10. Future Work

Areas for further research

Automated Segmentation Updates: Develop a system to periodically re-run the segmentation process to capture shifts in customer behavior. Automating these updates could ensure that segmentation remains relevant and reduces the time required for manual analysis.
Investigate how external factors, such as economic shifts, competitors’ promotions, or social trends, influence customer purchasing behavior. This research could provide insights into how sensitive different customers segmetns are to market changes, which would influence our promotional campaigns to retain customers.
Investigate behavioral triggers such as time-of-day that influence purchasing decisions. This could involve studying whether specific times of the day or year lead to higher purchase volumes or engagement, helping the business optimize promotional timing.
Investigate how various combinations of marketing channels interact and influence overall customer behavior. This can help to identify combined interactions of channels that can lead to improved engagement and sales.
Conduct longitudinal studies to understand how consumer preferences and behaviors change over time, particularly in response to seasonal variations, economic shifts, or emerging trends.
Assess how external market factors, such as economic downturns, cultural shifts, or technological advancements, impact customer purchasing behavior and marketing effectiveness.

Investigate impacts of Macroeconomic Indicators on Inventory Optimization and how macroeconomic trends such as inflation and consumer confidence influence demand patterns could enhance the inventory model’s ability to adapt to larger market conditions. These insights could guide inventory adjustments based on expected changes in consumer behavior.
Other than price elasticity of demand, we can also look at the price elasticity of supply to determine an optimal price to maximise revenue. Doing so can help identify pricing strategies that maximise revenue while considering both consumer demand and supply constraints.
Delve into understanding how external factors, such as macroeconomic trends or regional disruptions, influence delivery delays and supplier performance. Incorporating external datasets, such as weather patterns or transportation schedules, could enrich the analysis and provide deeper insights into supply chain dynamics. Additionally, investigating customer behavior patterns, including the relationship between satisfaction scores and delivery times, might yield valuable findings to optimize customer experience.

Reality of Customer Reviews: Customer reviews may be inaccurate of a customer’s opinion of the product; they can be made by cynical people from our competitors or made by customers who are sarcastic. Facing such issues, we deploy methods to prevent these people from flooding the review section. We may:

Further modify our systems or develop one with sarcasm detection capabilities to identify and flag such reviews.
While we already have information on our customers, further implement bot detection techniques such as CAPTCHA to prevent our competitors from dropping bot-driven reviews.

After identifying the customer segments, we can test and optimize promotional campaigns for segments. This can be done by conducting A/B testing on targeted promotions for different customer segments and evaluating how other engagement metrics (conversion rate etc.), and not just churn rate, are affected. This approach would allow the business to better refine its strategies by seeing which promotions are most effective for each segment, leading to data-driven promotional optimizations. The details are in Recommendations section of the wiki

We can integrate real-time demand forecasting into inventory automation. Currently, reorder thresholds and restock levels are set manually from linear programming calculations. If new results from the demand forecasting and linear programming can constantly feed values for inventory level tracking, real-time analytics would allow dynamic updates to reorder points, safety stock levels, restocking amount and lead times, resulting in more agile inventory management.
Rather than a singular pricing model, we can implement separate pricing models for each category. This would allow each model to capture relationships between relevant features and the optimal price more accurately. This is because different product categories may have different price ranges despite having similar feature inputs.
By implementing advanced machine learning models, such as Neural Networks, to better capture complex relationships in the data. Feature engineering could also be improved by integrating geospatial data for more precise predictions of delivery delays. Finally, creating automated real-time dashboards to monitor supplier performance and order fulfillment rates would enable stakeholders to make data-driven decisions quickly and efficiently.

More Complex Customer Review Analysis: We have information about our customers, stored in our database. An expansion into customer review sentiment analysis would be to separate customers into different age groups and demographics, and determine which group leaves the greatest number of positive/negative reviews. This helps with designing customer retention strategies. As this project only focuses on a few categories, we can expand our analysis to more categories. This helps with expanding our platform and serving more customers.