For information and reference only – not to be submitted as your own work. This is a model-answer extract showing how a distinction-level LOG309 End-of-Course Assessment can be structured; the prompt, the AI output and every figure shown are illustrative.
LOG309 Industry 4.0 Logistics Applications is a Singapore University of Social Sciences (SUSS) course on connected supply chains, data analytics and smart warehousing. The July 2026 End-of-Course Assessment (ECA) is split into a 65-mark written section, a 3 to 6 minute video presentation and a slide deck of up to 15 slides. This extract covers Section A in full outline and then gives a planning guide for Sections B and C.
| Part | Marks | Approx. length | What the extract below shows |
|---|---|---|---|
| Q1(a) Technology, prompt, output, analysis | 20 | Prompt up to 50 words; output up to 200; analysis about 200 | AI-driven demand sensing, with a labelled illustrative prompt and output |
| Q1(b) Prompt reflection | 5 | About 50 words | Three specific prompt fixes |
| Q2 API mini design document | 20 | About 200 words plus table | A hypothetical landed-cost risk signal API |
| Q3 Constrained Industry 4.0 proposal | 20 | About 200 words | A digital supply chain twin under two constraints |
| Section B video / Section C slides | 25 / 10 | 3–6 minutes / up to 15 slides | Timed plan and slide-by-slide outline |
Scenario read-through
The company supplies products with unpredictable demand. Supplier disruption, tariff changes and maritime transit risk produce shortages and push up total landed cost. Notice that the scenario contains two different failures: a demand-side failure (the forecast is wrong) and a supply-and-cost-side failure (the inbound plan breaks). A distinction answer keeps these separate, because the weakness found in Q1 should expose exactly this gap.
Question 1(a): Technology, GenAI prompt and analysis
Technology selected
AI-driven demand sensing: machine learning models that refresh short-horizon forecasts from recent orders, point-of-sale data, promotions and external signals, then feed revised numbers into replenishment planning.
Illustrative prompt (47 words)
Act as a supply chain analyst. A Singapore distributor faces unpredictable demand, supplier disruptions, tariff changes and maritime delays that cause stockouts and higher landed cost. Recommend one Industry 4.0 technology, explain how it works, give one strength, one weakness, and one data requirement. Maximum 200 words.
Illustrative GenAI output (136 words)
Illustrative only. In your own submission, run your own prompt, paste the tool’s real full output and state the tool and date.
Technology: AI-driven demand sensing. It applies machine learning to recent orders, point-of-sale data, promotions and external signals such as events and port congestion, updating short-horizon forecasts daily instead of monthly. Replenishment and safety stock then respond to current conditions.
How it works: Models learn demand patterns by product and location, detect sudden shifts and push revised forecasts into inventory planning, which adjusts reorder points and purchase timing.
Strength: It improves short-term accuracy for volatile items, reducing stockouts and avoiding excess safety stock that raises holding cost.
Weakness: It explains demand, not supply or cost. It cannot natively see supplier delays, tariff changes or vessel schedules, so a forecast can be accurate while the replenishment plan still fails.
Data requirement: Twelve to twenty-four months of clean, SKU-level sales history with consistent product codes, plus near-real-time order feeds.
Analysis (strength and weakness)
Demand sensing addresses the first half of the scenario directly. Short-horizon models react to order patterns days rather than weeks earlier, which is the situation where unpredictable demand hurts most. Evidence from large retail forecasting contests supports the direction: the M5 competition found that machine-learning methods using related series and external information outperformed standard statistical benchmarks (Makridakis et al., 2022). The strength is therefore lower forecast error on volatile items, translating into fewer stockouts without buffering every SKU.
The weakness is scope. The output correctly identifies that demand sensing sees demand, not supply. Supplier disruption, tariff changes and maritime delays alter when stock arrives and what it costs on arrival, so an accurate forecast still meets an inbound plan that is stale. Ivanov et al. (2019) describe how disruptions propagate as a ripple effect, which a demand-only model cannot anticipate. I would also flag one accuracy check on the GenAI output: it names port congestion as a signal without saying where that data comes from, so I adapted that point in Q2.
Hire a Professional Essay & Assignment Writer for completing your Academic Assessments
Native Singapore Writers Team
- 100% Plagiarism-Free Essay
- Highest Satisfaction Rate
- Free Revision
- On-Time Delivery
Question 1(b): Improving the prompt
The prompt gave the model a role, the context and a word cap, but it left the company’s size, product type and lead times undefined, so the answer stayed generic. I would add those facts, require the model to state its assumptions, ask for a table comparing two technologies before choosing one, and request one citation-ready source. Specific inputs would make the output testable rather than plausible.
