The operator's guide to
Inventory Planning

Most inventory systems tell you what happened last week and leave you to work out what to do about it. This one runs the calculation every night and arrives at concrete recommendations. When you log in, your job is not to hunt for problems — it is to review the exceptions the system has already found and accept, adjust, or reject them.

Three principles run through the product:

• Exception-driven. You should not scan 23,000 products every morning. You should see the handful that need a human decision today. Everything else runs automatically.
• Nothing is hard-coded. Weeks of supply, minimum presentation quantities, size curves, order minimums, approval levels — all of these are your configuration. The engines only ever do what it tells them.
• Every number is explainable. Each recommendation traces back to a forecast, a constraint and a stock position, and every change anyone makes is recorded with a before and after.

Your workspace

PACX creates a workspace for you at your own address. Everything in it — products, stores, forecasts, plans — belongs only to you. Users from other workspaces cannot see it, and yours cannot see theirs.

Who does what

Your administrator assigns each person one or more roles. Roles decide which controls you see and which actions the server will accept.

The two dates in the header

Refresh dates:   Inventory. Refresh Date: 07-25-2026   |   Transaction. Refresh Date: 07-25-2026

• Inventory Refresh Date — how current your stock positions are.
• Transaction Refresh Date — how current your sales and purchase-order data is.

They advance independently. If one is older than you expect, that data feed has not arrived — treat the day's recommendations with care and tell your administrator.

Every night the platform runs a fixed sequence. Each step is recorded, and the full ledger is visible on the Admin page.

The dark rail on the left of the app is your menu.

Store Inventory — the ten KPIs

Each card is a number across your whole eligible product × store estate.

Vendor-DC — five KPIs

Below these sit the four buying exceptions — Expedite, Upcoming, Pending, Transfer — and the Deep Dive chart described in section 09.

Forecast — accuracy

Cards for Last Week, Last 4 Weeks and Last 8 Weeks:

Actuals

Units actually sold in that window.

IA Forecast Accuracy

How close the raw engine forecast was: 1 − (total absolute error ÷ total actuals), as a percentage.

Adjusted Forecast Accuracy

The same measure for the forecast after your planners' adjustments.

Override Success

Of the rows a human adjusted, the share where the adjustment actually landed closer than the machine.

The list

Every plan with its status; filter with the pills, click a name to open it.

The review screen — summary cards

The three tabs

Product View — one row per product, drilling down: Stores › shows that product across its stores, then Sizes › shows the size split. Watch Aggregated Demand (Target WOS), which covers your target weeks of supply plus the delivery lead time — the engine accounts for stock stores burn while the truck is moving. And # Allocated vs # Eligible Stores: 37 of 75 means the other 38 already have enough.

Store View — the same plan seen store-first, for “what is store 115 getting in total?”

Store Capacity — the safety check. A red, negative Net Available Capacity (say −8) means this allocation pushes a store past its physical capacity.

Editing a plan

Switch View Mode → Edit Mode; Allocated Quantity becomes editable at product and store level. On Save Edits the server:

• spreads a product-level change down across its stores, and a store-level change down across its sizes, in proportion to the current mix;
• rolls totals back up so every level agrees;
• re-checks DC availability — asking for more than exists is blocked and nothing changes;
• re-checks capacity — a breach is a warning, not a block;
• records every change in the Audit trail.

If a colleague changed the plan while you were editing, you are told and the plan reloads — no silent overwrites.

Finishing

Move to Order Batching → Finalize → Send downstream. Download file gives you the CSV to inspect and changes nothing. Send downstream is the real transmission, restricted to Super Users and Admins.

The forecast is a stack, not a number

For any product, channel and week there are several numbers:

Edit a multiplier, press Save multipliers, and Adjusted User and Final recalculate. This is what allocation and ordering consume — one number drives both.

Demand selection

Per product and channel, choose whether Final Forecast follows the IA forecast, your Adjusted User forecast, or MFP. Set All applies a choice across everything, with a confirmation.

