Skip to content

NewExplore decision agents and merchant-set guardrails

See how they work

AI-native retail planning · Forecast → allocation → order

Retail planning that decides.

PACX is an AI-native retail planning platform for demand forecasting, allocation and ordering, with connected recommendations, configurable constraints, staged approvals and auditable actions.

Book a demo
See the decision chain

Prefer async? Get the 6-minute walkthrough by email.

Forecast, allocation and ordering workflows · Configurable constraints and approval paths · Recommendations with recorded inputs

Illustration of the PACX planning workspace for one SKU: a 12-week demand forecast with its confidence band and drivers, a ledger of guardrail-checked decisions, and a store allocation heat strip — fictional data.

Customer references are published only with written approval

    The problem

    Trillions in inventory, steered by gut feel.

    Retailers still make their biggest financial decisions — what to buy, where to send it, what to charge, when to mark it down — in spreadsheets and planning suites designed twenty years ago.

    The result is the same every season: too much of the wrong thing, too little of the right thing, and margin burned in end-of-season markdowns.

    One retail season: planned receipts against actual demandDiagram, not real data. Actual demand runs above the planned receipts in the early weeks (stockouts), falls below the plan from mid-season (overstock), and the final weeks sit in a hatched markdown zone where margin is given away.markdown zone← where margin goesW1W4W8W12W16actual demandplanned receipts
    Illustrative season curve: stockouts early, overstock late, margin given away in the markdown weeks.

    One season, as it usually goes: planned receipts against actual demand — stockouts early, overstock late, and margin given away in the markdown weeks. Illustrative, not customer data.

    $1.7T

    lost to inventory distortion every year. IHL Group, 2026

    40%

    of a merchant's time goes to spreadsheet and data work. McKinsey, 2026

    71%

    of merchants say AI merchandising tools have had limited or no effect. McKinsey, 2026

    Sources last reviewed 22 Aug 2026

    The shift

    Not another disconnected dashboard.

    A forecast matters when a planner can carry it into a constrained allocation or order. PACX keeps the inputs, recommendation, review and audit trail connected so the next action can be inspected before it moves.

    1. Forecast

      Demand is modelled at the SKU-store level where planning decisions are made.

    2. Decide

      Selected demand becomes a constraint-aware allocation or ordering recommendation.

    3. Approve

      Configured workflows route decisions through the required approval path and retain the action record.

    Retail planning framework

    Six decisions, one connected chain.

    The decision-chain framework maps the six decisions that shape a retail season. PACX currently provides forecasting, allocation and ordering workflows; the adjacent chapters show the context those workflows receive from, and hand back to, the wider planning process.

    Skip the walkthrough ↓

    01

    Demand forecasting

    Demand, learned at the level decisions are made.

    PACX stores forecast output at SKU-store-week grain and provides a workbench for demand-source selection, hierarchy adjustments, discount-driver scenarios and versioned finalisation. Lower and upper forecast bounds can travel with the granular output.

    Forecast, 12 weeksWomen's block heel · Tan · 38 · North · 212 stores
    Demand forecast, Women's block heel · Tan · 38, cluster North · 212 storesLine chart of units per week: eight weeks of actuals rising from 96 to 134 units, then an eight-week forecast peaking at 184 units in week 4 (festive uplift) and easing to 139, with a P10 to P90 confidence band. MAPE 11.8% on holdout.100150200u/wknowW-8W-4W1W5W8actualforecastP10–P90

    Drivers

    • Festive uplift+18%
    • Promotionnone
    • Weatherneutral

    MAPE 11.8% on holdout

    Nearest neighbours · new product

    • 01Women's kitten heel · Tan
    • 02Women's block heel · Black
    • 03Women's strappy sandal · Camel
    Illustrative reconstruction · fictional data

    02

    Merchandise financial planning

    A plan that reconciles to the forecast, not the other way round.

    In the framework, merchandise financial planning supplies a governed demand and open-to-buy context for later decisions. PACX can ingest an MFP demand layer and preserve the selected source; it does not present this chapter as a complete MFP workspace.

    Open-to-buyWomen's block heel · Tan · 38 · North · 212 stores

    Forecast · W1–W8

    Demand sparkline, Women's block heel · Tan · 38 — actuals and 8-week forecast
    Merchandise financial plan for Women's block heel · Tan · 38: sales, margin percent, inventory and receipts by month, February to July, in ₹ lakh. Variance against the forecast of 4 percent or more is flagged.
    ₹ lakhFebMarAprMayJunJul
    Sales118124131152+6%139121−5%
    Margin %46%46%45%44%45%43%
    Inventory412398405371−4%342300−6%
    Receipts96110128+5%887452

    Values in ₹ lakh · variance vs forecast flagged at ≥ 4%

    Illustrative reconstruction · fictional data

    03

    Assortment and size curves

    Buy the right breadth, depth and sizes.

    Assortment quantities can enter the allocation workflow as a demand source. Approved size curves then spread demand at global, store-group or store scope, with the most specific configured curve taking precedence.

