Mismatch between supply and demand and difficulty in judging inventory cycles
Complete English presentation of the source project analysis, preserving the full narrative, tables, figures and evidence boundaries.
Customer Case · Pain Point 3
Mismatch between supply and demand and difficulty in judging inventory cycles
Establish Integrated Decision-Making of Physical Flow—Inventory—Curve—Profit
| Case Claim: Unit Inventory Risk-Adjusted Holding Return at Target Service Level |
|---|
Applicable Readers: Management, Trading, Operations, Risk Control, Finance, Legal/Compliance, and Data Teams
Executive Summary
This case constructs a complete solution from diagnosis to implementation for the 'mismatch between supply and demand and difficulty in judging inventory cycles.' The report uses anonymized representative oil and gas trade scenarios, focusing not on providing a single-point tool, but on incorporating business decisions, physical execution, risk capital, evidence chain, and management responsibility into the same closed loop.
| Core judgment: Establish an integrated decision-making process of physical flow—inventory—curves—profit; the North Star metric is: the unit inventory risk-adjusted holding return under the target service level. |
|---|
| Customer Business Profile | Case-based parameters |
|---|---|
| Representative business | Finished oil imports, coastal storage areas, and regional distribution |
| Planning cycle | Procurement decisions lead demand by 30–60 days |
| Key constraints | Tank capacity, vessel schedule, terminal window, minimum inventory, and financing limit |
| Common risks | High prices to replenish stock during peak season shortages, inventory accumulation and monthly margin losses during off-season |
Observable failure signals
The total inventory seems reasonable, but the available inventory, in-transit inventory, and quality-restricted inventory are not separated.
Only look at the official inventory weekly report, ignoring port congestion, refinery maintenance, and hidden flows
After the monthly spread turns from contango to backwardation, inventory replenishment continues according to the original pace, and holding profits rapidly deteriorate.
Project Goals
Without sacrificing trading compliance, control independence, and cash security, turn unexplainable losses into measurable, accountable, predictable, and controllable operational variables.
1. Professional Diagnostic Framework
The project starts from settled transactions and real business events, replaying contracts, prices, goods flow, inventory, documents, credit, cash, and final profit and loss according to a unified transaction number.
| Diagnostic Module | Professional testing | Output |
|---|---|---|
| Inventory availability | Distinguish between book, physical, available-for-sale, committed, in-transit, and restricted inventory | Avoid false safety stock |
| Physical equilibrium | Supply Import Transfer In − Demand − Export − Loss | Establish daily/weekly rolling balance |
| Inventory economics | Spot price, monthly spread, storage, capital, loss, and option value | Determine Hold/Release |
| Situational pressure | ETA, demand, refinery operation, policy and weather joint distribution | Quantify stockouts and inventory accumulation at the tail end |
Diagnostic methods
Select 20–30 settled transactions covering normal, abnormal, and loss scenarios.
Rebuild the event timeline, data lineage, and chain of responsibility from quotation to final settlement.
Determine whether the loss could have been avoided by assessing counterfactual scenarios, and calculate the control costs and benefits.
Distinguish between uncontrollable industry fluctuations, manageable risks, and preventable execution defects.
2. Representative Transaction Examples
The following amounts and indicators are professional case data, used to illustrate causal chains and management actions, and do not represent the audit facts of any specific client.
Representative Transactions: Key Changes from Business Expectations to Final Economic Outcomes
The source-document visual is presented here as an English-native analytical frame. The adjacent English narrative and tables preserve the full evidence and quantitative context.
| Project | Impact/Result | Professional explanation |
|---|---|---|
| Planned Contribution | +120 | According to peak season sales and normal turnover |
| Demand is lower than expected | −32 | Regional consumption and terminal pickups are weak |
| Two ships concentrated at the port | −18 | Container shortage and surcharge for transshipment |
| Worsening menstrual irregularity | −26 | Holding gains turn into negative carry |
| Funds/Warehousing/Loss | −21 | Turnover extended by 18 days |
| Ultimate economic profit | +23 | The return on capital has declined significantly |
| Example Insights: Surface problems are usually just the final manifestation; the real value leakage comes from data, processes, authorization, and economic metrics not being linked with business events. |
|---|
3. Root Cause Analysis
Demand forecasting replaces probability distribution with single-point sales planning
The inventory caliber does not deduct committed, quality-restricted, and unavailable quantities
Purchasing, shipping, storage area, and sales use different time granularities
Did not combine futures curve, storage, and financing to calculate inventory carry
Replenishment approval lacks a symmetrical comparison of stockout costs and inventory accumulation costs
Root cause structure
| Level | Question | Management consequences |
|---|---|---|
| Commercial Design | The quotation, terms, or combination logic do not cover all risks | Expected profits are inherently high |
| Execution Control | Events did not trigger tasks, recalculation, and upgrades | Loss accumulates during the process |
| Data system | Objects, versions, and responsibilities are not unified | Unable to see the real situation in time |
| Organizational Motivation | Disconnection between returns and risks, cash, and control | Erroneous behavior is repeatedly rewarded |
4. Solution: Five-layer closed-loop control
| Hierarchy | Core Competence |
|---|---|
| 1 Inventory Digital Twin | Unify inventory according to batch, tank number, quality, ownership, commitment, and availability |
| 2 Rolling Physical Balance | Integration of refinery operations, imports, AIS, storage area in and out, and customer pick-up |
| 3 Probability Prediction | Output P10/P50/P90 demand, ETA, and inventory path |
| 4 Inventory Optimization | Optimize replenishment under service level, tank capacity, cash, and curve constraints |
| 5 Event Closed Loop | Delays, quantity changes, refusals, and demand fluctuations automatically trigger recalculation |
Operating mechanism
Business events enter the unified data layer and retain the source, timestamp, and version.
