Shippit Demand

Package Allocation V2 / Internal explainer / Aug 2026

Explainer

Accuracy is the product

New allocator. Same inputs. Better packs.

A new package allocator. Same merchant inputs. Better packs. Cost is a watch metric. Dims are upside, not the gate.

Wrong pack

Wrong pack becomes a wrong quote. The warehouse then changes it.

Warehouse swap

Operators do not trust the suggestion. They swap, split, or pick a satchel by hand.

Satchels weak

Fashion and retail leave in satchels. Today barely has them. A standard satchel can overwrite the submitted weight.

Over 25 kg

Carriers cap a parcel at 25 kg. Today we fail the order.

Create only

Create. Allocate. Quote. Book.

Allocation runs at order create. It does not run again at pack time. Quote sees the pack V2 just picked.

FIG 0.2 End to end

1 of 4

Shapes not results

Four orders. Today versus V2.

Scenario shapes. Not measured lift. Do not invent percentages. Do not put dollar savings on a slide.

01 Weight only · FIG 1.1

5.5 kg fashion. No dims.

Five apparel SKUs. Shopify-style create. No catalog.

Today

7 kg carton

5.5 kg is too small. Lands in the 7 kg box. Quote is for 7 kg. Warehouse often swaps it.

Fit-first pack

Actual package, not the next rate-card bucket. Weight-only is enough. Quote is for that pack.

02 Combine + overflow · FIG 1.2

Three items. Two fit.

Combined they do not fit one box. Two do. One does not.

Today

1 huge box or 3 boxes

V1 is one parcel or many. It cannot combine three then overflow. One mystery pack, or three.

Box A + overflow B

Combine what fits. Overflow the leftover into a smaller box. Two parcels. How we pack, not one mystery box.

03 Over 25 kg · FIG 1.3

31 kg. Weight only.

AusPost or another cap-at-25 carrier is in the quote set.

Today

One illegal 31 kg

The order fails. Nobody books until a human splits it.

25 kg + 6 kg

We split. We do not fail. Then quote. Then label.

04 Unit detect · FIG 1.4

210 × 150 × 20. Soft goods.

Merchant meant mm. Today we often read them as cm.

Today · as cm

Huge carton

No auto unit detect. Dims look huge. Wrong box, or dims ignored.

V2 · as mm

Satchel or flat pack

Auto-detect m / cm / mm. Dims help when they exist. Foldables still open.

Four decisions

What the allocator actually decides.

This order, every create. It does not decide the carrier. Rules engine, then cheapest or fastest, comes later.

FIG 0.3 Decide

01 Pack type

Box or satchel.

Fashion and retail should be able to land on a satchel. Pallet is later, not this cut.

02 How many

One, or several.

Several when items overflow, or when a split is required at 25 kg.

03 Size

From their presets.

Including tighter custom presets and CSV. Not only the coarse 5 kg / 7 kg brackets.

04 Weight

Dead weight first.

If they sent it, use it. If that field is blank, volumetric. V1 satchel overwrite is left behind.

Volumetric in Shippit today is L(cm) × W(cm) × H(cm) × 250 / 1000 kg. V2 does not change that formula. [verify]

Required path

Weight is enough. Catalog is not a gate.

API create with weight, and optional dims, must succeed when catalog is empty. Catalog is a supplement for merchants who do not send them.

FIG 0.4 Inputs

Input	Required	Who sends it	What V2 does
Dead weight	Required	API / Connect / Shopify / UI	Baseline. Weight-only is the fair test.
Dims L × W × H	Optional	Same payload	Used when present. Auto-detect m / cm / mm.
SKU / product	Optional	Same payload	Match to catalog if one exists.
Product catalog	Optional	Settings, later	Supplement. Persist-to-product-table is incomplete. [verify]
Custom presets + CSV	Optional	Settings	Tighter boxes than 5 kg / 7 kg.
Compressibility / max stack	Engine	Defaults	Not a merchant field. Do not expose.
Fragile / no-rotate / DG	Out	Catalog later	Not used by V2 alloc this cut.
Packed vs product dims	Open		Do not block the baseline.

This cut

In. And out.

Floated, not locked: smart presets from history.

Allocate at order create only
Combine then overflow into a smaller box
Satchels
Auto unit detection m / cm / mm
Tighter custom presets, including CSV
Weight and dims on the create payload
Catalog as a supplement, never a gate
Auto-split over 25 kg
Dead weight first. Volumetric only if blank
One global banner. Account opt-in. Off = legacy
Shadow V2 vs V1 at create. Logging only

Out

Re-run at pack time
Re-allocate after consolidation
LaunchDarkly, or any per-merchant flag
DG pairing inside the allocator
Co-pack / segregation in the first release
Compressibility as merchant fields
Palletisation
Standalone 3PL cartonisation product
Live carrier quoting inside shadow
Foldables solved
Scoring V2 on the dim-sending minority

Locked 27 Aug

On per account. Off is legacy.

Shadow logs V2 vs V1. It does not change the live pack. Do not put this in a customer release until opt-in accounts are live on V2. No customer launch date is locked. [verify]

FIG 0.5 Live model

Global banner

Flip one V2 banner. Every account can see it.

Account opts in

V2 then allocates at create for that account.

Off = rollback

Off returns that account to the legacy allocator.

No flags

No LaunchDarkly. No per-merchant engineering flag.

In the room

Say this. Do not say that.

Pass line: swap the algorithm, keep merchant behaviour, keep the same inputs. Accuracy is clearly above today. Score weight-only first.

Say

New allocator. Same inputs. Better packs.
Weight is enough. Dims help.
Satchels, not only boxes.
Over 25 kg we split. We do not fail the order.
On per account. Off if you want.
We are making the allocator merchants already use more accurate. We are not selling a standalone 3PL cartonisation product.

Do not say

Cost goes down
Catalog required
3D packer / 3D spatial fit
Re-run at pack time
Track A
Standalone 3PL / palletisation
Foldables solved
15% / 38% / 50%
Override rate as a customer claim
Lift on accounts that did not opt in
A named competitor comparison

The job

If we swap the algorithm and behaviour stays the same, accuracy must go up.

That is the product. A cheaper pack is a watch metric, not the headline.