Inventory Optimization

Segmentation-driven policy design + multi-objective safety-stock simulation + holdout backtest validation. The right way: classify first (ABC×XYZ), tune per-segment, validate on real history.

How to use this page

Step 1 — Segmentation. Classify every SKU into a 9-box (ABC by revenue × XYZ by demand CV). Each cell ships with a recommended fill-rate target (β).
Step 2 — Policy optimizer. Per-(SKU × DC) analytic (s, Q) + Monte Carlo with empirical demand bootstrap (real forecast residuals, no Normal assumption) and empiricallead-time bootstrap (real PO actuals).
Step 3 — Safety-stock frontier. Sweep 10 SS formulas against stochastic demand × stochastic lead time; pick the cheapest feasible Pareto point.
Step 4 — Backtest. Replay the last N weeks of real actuals against both stored and recommended policies — the ground-truth validation that catches the calibration errors the analytic flow misses.

ABC × XYZ — recommended policy per cell

ABC by cumulative revenue contribution (80 / 15 / 5). XYZ by demand CV (≤ 0.5 / ≤ 1.0 / > 1.0). Cell shows: # SKUs · revenue share · recommended fill-rate target (β). CZ candidates should be reviewed for delisting before optimizing.

	X — steady (CV ≤ 0.5)	Y — variable (CV 0.5–1.0)	Z — lumpy (CV > 1.0)
A — top 80% rev	AXβ 99% 7 SKUs 73.0% revenue Tight (s, Q), high service, forecast-driven, weekly review	AYβ —% 0 SKUs — Tight (s, Q), wider SS buffer for promo / Tet variance	AZβ —% 0 SKUs — Periodic review, big SS buffer, watch obsolescence
B — next 15%	BXβ 97% 2 SKUs 19.0% revenue Standard (s, Q), monthly review	BYβ —% 0 SKUs — Standard (s, Q) with promo overlay	BZβ —% 0 SKUs — Periodic, conservative SS
C — long tail	CXβ 92% 1 SKU 8.0% revenue Bulk orders, low review frequency, JIT-style	CYβ —% 0 SKUs — Min/max levels, low service target	CZβ —% 0 SKUs — Candidate for delisting before optimizing

X — steady (CV ≤ 0.5)

Y — variable (CV 0.5–1.0)

Z — lumpy (CV > 1.0)

A — top 80% rev

AXβ 99%

7 SKUs

73.0% revenue

Tight (s, Q), high service, forecast-driven, weekly review

AYβ —%

0 SKUs

—

Tight (s, Q), wider SS buffer for promo / Tet variance

AZβ —%

0 SKUs

—

Periodic review, big SS buffer, watch obsolescence

B — next 15%

BXβ 97%

2 SKUs

19.0% revenue

Standard (s, Q), monthly review

BYβ —%

0 SKUs

—

Standard (s, Q) with promo overlay

BZβ —%

0 SKUs

—

Periodic, conservative SS

C — long tail

CXβ 92%

1 SKU

8.0% revenue

Bulk orders, low review frequency, JIT-style

CYβ —%

0 SKUs

—

Min/max levels, low service target

CZβ —%

0 SKUs

—

Candidate for delisting before optimizing

10 SKUs classified · ABC by revenue, XYZ by weekly-demand CV. β = fill-rate target the policy optimizer should aim for in each cell.

SKU classification

Top 200 by revenue. Each SKU’s segment maps to a fill-rate target.

SKU	Segment	Weekly demand	CV	Revenue share	Cum. rev	β target
MCH-KOK-004 Kokomi 90 Chicken Noodle	AX	2,057.9	0.06	11.46%	11.5%	99%
MCH-NAM-002 Nam Ngu Fish Sauce 750ml	AX	1,955.2	0.05	10.88%	22.3%	99%
MCH-OMA-003 Omachi Spicy Beef Noodle	AX	1,898.1	0.05	10.57%	32.9%	99%
MCH-CHIN-009 Chinsu Chili Sauce 250g	AX	1,844.6	0.05	10.27%	43.2%	99%
MCH-NAM-010 Nam Ngu Premium Fish Sauce 500ml	AX	1,812.6	0.07	10.09%	53.3%	99%
MCH-WAK-006 Wake-Up 247 Energy Drink	AX	1,785.9	0.04	9.94%	63.2%	99%
MCH-COMB-008 ComBo Pho Bo 70g	AX	1,762.3	0.06	9.81%	73.0%	99%
MCH-VIN-005 Vinacafe 3-in-1 Original	BX	1,713.1	0.05	9.54%	82.6%	97%
MCH-CHIN-001 Chinsu Premium Soy Sauce 500ml	BX	1,694.1	0.06	9.43%	92.0%	97%
MCH-VIVA-007 Vivant Mineral Water 500ml	CX	1,439.9	0.05	8.02%	100.0%	92%