    Size curvesWomen's block heel · Tan · 38 · North · 212 stores
    Forecast · W1–W8Demand sparkline, Women's block heel · Tan · 38 — actuals and 8-week forecast
    RecommendedLast year
    North · 212 stores
    Size curve — recommended buy share vs last yearHistogram of buy share by size for North · 212 stores. Recommended shifts depth to size 38 at 28%, up from 22% last year.3536373839404128%

    Depth +6 pts on 38 — last season broke in week 3.

    Recommended buy share versus last year, North · 212 stores: size 35 4% (last year 8%), size 36 12% (last year 14%), size 37 22% (last year 18%), size 38 28% (last year 22%), size 39 19% (last year 18%), size 40 11% (last year 13%), size 41 4% (last year 7%).

    Range · depth recommended

    • Women's block heel · Tan420 u+15%
    • Women's block heel · Black360 u+8%
    • Women's kitten heel · Tan240 uhold
    • Women's strappy sandal · Camel180 u−10%
    Illustrative reconstruction · fictional data

    04

    Allocation and replenishment

    The right units in the right store, before the gap opens.

    PACX builds initial allocation plans from approved demand, eligibility rules, size curves, available inventory and scoped weeks-of-supply or unit constraints. Ordering and transfer workspaces carry recommendations through adjustment, approval and dispatch states.

    Allocation · replenishmentWomen's block heel · Tan · 38 · North · 212 stores
    Forecast · W1–W8Demand sparkline, Women's block heel · Tan · 38 — actuals and 8-week forecast

    Units allocated by store · North · 212 stores · 24 shown

    Allocation intensity for 24 stores in cluster North: darker means more units; 15 of 24 stores at high allocation.

    Replenishment · Northconstraint

    • Indiranagar84 ucapacity 120
    • Banjara Hills66 upack size 6
    • Salt Lake48 ulead time 5d
    • Aundh36 u⚠ minimum 24
    • Koramangala → Indiranagar24 utransfer · 2d
    • Jubilee Hills → Banjara Hills12 utransfer · 1d
    Illustrative reconstruction · fictional data

    05

    Pricing and promotions

    Price and promote for margin, not volume alone.

    The forecast workbench can recalculate a demand scenario from a configured discount and elasticity coefficient. Candidate-price optimisation, promotion calendars and commercial approval remain part of the surrounding planning process, not a published PACX module claim.

    Pricing · elasticityWomen's block heel · Tan · 38 · North · 212 stores
    Forecast · W1–W8Demand sparkline, Women's block heel · Tan · 38 — actuals and 8-week forecast
    Price elasticity, Women's block heel · Tan · 38Demand index against price index: demand falls smoothly from 1.62 at 0.7 times the current price to 0.71 at 1.2 times. The current price point is marked at 1.0.1.0×1.5×demand index0.8×1.0×1.2×price indexcurrent price · 1.0×

    Festive promo · W4

    Promo uplift
    +34% units
    Incremental margin
    +₹1.8L
    Cannibalisation
    −9% on Women's kitten heel · flagged
    Illustrative reconstruction · fictional data

    06

    Markdown optimisation

    Clear the season on your terms.

    The framework closes with the markdown decision: remaining inventory, time, expected demand and margin constraints determine the trade-off. This chapter is an explanatory planning model and illustrative interface, not a claim of a released optimisation workflow.

    Markdown scheduleWomen's block heel · Tan · 38 · North · 212 stores
    Forecast · W1–W8Demand sparkline, Women's block heel · Tan · 38 — actuals and 8-week forecast

    Sell-through target 85% by W12

    Margin floor 42%

    Markdown depth per week for Women's block heel · Tan · 38, weeks 9 to 12, by store scope. A dash means no markdown that week.
    ScopeW9W10W11W12
    North · Core · 96 stores–20%20%30%
    North · Growth · 71 stores20%20%30%30%
    North · Outlet · 45 stores30%40%40%50%

    Markdown depth by store · North · 212 stores · 24 shown

    Markdown depth for 24 stores in cluster North: darker means a deeper markdown; 6 of 24 stores at high depth.

    → every outcome feeds the next forecast

    Illustrative reconstruction · fictional data

    → Every outcome feeds the next forecast.

    Explore the framework

    Why PACX

    Three design choices keep decisions usable.

    Models at the level of the decision.

    Forecasts are organised around the product and location contexts planners act on.

    Read the evidence policy →

    Decisions stay connected to outcomes.

    Planning inputs, decisions and observed outcomes can be evaluated as one continuous record.

    How claims are validated →

    Recommendations reach a workflow.

    Forecasts can flow into allocation and ordering decisions with selected sources, constraints, approvals and audit records preserved.

    PACX is designed to connect planning inputs to governed operational decisions.

    Operating metrics are withheld until their evidence records are approved.

    Decision agentsNew

    Automation earns its permissions.

    PACX planning workflows can generate recommendations, apply configured constraints, auto-approve eligible allocation plans, route other decisions through staged approval and record changes in an append-only audit log. Broader decision-agent availability is stated only for a documented module and customer scope.

