Unit 3 - Methodology and Research Methods

For my Proposal Presentation and MSc capstone, I have chosen the topic "Integrating Demand Forecasting and Personalised Product Recommendation in Nutraceutical E-Commerce: A Machine Learning Approach Using Multichannel Behavioural and Sales Data". This project is directly aligned with my professional role as a web developer in a nutraceutical company, offering mutual value by addressing real-world marketing and inventory challenges through academic research.

The proposed methodology follows a Design Science Research (DSR) approach, which is well-suited for artefact creation, iterative model development, and practical evaluation. The research will proceed through four key phases:

1. Data Engineering – Integration and cleaning of multichannel data, including sales (Shopify), user behaviour (GA4), and advertising metrics (Google and Meta Ads).
2. Model Development – Implementation of forecasting models (ARIMA, Prophet, LSTM) and recommendation algorithms (Matrix Factorisation, Neural Collaborative Filtering, Hybrid models).
3. System Integration – Forecast outputs will influence recommendation logic, ensuring relevance while avoiding out-of-stock or low-demand SKUs.
4. Evaluation – Using quantitative metrics (RMSE, MAE, precision@k, NDCG) and qualitative feedback through simulated user testing with Likert-scale questionnaires.

Data collection will centre on secondary system data from existing platforms, supplemented by small-scale user testing to assess perceived recommendation relevance and trust.

To execute this project effectively, I will need to enhance skills in time-series modelling, recommender systems, data engineering, and evaluation methodologies, alongside ensuring ethical data handling under GDPR standards.