy

designed specifically for low-scale retailers who operate with limited
technical resources and minimal data. Traditional static pricing methods
often result in overstocking, missed sales, and poor market responsiveness.
The proposed system, Cogni Price, integrates simple rule-based logic with
Bayesian Optimization to enable real-time, adaptive pricing. It utilizes
basic inputs such as inventory levels and sales trends—eliminating the need
for large datasets or high-end infrastructure. Simulations conducted in small
retail scenarios demonstrate improvements in revenue, inventory
turnover, and price competitiveness, highlighting its practicality and
scalability for small businesses.

03
Introduction
Problem Statement

Small-scale retailers struggle to implement dynamic pricing due to

limited technical infrastructure, high data requirements, and the
complexity of existing AI models. Traditional static pricing methods
lead to lost revenue, poor inventory turnover, and reduced
competitiveness. There is a need for a simple, affordable, and
intelligent pricing system that works with minimal data and helps
small retailers adapt to changing market conditions in real-time.

04
Introduction
Objectives of the Project:

The primary objective is to develop a hybrid dynamic pricing

strategy for low-scale retailers that leverages minimal data to optimize
pricing decisions and inventory performance in real time.
Sub-Objectives:
•✅ Design a rule-based system to handle basic pricing logic using
inventory and sales data.
•✅ Integrate Bayesian Optimization to fine-tune prices and maximize
revenue with minimal data input.
•✅ Ensure the system is cost-effective, interpretable, and easy to
deploy without complex IT infrastructure.
•✅ Validate the framework through simulations and performance
comparisons with static pricing methods.
•✅ Enable small retailers to respond to market trends and competitor
pricing effectively.
05
Literature Review

Techniques / Performance
Author Name Title of the
S.No Problems Considered Algorithms Metrics Limitations
& Year Paper
Implemented Considered
Adaptive
Pricing Inflexible pricing No real-time
Johnson & Lee, Rule-based Sales volume,
1 Strategies for models for small adaptability, ignores
2019 dynamic pricing profit margin
Small retailers competitor pricing
Retailers
Simplified
Dynamic Complexity of dynamic Heuristics,
Davis et al., Profit margin, Lacks flexibility and
2 Pricing for pricing for low-scale statistical
2017 sales growth scalability
Small retailers models
Businesses
Dynamic
Heuristic pricing Cannot capture
O’Connor & Pricing in the Dynamic pricing with Revenue trend,
3 using short-term complex market
Singh, 2019 Context of minimal data pricing accuracy
sales data dynamics
Limited Data
Cost-Effective Rule-based Limited vendor
Brown & Davis, Dynamic High implementation systems using ROI, ease of use, support,
4
2021 Pricing costs for SMEs open-source revenue growth customization
Solutions tools challenges
Competitive
Competitor Market share,
Li & Zhang, Pricing Lack of adaptation to Requires real-time
5 price scraping pricing
2022 Strategies in competitor pricing external data access
algorithm competitiveness
Retail Markets
07
Literature Review

Inventory
Management Pricing based Ignores competitor
Garcia & Overstocking and Inventory turnover,
6 through on stock levels pricing and customer
Kumar, 2021 stockouts holding costs
Dynamic & sales velocity behavior
Pricing
Evaluating the
Profitability, price
Effectiveness Variability in model Comparative Depends on data
Hernandez et optimization
7 of Dynamic performance across pricing model quality, short-term
al., 2020 accuracy, sales
Pricing conditions evaluation variability
volume
Models
Customer-
Centric Need for Customer Repeat purchase Privacy concerns
Martinez et al.,
8 Dynamic personalization in behavior-based rate, customer and data availability
2019
Pricing pricing dynamic pricing satisfaction issues
Models
Real-Time
Time-series
Chen et al., Pricing Delayed response of Price elasticity, Requires continuous
9 forecasting-
2020 Models in E- existing models conversion rates data feeds
based model
commerce
Dynamic
Pricing in Revenue
Machine High data
Smith et al., Retail: A Inefficiencies of static optimization,
10 Learning-based dependency and
2018 Review of pricing in retail customer
pricing models computational cost
Algorithms & satisfaction
Applications
08
Existing System / Methodology
Used
Overview
Existing dynamic pricing systems predominantly rely on machine learning
models and real-time data for price optimization. These are widely used by
large e-commerce platforms and retail chains.
Algorithm used
• Linear & Logistic Regression
• Time-Series Forecasting
• Deep Learning Models
• Reinforcement Learning
• Rule-based Heuristics (for basic applications)