Rules and models calculate economic impact, risk exposure, and disposal priority.
The responsible person receives the task and executes or escalates approval within the scope of authorization.
The results are written back to the profit, risk, cash, and evidence ledgers, forming review data.
| Governance Principles: The system is responsible for identification, calculation, recommendation, and record-keeping; business responsibility, independent review, and approval of major exceptions are still undertaken by clearly designated individuals. |
|---|
5. Implementation Roadmap and Governance
| Stage | Time | Key deliverables | Acceptance |
|---|---|---|---|
| January–February | Unified inventory and traffic/caliber | Account discrepancy <1% | |
| February to May | Establish a 13-week rolling balance | Weekly forecast rolling update | |
| May–September | Online Scenarios and Inventory Replenishment Optimization | Out-of-stock/overstock warning in advance | |
| September to December | AI ETA and Demand Coordination | Automatically generate interpretable suggestions |
Project Governance
| Character | Primary responsibility |
|---|---|
| Business Manager | Define business objectives, acceptance processes, and outcomes |
| Product/Data Manager | Unified objects, standards, interfaces, and quality SLA |
| Risk/Compliance/Legal | Define hard rules, limits, exceptions, and independent challenges |
| Operations/Finance | Confirm events, costs, cash, and final settlement |
| Management Committee | Resources, cross-departmental conflicts, and major exception decisions |
The first 90 days
Complete the risk/value dictionary, representative trade replay, and baseline measurement.
Select a high-frequency product and carry out parallel trial operations along two typical business paths.
First establish a manually operable control loop, then gradually automate it.
Review anomalies, false reports, missed reports, user adoption, and actual value every two weeks.
6. Outcome Indicators and Business Value
The target range should be calibrated based on client size, product liquidity, jurisdiction, and risk tolerance; the table below is used for pilot run acceptance design.
| Indicator | Before Implementation / Baseline | 12-month goal |
|---|---|---|
| Demand MAPE | 28% | ≤15% |
| Sellable Inventory Accuracy | 88% | ≥98% |
| Inventory turnover days | 42 days | 30–34 days |
| Number of emergency restocks per month | 6 | ≤2 |
| Tank capacity utilization peak | 98% | ≤90% |
| Polaris metric: Unit inventory risk-adjusted holding return under target service level |
|---|
Value realization logic
Direct value: reducing losses, fines, discounts, capital occupation, or execution leaks.
Risk value: Reduce tail losses and the probability of major disruptions.
Efficiency Value: Shorten the cycles of quoting, reviewing, investigating, reconciling, and closing accounts.
Capability value: Transform personal experience into reusable data, rules, and organizational processes.
7. AI Evolution and Control Boundaries
AI Applicability
Predict ship ETA, port queuing, and customer pickup probability
Integrating price curves and physical flows to identify inventory turning points
Simulation of procurement, resale, exchange, ship extension, and cross-warehouse transfer plans
Explain whether the forecast changes come from demand, arrivals, quality, or the curve
Control boundaries that must be retained
The optimization objective must simultaneously constrain both service level and cash.
Using manually set safety stock thresholds during abnormal periods
The model distinguishes between observable facts and inferred flows
Key purchases are still approved according to the authorization matrix
| Stage | AI Character | Human responsibility |
|---|---|---|
| Data Assistant | Extraction, Association, Verification, and Summary | Confirm key facts |
| Monitoring and Prediction | Exceptions, Probability, and Scenarios | Determine business meaning |
| Program Collaboration | Compare actions, costs, and constraints | Approve and take responsibility |
| Closed-loop learning | Review results, update parameters | Governance Models and Rules |
| AI Principles: Traceable, Explainable, Stoppable, Auditable. Any recommendation must display the corresponding transaction, data source, assumptions, confidence level, residual risk, and failure conditions. |
|---|
8. Conclusion and Next Steps
Professional conclusion
| Final judgment: establish an integrated decision-making process of physical flow—inventory—curves—profit. When the 'unit inventory risk-adjusted holding return at the target service level' consistently enters the target range, and surface performance is not achieved by increasing invisible risks, it indicates that the capability can be replicated. |
|---|
Recommended next step
Conduct a 6-week diagnosis and complete a replay of 20–30 settled transactions.
Establish a baseline for value leakage, control gaps, data discrepancies, and a priority list.
Run a trial for 90 days with a single product/path, and expand only after verifying the metrics.
Incorporate final economic results, venture capital, and quality control into the continuous operation mechanism.
Caliber and Limitations
This report is prepared based on the logic of the aforementioned oil and gas trade cases, with clients, transactions, amounts, and indicators anonymized, case-based, or within target ranges. Formal implementation must be calibrated with actual contracts, accounting policies, risk limits, regulatory requirements, and professional opinions on local laws, taxation, and customs; this report does not constitute legal, tax, audit, or investment advice.