Drivers of Forecast

Currently one driver: discount. Enter a planned discount for a week and the forecast recalculates using that product's price elasticity, showing the uplift factor applied. This answers “what if we run 15% instead of 10%?”

View / edit Hierarchy Forecast

Total → Class → Style-Color → Store, weeks across. Edit at any level and the change spreads down proportionally to the underlying rows and rolls back up. Blank, non-editable cells are weeks before a product launches.

Save & Finalize Forecast

Freezes the current forecast as a numbered vintage and locks it against further edits. Do this when you are happy: allocation and ordering then work from a stable basis, and accuracy is measured against exactly the numbers you committed to.

Deep Dive — the 52-week picture

Ten series on one chart. The shaded area is Projected DC Inventory week by week; the red line is Safety Stock; the bars are incoming receipts and outflow to stores.

Expedite Orders — act now

Products where the DC runs out before the next scheduled order can arrive.

Stockout Date & Recovery Window

When, and how long you have.

Lost Sales Qty / Potential Sales Gain

What inaction costs — your prioritisation number.

Off Cycle Qty & Receipt Date

The emergency order the system suggests, in eaches and packs.

Earliest Planned Cycle

What would have arrived anyway, and when.

Two actions: View PO lists existing purchase orders arriving after the stockout — often the cheapest fix is to phone the vendor and pull one forward. Accept off-cycle creates the emergency order and puts it into the approval flow.

ROQ Scheduler — the regular buying rhythm

Rows are product × DC; columns group by month. Each period shows the recommended quantity in packs and eaches, plus the vendor's MOQ. Type into the eaches box and the system rounds up to whole packs and whole order multiples — you cannot accidentally order 101 units when the vendor ships in sixes.

Tick the rows you want and press Approve Orders. If any are red you must explicitly choose:

• Round up — raise to the MOQ. You buy more than you need; excess stock.
• Keep — order below MOQ anyway, if your vendor allows it.
• Drop — skip this cycle and accept the stockout risk.

The system will not choose for you: all three are legitimate, and only you know the vendor relationship.

Pending Orders, Scenarios, Transfers

Pending Orders — where two-level approval is in force, everything a buyer approves waits here for a manager, and a different person must approve it.

Scenarios — create one, change some order quantities, and Compare shows the delta in units and cost against the live plan without touching it. Already-approved rows are never overwritten.

Transfer Inventory — store-to-store rebalancing where the recovered margin beats the transfer cost.

How editing works

Grids are edited in place; changed cells turn amber. On Save each row is validated independently, and rows that fail stay amber with the reason. Some rules apply across rows and reject the whole save:

• Overlapping status periods for the same product or store.
• Size-curve fractions that do not sum to 100% for a scope.
• min WOS ≤ target WOS ≤ max WOS must hold.

Audit records every configuration change, forecast adjustment, allocation edit, approval and dispatch, with actor, timestamp and a before/after diff. This is your answer to “who changed the minimum, and when?”

Admin shows the nightly batch ledger: each run's steps as coloured pills, with row counts or the error on hover. You can re-run the whole batch, or re-run from a single step if one part failed.

1. Historic forecast snapshots are stored, but browsing a past snapshot in the workbench is not yet available. The header shows which vintage you are on.
2. Order recommendations aggregate sizes at product × DC level. Size-level ordering is planned.
3. Transfer recommendations are PACX's design, not a replication of an established standard — treat them as a starting point for discussion.
4. Three of the five cohort cards are marked Coming Soon.
5. Data feeds: the forecast import is live. Depending on your onboarding stage, other feeds may still be loaded by the PACX team rather than automatically.
6. Cross-product effects are not modelled in scenarios.

The numbers look wrong

Check the two refresh dates first, then Admin for a failed batch step.

A button is missing or greyed out

Almost always a role or plan-status restriction — see section 02, or ask your administrator.

A save was rejected

The message names the rule: overlapping periods, curve sum, or insufficient DC stock.

Anything else

Your PACX contact, quoting the page and — if you have it — the correlation ID from the error.