Landing process · five steps
1Data preparationImport ~425 days of daily sales history, validate channel/category/SKU mapping
2Target entryEnter the annual budget top-down with unified standards
3Smart splitOne-click split to SKU × month by structural weights, generating a draft
4Decide in-meetingProject the simulation sheet in the planning meeting; parameters recalculate live, key items fine-tuned
5Snapshot archiveDraft→Active auto-generates version snapshots, with rollback and comparison
Roles involved
BuyerPlanning leadFinance
Key deliverables
- SKU × month sales plan (Rolling 15~18M)
- Version snapshots and split-weight ledger
- Budget variance baseline for later closed-loop comparison
Balanced note: efficiency gains depend on data maturity and organizational cooperation — validate with a POC first, then roll out gradually.
Landing process · five steps
1New-item mappingBuyers confirm Parent Code mapping, establishing old-new series relationships
2Baseline estimateNew items anchor to similar-item baselines for the first order, converging with actual sales
3Multiplier overridePromo calendar flagged; manual multipliers require reasons, fully logged
4Peak removalAbnormal promo peaks auto-flagged and stripped when computing next year's baseline
5Post-promo reviewForecast vs. actual variance compared, multiplier experience retained as reference
Roles involved
BuyerCategory lead
Key deliverables
- New-item first-order estimates and cannibalization quantification
- Promo multiplier intervention ledger (who/when/why)
- Post-promo variance review report
Balanced note: multipliers are human judgment; the system's value is making judgment traceable and reviewable, building it into an experience asset.
Landing process · five steps
1Snapshot importDaily/weekly inventory snapshots imported; the ledger records only physical changes
2Heatmap diagnosisRed/blue/green zones update automatically, showing stockout and surplus risks
3Expiry viewSeparates normal consumption from expiry write-offs; high-risk batches conservatively estimated
4DRP computationNet-demand formulas run automatically, generating suggested replenishment
5Execution feedbackOrders to HQ after confirmation; actuals flow back and variance feeds the next round
Roles involved
Inventory planningBuyer
Key deliverables
- Inventory heatmap and expiry write-off alert list
- Suggested replenishment (net purchase suggestions)
- Weekly meeting decision dashboard
Balanced note: the system suggests, people decide — whether and how much to replenish is confirmed by the business team.
Solution modules
Planning · Budget managementTarget derivation
How it solves it
- Annual budgets are managed by version number after entry, with clear active/archived status
- Channel allocation and version comparison, differences shown intuitively
- Target derivation supports AI re-derivation by product line/region, taking effect after batch confirmation
- Version history auto-retained; historical snapshots reviewable
Demo example: BV-2026-01 active at ¥25.00M; BV-2026-00 archived at ¥17.50M.
Balanced note: split weights and target confirmation remain human decisions; the system provides derivation suggestions and comparison tools.
Solution modules
Smart optimization · Forecast engineTrend modeling
How it solves it
- Multi-algorithm forecast tasks (ARIMA/Prophet/LightGBM/Ensemble) with visible progress and write-back status
- Trend types capture business rhythms (e.g., 618 + Double-11 twin peaks, back-to-school minor peak); weekly daily composition configurable
- Anomaly cleaning rules configurable: stockouts/promos cleaned, new items/near-expiry keep original values
- Model versions and fallback mechanism (3σ + moving average) transparently visible
Demo example: demo model version forecast-v2.1, MAPE ~22% (sample value).
Balanced note: model accuracy depends on data quality and business changes; regular review and retraining are required.
Solution modules
Planning · Sales plansWorkflow management
How it solves it
- Each channel plans independently, flowing draft→submitted→active
- Dual views for multi-channel aggregation: quantity rollup (SKU × channel × month) and amount vs budget (variance rate)
- Version history supports snapshot viewing and Excel export
- Submission approval runs through workflows with clear accountability
Demo example: SP-TM/SP-JD active; the demo holds 1 sales plan pending approval.
Balanced note: aggregation is a presentation and comparison tool; variance causes still need business interpretation.
Solution modules
Weekly simulation centerMaster-data SKUPromo configuration
How it solves it
- SKU management maintains PARENT codes and CARRY OVER flags, carrying old-new substitution relationships
- What-if scenario simulation: expected incremental volume/revenue and confidence quantified; apply after comparison, copyable and revertible
- Promo configuration defines promo patterns and effect parameters for simulation reference
Demo example: sample scenario "Double-11 Tmall G-Shock promo simulation" +35%, confidence 82%.
Balanced note: confidence is a model reference metric; whether to execute a scenario is a business decision.
Solution modules
Planning · Inventory planningHome overview
How it solves it
- After opening stock is imported, daily simulation (571 days in the demo) shows coverage weeks automatically
- Coverage heatmap in four bands: <1 month / 1~2 / 2~4 / >4 months, showing stockout and surplus zones
- Inventory snapshot management: warehouse/DC/store stock and safety-stock weeks in one table
- Batch expiry alerts concentrate high-risk batches on the home page
Demo example: the demo home page includes coverage heatmap and batch expiry alert cards.
Balanced note: the heatmap rests on snapshots and forecast assumptions; update snapshots regularly to keep it useful.
Solution modules
Buying engineBuying plan
How it solves it
- Transparent formulas: SS=Z×σ_d×√(LT+RT); net demand=forecast sales−available stock+safety stock; buy qty=MAX(net demand,MOQ)
- AI auto-generates Rolling buying plans; historical versions viewable/exportable
- Purchase inbound summary matrix supports supplier reconciliation and funding schedules
- Task statuses DRAFT/IN_REVIEW/APPROVED, approvals logged
Demo example: demo task amounts ¥2,156,000 / ¥9,760,000.
Balanced note: parameters like service level/lead time must be tuned to each company; results change with parameters.
Solution modules
Smart optimization · End-of-life optimizer
How it solves it
- AI identifies declining SKUs and lists end-of-life candidates
- Quantifies surplus units, estimated loss and recoverable amount
- Suggested actions: pause replenishment + promo discount / reduce replenishment + promo discount, executed after human confirmation
Demo example: 2 end-of-life candidates, 3,800 surplus units, estimated loss ¥66K, recoverable ¥43K.
Balanced note: recoverable amounts are model estimates; actuals depend on timing and market response.
Solution modules
Data integrationDimensional modelingData inspection
How it solves it
- Four channel types: ERP daily sales (API), WMS stock (KAFKA), PLM master data (DATABASE), e-commerce orders (WEBHOOK)
- Dimensional modeling presets PSI_WEEKLY/PSI_MONTHLY/FORECAST_M/BUDGET_FY/ACTUAL_DAILY models
- Quick upload via Excel templates eases the transition; no process changes initially
- Data inspection auto-checks consistency issues like orphan categories and stale references
Demo example: all 4 data sources in the demo are synchronized.
Balanced note: integration quality depends on both systems' openness and interface contracts; joint testing and acceptance are required.
Solution modules
Operation auditWorkflow managementPlatform audit
How it solves it
- Operation audit records: who/when/which module/what action/change summary
- Budget and purchase-plan approval flows clearly defined; todos/instances viewable
- Team members authorized by RBAC roles; the platform adds tenant-level audit and security events
Demo example: the demo holds audit records such as plan submission/approval.
Balanced note: audit is a record-keeping and accountability tool; it delivers value only when policies are enforced.