08
Existing System / Methodology
Used
Performance Metrics Considered
• Revenue Growth
• Price Elasticity
• Inventory Turnover
• Conversion Rate
• Customer Retention

Limitations
• High data volume and infrastructure requirements
• Limited adaptability for small retailers
• Lack of interpretability in complex models
• Real-time competitor tracking is resource-heavy

9
Proposed System / Methodology
Module Diagram:

10
Proposed System / Methodology
Module Description
Data Input Module
Collects basic data such as:
•Inventory levels

•Recent sales data

•Optional competitor prices

•External triggers (e.g., season, events)

Rule-Based Engine
Applies predefined business rules for pricing:
•E.g., If stock < threshold → increase price

•Ensures quick, logic-based filtering

Bayesian Optimization Module

Fine-tunes prices using probabilistic search:
•Optimizes revenue, stock turnover

•Works effectively with minimal data

11
Proposed System / Methodology
Module Description (Contd.)

Pricing Decision Engine

Combines rule output + optimized value:
•Selects best price within business constraints

•Adjusts based on past feedback

User Interface / Integration

•Google Sheets / Firebase interface

•Simple UI for non-technical users

•Low-code deployment friendly

12
 Proposed System / Methodology
Pseudo Code
# Step 1: Data Collection
data = collect_data(sources=["database", "csv", "logs", "docs"])

# Step 2: Data Preprocessing

cleaned_data = preprocess_data(data)
features = extract_features(cleaned_data)

# Step 3: Rule-Based Filtering

rule_based_output = apply_rules(features)
# Example: price range constraints based on stock, demand, category

# Step 4: Bayesian Optimization Setup

def objective_function(price):
return simulate_revenue(price, features)
# Define search bounds using rule-based output
price_bounds = get_bounds(rule_based_output)

13
Proposed System / Methodology
Pseudo Code (Contd.)
# Step 5: Bayesian Optimization
optimal_price = bayesian_optimize(objective_function, bounds=price_bounds)

# Step 6: Final Hybrid Price Decision

hybrid_price = finalize_price(rule_based_output, optimal_price)

# Step 7: Deployment
send_to_interface(hybrid_price)

# Step 8: Feedback Loop

feedback_data = collect_feedback()
update_model(feedback_data)

14
Proposed System / Methodology
Performance Metrics Considered
📈 Revenue Growth
Measures increase in overall sales income after applying dynamic
pricing.
🔄 Inventory Turnover Rate
Evaluates how quickly stock is sold and replaced over a time period.
💰 Profit Margin
Assesses the difference between cost price and dynamically
adjusted selling price.
⚖️Price Competitiveness
Comparison of proposed prices with competitor prices for similar
products.
🤝 Customer Retention (Optional) verbally discussed
Repeat purchase rate in response to fair and optimized pricing.

15
Proposed System / Methodology
Tools / Software Required
💻 Programming & Modeling
•Python – core implementation language

•Google Colab – development & testing

•Scikit-learn – rule-based logic & ML utilities

•Optuna / GPyOpt – Bayesian Optimization modules

📊 Data Handling
•Google Sheets – user-friendly input/output interface

•Firebase Realtime Database – for real-time deployment simulation

🔧 Deployment & Automation

•FastAPI – lightweight API creation

•Zapier – for low-code process automation

•Node.js – integration or UI-based customization

🧪 Simulation & Testing

•Simulated retail data – custom CSV/Sheet inputs

16
System Implementation
Experimental Setup
Environment:
Development Tools: Python, Google Sheets, Firebase, FastAPI
Libraries: Scikit-learn, Optuna, GPyOpt (for Bayesian Optimization)
Platform: Google Colab for testing and simulation

⚙️System Configuration:
Developer Machine:
•Intel i5 Quad-Core Processor
•16 GB RAM
•Windows 10
Server Setup (Simulated):
•8-Core CPU, 32 GB RAM
•Windows Server Environment

17
System Implementation
Experimental Setup (Contd.)
🛒 Dataset:
Simulated inventory and sales data for 3 retail shops (A, B, C)
Product types: Fashion and seasonal items
Input Variables: Stock level, price, sales history, competitor price

🧪 Testing Scenario:
Simulated real-world events:
Sales events, stock clearances, demand surges
Comparison between:
Static Pricing vs Hybrid Dynamic Pricing (Proposed)

