The four lanes
Every SKU decision travels four lanes. Skip a lane and you pay later in damage, DIM, or brand inconsistency.
Lane 1 — Geometry (fit the product, not the shelf)
- Outer dimensions that carriers actually bill
- Nesting / multipack options
- Orientation constraints (this side up, do not stack)
Lane 2 — Protection (survive the network)
- Fragility class (drop, crush, puncture, shock)
- Cushioning type and thickness
- Climate / moisture exposure
Lane 3 — Presentation (what the customer feels)
- Mailer vs shipper vs retail-ready
- Inserts, scent, reveal sequence
- Photography vs reality gap
Lane 4 — Compliance & reverse
- Labeling, hazmat flags, battery rules
- Recyclability claims you can defend
- Returnability / reuse of the same pack
SKU classes (pick one before you prompt)
| Class | Examples | Default posture |
|---|---|---|
| A — Soft goods | apparel, textiles | Flat poly/mailer first; box only if fold memory matters |
| B — Rigid durable | tools, hard goods | Shipper + minimal void; focus on puncture |
| C — Fragile | ceramics, glass, cosmetics jars | Dual-wall / inserts; test drop before brand spend |
| D — Liquid / leak risk | bottles, kits | Secondary containment + orientation |
| E — Kit / multi-component | subscription boxes | Partition logic before pretty print |
| F — Oversize / awkward | lamps, framed art | Custom geometry; AI for cut plan, human for freight mode |
20-minute AI workflow (new SKU)
- Measure truthfully — L×W×H of product as packed orientation, weight, fragility notes, retail vs ship-only.
- Run the geometry prompt (see prompts #1–3) — ask for 3 outer-dimension candidates with estimated DIM weight for UPS/FedEx/USPS rules of thumb.
- Protection pass — ask AI for cushioning options ranked by damage risk vs cost for your class.
- Cost pass — materials + estimated DIM delta vs current pack (or vs a naive “next size up” box).
- Presentation pass — only after geometry/protection stabilize. Brand moments on a leaking bottle are vanity.
- Compliance pass — batteries, aerosols, claims language, recycle marks.
- Pack-out card — one-page station instructions a seasonal hire can follow.
Go / no-go gates
Do not order custom print until:
- Damage assumption is explicit (target <1% unless category norms differ)
- Outer size is stable across 20 sample orders
- Void fill method is chosen (paper, air, foam-in-place, molded pulp, etc.)
- Someone timed a real pack-out (labor is a packaging cost)
Do not claim “100% recyclable” until:
- Every component has a known stream in your top 3 ship-to regions
- Inks/coatings/windows/tapes are not silent contaminants
Decision matrix: mailer vs box vs custom
| Signal | Prefer |
|---|---|
| Soft, compressible, low scratch risk | Padded mailer / poly mailer |
| Needs stacking strength or retail shelf | RSC shipper / retail-ready carton |
| High brand ASP + unboxing KPI | Custom mailer or printed shipper after size lock |
| Multi-SKU kits with movement risk | Partitioned shipper or tray + sleeve |
| Very low volume (<50/mo) | Stock sizes + label/branding overlay |
How AI should talk to your BOM
Ask models to output structured fields you can paste into ops tools:
outer_l_w_h_inestimated_dim_weight_lbprimary_materialcushioningclosurepack_time_seconds_estimaterecyclability_notesriskstest_plan
If the answer is only adjectives (“eco-chic, premium”), re-prompt for numbers.