Question 2: Hypothetical API mini design document
The Q1 weakness is that demand sensing cannot see supply-side and cost-side shocks. The proposed fix is a hypothetical Landed Cost Risk Signal API that sits beside the demand-sensing engine and returns a risk-adjusted inbound view for each SKU.
| Element | Design (hypothetical) |
|---|---|
| Endpoint | POST /v1/landed-cost-risk, returning JSON |
| Inputs | SKU, origin, port pair, order quantity, HS code, Incoterm; supplier on-time-in-full history from the ERP; booking and vessel schedule data from the transport management system; current tariff table from a customs data feed; the latest demand-sensing forecast |
| Processing | Validate and standardise codes; estimate a delay distribution from vessel schedule history; recalculate duty and freight per scenario; compare projected arrival with sensed demand to compute stockout probability and landed cost per unit |
| Outputs | Stockout probability by week; expected and worst-case landed cost per unit; recommended action (expedite, reroute, split order, hold buffer); confidence score |
| Integration | Reads from ERP and TMS through scheduled pulls; writes to the planning module and a control tower dashboard; returns revised reorder suggestions to the demand-sensing engine |
| Controls | Token authentication and role-based access; request logging; schema validation; rate limits; human approval above a cost threshold; model drift monitoring; versioned tariff data |
The link to Q1 is explicit: demand sensing supplies the sensed demand curve, while the API answers the question that curve cannot, namely whether stock will arrive in time and at what total landed cost. Without the forecast input, the API would not know what a delay is worth. Without the API, the forecast has no supply context. An interface specification such as OpenAPI 3.1.0 (OpenAPI Initiative, 2021) would document the request and response formats so planners and developers share one contract.
Question 3: Industry 4.0 technology under two constraints
Proposal: a digital supply chain twin, a virtual model of the inbound network that simulates shortage and landed-cost outcomes before orders are placed. It uses the GenAI output’s recommendation to pair demand sensing with supply visibility, and the twin is where that pairing is tested.
Constraint 1: supplier data quality issues. The twin does not assume clean feeds. Each supplier record receives a data-quality score based on completeness, timeliness and past accuracy; low-scoring inputs are down-weighted and replaced by category-level averages. Planners see a confidence band rather than a single number.
Constraint 2: unpredictable maritime delays. Rather than a fixed transit time, the twin runs probabilistic arrival scenarios built from historical voyage variation by lane. It reports a service-level range, for example the share of simulations in which stock arrives before the sensed stockout date.
Why the output is still useful. The twin ranks actions by expected shortage reduction and landed-cost impact, not by exact prediction. If one supplier’s data is poor and a port is delayed, it can still show that splitting an order or holding a small buffer on the highest-margin SKUs gives the most robust result (Ivanov et al., 2019). Where confidence is low, it says so, and the planner decides.
Section B: video plan (3–6 minutes, 25 marks)
| Time | Content | Marks it supports |
|---|---|---|
| 0:00–0:30 | Greeting, scenario and the plan for the talk | Organisation (8) |
| 0:30–2:00 | Q1 technology, strength and weakness | Organisation, pace and articulation (7) |
| 2:00–3:30 | Q2 API: show the design table, explain the link to Q1 | Eye contact (5), organisation |
| 3:30–4:45 | Q3 twin and the two constraints | Posture and body language (5) |
| 4:45–5:30 | Summary, limits, one closing line | Pace and articulation |
Aim for about 5 minutes so a slow start does not breach the 6-minute limit. Speak to the camera lens, keep notes to key words and record a full take before you edit.
Section C: slide-by-slide outline (up to 15 slides, 10 marks)
| Slide | Content |
|---|---|
| 1 | Title, course code, your details as required |
| 2 | Scenario and the two failure types |
| 3 | Technology: AI-driven demand sensing |
| 4 | GenAI prompt and word count |
| 5 | GenAI output summary and what you adapted |
| 6 | Strength and weakness |
| 7 | Prompt improvement |
| 8 | API purpose and flow diagram drawn by you |
| 9 | API design table |
| 10 | Controls and integration |
| 11 | Q3 digital twin |
| 12 | Two constraints and the adapted output |
| 13 | Conclusion |
| 14 | GenAI attribution |
| 15 | References in APA |
Use your own layout, not an internet template, and keep each slide to one message. Diagrams you draw yourself earn the creativity marks.
GenAI attribution (as the brief requires)
State the tool and version, the exact prompt with word count, the full output, and what you changed. In this extract the change was the Q2 source for congestion data, because the output did not name one. APA 7th guidance treats the developer as author, for example OpenAI (2023); check the current APA Style page for the latest wording (McAdoo, 2025).
For the full task wording on the platform, see the SUSS assignment questions hub (the LOG309 ECA question page will be linked when live) and the related LOG309 group-based assignment question. A step-by-step guide is in how to approach the LOG309 ECA. For related reference extracts, see LOG307 optimisation and simulation and LOG321 trade regulations. Support pages: SUSS assignment help and information technology assignment help.
If your ECA is due on 30 October and you want an expert to check your prompt, API table or slides, submit your assignment details for a quote.
About the expert
This subject area falls within the specialisations of Wei Chen, who holds a Ph.D. in Computer Science from the University of Hong Kong and has over 10 years in the field, with listed subjects including artificial intelligence, machine learning, data mining and natural language processing.
References
Ivanov, D., Dolgui, A., & Sokolov, B. (2019). The impact of digital technology and Industry 4.0 on the ripple effect and supply chain risk analytics. International Journal of Production Research, 57(3), 829–846. https://doi.org/10.1080/00207543.2018.1488086
Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2022). M5 accuracy competition: Results, findings, and conclusions. International Journal of Forecasting, 38(4), 1346–1364. https://doi.org/10.1016/j.ijforecast.2021.11.013
McAdoo, T. (2025, September 15). How to cite ChatGPT. APA Style. https://apastyle.apa.org/blog/how-to-cite-chatgpt
OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model]. https://chat.openai.com/chat
OpenAPI Initiative. (2021). OpenAPI Specification v3.1.0. https://spec.openapis.org/oas/v3.1.0.html