📊 Tools Used:
Firebase Realtime DB – for deployment simulation
Google Sheets – for manual input & visibility
Low-code platforms Zapier – for flow automation
18
System Implementation
Input & Output Values
Customer
Stock Price Satisfactio
Price Revenue Overstock
Shop Product Turnove Competitiveness n (1-5)
(INR) (INR) Rate (%)
r Ratio (%)

Shop_A Jean 1355.13 95308.52 3.03 17.05 86.50 4.57

Shop_A Shirt 924.73 103836.14 3.14 17.49 80.83 3.15
Shop_A T-shirt 1480.23 105739.67 4.43 24.96 87.79 4.32
Shop_B Jean 1423.56 110567.88 4.10 18.23 82.74 4.23
Shop_B Shirt 1005.76 119834.21 3.88 16.67 75.50 3.98
Shop_B T-shirt 1556.93 124678.99 4.90 20.14 88.34 4.49
Shop_C Jean 1308.47 98562.33 3.65 19.42 84.61 4.12
Shop_C Shirt 975.43 101258.92 4.01 20.34 79.60 3.74
Shop_C T-shirt 1467.34 115490.22 4.20 25.65 85.32 4.27

19
System Implementation
Performance Metrics values Measured

Performance Metric Static Pricing Dynamic Pricing

Monthly Revenue (INR) $100,000 $115,000

Overstock Rate (%) 25% 20%

Stock Turnover Ratio 4 4.5

Price Competitiveness (%) 60% 85%

Intervention 8 manual adjustments 2 automated actions

20
System Implementation

21
Results Discussion
Comparison Table of Existing and Proposed System
Performance Static Dynamic
Metric Pricing Pricing Improvement Improvement %

Monthly
Revenue (INR) $100,000 $115,000 15000 15%
Overstock
Rate (%) 25% 20% -5% 5%

Stock
Turnover Ratio 4 4.5 0.5 13%

Price
Competitivene
ss (%) 60% 85% 25% 25%

8 manual 2 automated
Intervention adjustments actions -6 interventions 25%

22
Results Discussion

Summary
The results of the proposed hybrid dynamic pricing system demonstrate
significant improvements over traditional static pricing. Retailers
implementing the model experienced up to a 15% increase in revenue, with a
notable 13% improvement in inventory turnover, effectively reducing both
overstock and stockout scenarios. The integration of Bayesian Optimization
fine-tuned pricing decisions, contributing to a 5% decrease in over stock.
Moreover, the system-maintained price competitiveness within 10–25% of
external market benchmarks, enhancing responsiveness to competitor actions.
The use of rule-based logic reduced the reliance on manual interventions,
enabling smarter decision-making with minimal human effort. Simulated trials
across different retail scenarios—such as Shops A, B, and C—showed
consistent outperformance of the hybrid model, especially during demand-
sensitive periods like sales events and festive seasons.
23
Results Discussion
Sample Snapshot of Output

24
Conclusion & Future Enhancement
Conclusion
The "Cogni Price" model is a lightweight, hybrid dynamic pricing
system tailored for small-scale retailers. It blends rule-based logic with
Bayesian optimization to offer intelligent, adaptive pricing without the need
for extensive data or complex infrastructure. Built using low-code tools like
Google Sheets and Firebase, it allows retailers to automate price updates
based on demand, inventory, and market conditions. The model
demonstrated improvements such as a 15% revenue increase, 5% reduction
overstock, improved price competitiveness by 25% and higher customer
satisfaction for stimulated datasets for the shops. Its transparent, user-
friendly design encourages trust and control, making AI-driven pricing
accessible and practical for SMEs.

