LAVI.SAHU/OPS · FEED: SYNTHETIC
UPTIME 14Y · STATUS: ACCEPTING WORK
▸ OPERATOR PROFILECLEARANCE: PUBLIC

// SUPPLY-CHAIN PLANNING & TECHNOLOGY · 14 YEARS ON SHIFT

Planning is an operations problem.

Lavi Sahu. Fourteen years across SAP IBP, o9, Kinaxis, OMP and S&OP — pharma cold chain, FMCG, energy. I build the analysis behind planning decisions: resilience, demand quality, integrated business planning. Reproducible, from first principles. Increasingly: applied AI in planning workflows.

14
Years planning
5
Platforms
3
Industries deep
20+
Programs delivered
▸ DEMAND SIGNALSYNTHETIC
smooth61%
erratic22%
lumpy12%
intermit5%

ADI / CV² segmentation · Syntetos–Boylan–Croston cutoffs

▸ FVA MONITORSYNTHETIC
forecast actual
+6.2FVA pts vs naïve
18.4%MAPE
−1.2%Bias
▸ WORK/01 · RESILIENCETTR/TTS

Supply Chain Resilience

Which node, if it fails, hurts most — and for how long?

MethodTime-to-Recover vs Time-to-Survive stress testing, in the Simchi-Levi tradition. Fail every node on schedule; measure exposure where TTS < TTR.
OutputA ranked map of structural fragility — which failures the network absorbs, and which it can't.
github.com/LaviSahu/supply-chain-resilience
▸ WORK/02 · DEMAND DXFVA · ADI/CV²

Demand Planning Diagnostics

Is our forecast actually adding value — and where is it worst?

MethodForecast Value Added (Gilliland) against a naïve baseline, crossed with ADI/CV² demand segmentation (Syntetos–Boylan–Croston).
OutputWhere planner touches help, where they hurt, and which demand classes deserve which method. Every number computed, not asserted.
github.com/LaviSahu/demand-planning-diagnostics
▸ WORK/03 · S&OPRCCP

S&OP Integrated Planning

Base vs upside vs constrained — what do we commit, and what does it cost?

MethodS&OP / IBP scenario reconciliation with rough-cut capacity checks, in the Wallace / Oliver Wight tradition.
OutputOne commit number per cycle, with the cost of each alternative stated — the plan as a decision, not a spreadsheet.
github.com/LaviSahu/sop-integrated-planning
▸ EVENT LOGSTREAMING · SYNTHETIC
    ALL EVENTS SIMULATED · DETERMINISTIC SEED 0x00000000
    ▸ NODE STATUSSTRESS HARNESS
    NodeTTRTTSState
    NAG-014d11dOK
    CHE-026d9dOK
    BLR-DC11d8dEXPOSED
    HYD-037d7dWATCH
    PNQ-XD3d12dOK

    Rule: TTS < TTR → the network runs out before it recovers. That gap is the finding.

    ▸ STACKPKG
    pkg list --scope=planning
    sap-ibp▮▮▮▮▮▮▮▮▮▮core
    o9▮▮▮▮▮▮▮▮▮▮scenario
    kinaxis▮▮▮▮▮▮▮▮▮▮concurrent
    omp▮▮▮▮▮▮▮▮▮▮ops
    sap-apo▮▮▮▮▮▮▮▮▮▮legacy

    bars = depth of use, not years

    ▸ RUNBOOK · PRINCIPLESREV 14
    1. Judgment over code craft. The model serves the decision, never the reverse.
    2. Bounded claims. Say what the analysis cannot see.
    3. Seed everything. Same inputs, same answer — reproducible or it didn't happen.
    4. Every number computed, not asserted.
    5. Synthetic data labelled SYNTHETIC. Including every feed on this page.
    6. The forecast is a decision, not a prophecy.
    ▸ CONTACT · CHANNEL OPENRESPONSE < 48H · IST