    • 09:14Executed

      Replenishment agent

      Proposed 1,240 units → Cluster North

      ✓ margin floor 42%✓ under order ceiling

    • 09:16Held, planner notified

      Markdown agent

      Second step on 38 SKUs held — sell-through above target

      ✓ sell-through target

    • 09:21Sent for approval

      Transfer agent

      16 store-to-store moves proposed

      ⚠ above tier-2 threshold

    • 09:27Executed

      Pricing agent

      3 price steps on slow sellers

      ✓ price floor

    • 09:33Executed

      Replenishment agent

      Proposed 480 units → Cluster West

      ✓ margin floor 42%✓ under order ceiling

    Margin floor
    42%
    Spend ceiling per order
    ₹2,50,000
    Store scope
    212 stores
    Approval tiers
    2
    Pause all agents
    off
    Audit log
    14,202 entries

    Illustrative agent activity · fictional data

    “By 2030, 50% of cross-functional supply chain management solutions will use intelligent agents to autonomously execute decisions.”

    Gartner, 2025

    “Over 40% of agentic AI projects will be canceled by the end of 2027.” Guardrails are why.

    Gartner, 2025
    Read about governed automation

    Results

    Customer outcomes are published only after validation and written approval.

    How it works

    A staged path from data to governed decisions.

    1. Connect

      Agree the decision scope, map the required data and validate access before implementation begins.

    2. Model

      PACX prepares and validates forecasting models against agreed baselines on your data.

    3. Decide and act

      Start with recommendations. Turn agents on one decision at a time, inside your guardrails.

    1. Stage 1

      Scope and data validation

    2. Stage 2

      Forecast evaluation

    3. Stage 3

      Governed decision workflow

    Ask PACXNorth · 212 stores

    Ask it the way a planner would.

    Which stores will break size 38 in the next 4 weeks?

    • Profiling 14 weeks of size-level sales for 212 stores
    • Finding nearest-neighbour styles for 3 new options
    • Validating model — MAPE 11.8% on holdout

    Decisions

    • 27 stores at risk of breaking size 38 by week 3
    • Recommended: 16 transfers, 2 reorders — all within margin floor
    • 1 above tier-2 ceiling, sent for approval

    Scripted preview · not a live query

    Book a demo

    Trust

    Every decision shows its why.

    Forecast inputs, planner adjustments, approval decisions and downstream actions retain their workflow context. PACX records changes in an append-only audit log so authorised teams can review who changed what and when.

    RecommendationWomen's block heel · Tan · 38

    Reorder 1,240 u → Cluster North

    Executed

    Why

    • Festive uplift+18%
    • Promotionnone
    • Weatherneutral

    MAPE 11.8% on holdout

    Guardrails checked

    ✓ margin floor 42%✓ under order ceiling

    Role-based permissionsBuilt in

    Planning capabilities and approval actions are checked against the assigned role and tenant scope.

    Audit logsBuilt in

    Recorded workflow changes include actor, role, before-and-after state, correlation and timestamp fields.

    Visit the trust centre

    Built for fashion, footwear, jewellery and specialty retail.

    By segment

    By role

    Questions retailers ask us.

    What is PACX?

    PACX is an AI-native retail planning platform for demand forecasting, allocation and ordering, with connected recommendations, configurable constraints, staged approvals and auditable actions.

    Explore the decision-chain framework
    Who is PACX for?

    Multi-store retailers in fashion, footwear, jewellery and specialty categories, and the merchandising, planning, finance and IT teams inside them. It replaces planning in spreadsheets and legacy suites; it does not replace the merchant's judgement about brand and range.

    About PACX
    What decisions does PACX automate?

    PACX currently supports configured auto-approval for eligible allocation plans and staged approvals for allocation, ordering and transfer workflows. The exact unattended scope is enabled per workflow; unsupported price, promotion or markdown automation is not implied.

    Read about governed automation
    How do guardrails work?

    Configuration defines eligible products and stores, scoped inventory constraints, allocation strategy, approval levels and role requirements. Decisions outside an automatic path remain in a review state, and recorded changes can be traced through the audit log.

    Define agentic AI in retail planning
    What data do we need to start?

    Data requirements depend on the first decision scope and the systems involved. PACX confirms the required fields, history and access method during technical discovery before making an implementation commitment.

    Read the data onboarding guide
    How long does it take to go live?

    Timing depends on scope, data quality, integration work and approval requirements. PACX agrees milestones after technical discovery rather than publishing a standard rollout promise without customer evidence.

    See the implementation approach
    Which systems does PACX integrate with?

    Integration scope is confirmed during technical discovery. PACX publishes a connector name only after its production status is documented; any required exchange method is specified in the implementation plan.

    Review the data contract approach
    How is PACX different from RELEX, Blue Yonder, o9 or Toolio?

    PACX is designed around one connected retail decision chain, explanations and merchant-set guardrails. During an evaluation, PACX maps those design choices against your current workflow rather than publishing unsupported claims about another vendor.

    Review the evidence policy

    See your decision chain run on your data.

    A 45-minute working session with a PACX planner: map one business question, the data it needs and the decision that follows.

    Book a demoGet the walkthrough by email