25
Conclusion & Future Enhancement
Future Enhancement

Future enhancements for the proposed system aim to expand its

adaptability and intelligence. One key improvement involves integrating
real-time data flows using Firebase, allowing instant price updates as
inventory or demand changes. Additionally, the system can be enhanced
with customer segmentation features, enabling personalized pricing
based on demographic or behavioral data. To further improve
competitiveness, browser extensions or API-based tools can be
incorporated to track live competitor pricing. Another promising
direction is the use of IoT-enabled inventory systems, such as smart
shelves, to automate stock updates and trigger pricing decisions. Lastly,
interactive dashboards using platforms like Google Data Studio or
Power BI can offer intuitive visualization of pricing insights and
performance metrics, supporting informed decision-making.
26
References
List of Journals referred
1. A. Sharma, N. Patel, and R. Gupta, “Leveraging Reinforcement Learning and Bayesian Optimization for
Dynamic Pricing Strategies in E-commerce,” Eur. Adv. AI J., vol. 11, no. 8, pp. 45–58, Jan. 2022.
[Online]. https://www.eaaij.com/index.php/eaaij/article/view/36
2. A. K. Kalusivalingam, A. Sharma, N. Patel, and V. Singh, “Leveraging Reinforcement Learning and
Bayesian Optimization for Enhanced Dynamic Pricing Strategies,” Int. J. AI ML, vol. 1, no. 3, pp. 12–25,
Apr. 2020. [Online]. Available: https://www.cognitivecomputingjournal.com/index.php/IJAIML-
V1/article/view/46
3. J. Johnson and K. Lee, “Adaptive Pricing Strategies for Small Retailers,” J. Retail Anal., vol. 12, no. 3, pp.
115–124, 2019.
4. M. Davis, S. Clark, and P. Adams, “Simplified Dynamic Pricing for Small Businesses,” Small Bus. Econ.
Rev., vol. 8, no. 2, pp. 44–51, 2017.
5. T. O’Connor and A. Singh, “Dynamic Pricing in the Context of Limited Data,” Retail Syst. Strateg., vol.
11, no. 1, pp. 88–96, 2019.
6. R. Brown and M. Davis, “Cost-Effective Dynamic Pricing Solutions,” Int. J. Retail Manag., vol. 13, no. 2,
pp. 152–160, 2021.
7. Y. Li and X. Zhang, “Competitive Pricing Strategies in Retail Markets,” Retail Pricing Intell., vol. 15, no.
1, pp. 45–55, 2022.
8. R. Garcia and P. Kumar, “Inventory Management through Dynamic Pricing,” Oper. Supply Chain J., vol.
14, no. 3, pp. 104–112, 2021.
9. L. Martinez, A. Roberts, and F. Cho, “Customer-Centric Dynamic Pricing Models,” J. Consum. Pricing
Strateg., vol. 9, no. 2, pp. 60–71, 2019
10. S. Ahmed, K. Das, and M. Roy, “Dynamic Pricing in Emerging Markets,” Emerg. Mark. Bus. Rev., vol. 27
7,
no. 1, pp. 35–46, 2017.
References
List of Journals referred
11. D. Hernandez, M. Carter, and J. Fox, “Evaluating the Effectiveness of Dynamic Pricing Models,”
Retail Decis. Sci., vol. 10, no. 4, pp. 210– 219, 2020.
12. A. Patel and S. Mehta, “Assessing the Risks of Dynamic Pricing,” J. Pricing Strateg. Ethics, vol. 16,
no. 1, pp. 75–84, 2022.
13. J. Smith, L. Wang, and H. Evans, “Dynamic Pricing in Retail: A Review of Algorithms and
Applications,” Artif. Intell. Retail, vol. 11, no. 1, pp. 10–22, 2018.
14. C. Chen, Y. Lin, and T. Zhao, “Real-Time Pricing Models in E- commerce,” E-commerce Technol.
Innov., vol. 13, no. 1, pp. 66–75, 2020.
15. T. Nguyen and R. Patel, “Impact of Dynamic Pricing on Customer Behaviour,” J. Retail Psychol.,
vol. 12, no. 2, pp. 98–108, 2018.
16. H. Lee and M. Kim, “Integrating Inventory Management with Pricing Strategies,” Oper. Manag.
Rev., vol. 10, no. 3, pp. 130–140, 2018.
17. M. Singh and Q. Zhao, “Real-Time Market Analysis for Pricing Decisions,” Retail Anal. J., vol. 13,
no. 3, pp. 155–163, 2020.
18. T. Khan and S. Roy, “Dynamic Pricing for Perishable Goods,” J. Perishable Prod. Manag., vol. 11,
no. 2, pp. 120–128, 2021.
19. H. Choi and J. Park, “Evaluating Customer Response to Dynamic Pricing,” Pricing Consum. Behav.
J., vol. 10, no. 4, pp. 190–200, 2018.
20. D. Wilson and F. Thompson, “Dynamic Pricing for Seasonal Products,” J. Seasonal Retailing, vol.
12, no. 1, pp. 80–90, 2019.

28
List of Publications
List of paper published / Presented in Conference

International Journal

Prathiba T, Dr. Singaravel G (2025) “A Data Driven Strategy For Dynamic

Pricing: Using AI to Address Inventory and Competition Issues in E-
Commerce” (Submitted)

29
List of Publications
Proof of Publication

30
32
Thank
you
33