Digital & AI

Costing & Commercial Technology

Cost engineering with AI predictions.

Advanced costing treats the cost sheet as a living model that is rebuilt at each development milestone, not a one-time exercise at order placement. Cost engineers separate direct material cost, cut-and-sew (CM/CMT) labour, trims, packaging, freight and duty into distinct cost drivers so that a fabric price change, a construction simplification or a freight mode switch can each be tested in isolation. AI-assisted costing tools are used to generate a first-pass estimate from historical style data and current material indices, but the estimate is always reconciled against a supplier's actual quotation, because model-generated costs reflect pattern-matching against past styles and can miss a genuinely new construction or a regional wage shift. The discipline of maintaining a should-cost model separately from the negotiated cost is what lets a team spot when a quote has drifted from what the underlying inputs justify.

Commercial costing decisions extend beyond the factory gate into landed cost and margin architecture: duty classification, freight terms, currency exposure and payment terms all move the real cost of a unit as much as fabric price does. Advanced teams model cost sensitivity to fabric consumption changes from marker efficiency, since even a one to two percent shift in yield can outweigh a modest negotiated fabric price reduction. Costing is also where sourcing, design and merchandising trade-offs get resolved quantitatively — a fabric substitution, a trim simplification or a construction change is evaluated on its net landed-cost and margin impact, not just its unit material cost, before it is approved.

How the work is done

1
Should-cost model build
Build an independent cost estimate from fabric consumption, trim BOM, labour minutes and overhead assumptions before requesting supplier quotations.
2
AI/historical benchmarking
Run the style against a model trained on comparable past styles and current material indices to sanity-check the should-cost estimate and flag outliers.
3
Supplier quotation and reconciliation
Collect supplier quotes and reconcile line by line against the should-cost model, questioning any category that deviates beyond an agreed tolerance.
4
Negotiation on drivers, not totals
Negotiate against specific cost drivers — fabric price, marker efficiency, labour minutes, packaging — rather than a single lump-sum target price.
5
Landed cost and margin modelling
Add freight, duty, currency and payment-term effects to convert factory cost into landed cost, then test margin against planned retail or wholesale price.
6
Cost sheet lock and variance tracking
Lock the cost sheet at order confirmation and track actual invoiced cost against it through production to catch scope creep or unauthorised substitutions.

Decisions you have to make

How much weight to give an AI-generated cost estimate versus a supplier quote?: Use the model estimate as a negotiation anchor and outlier detector, not a substitute for a real quote, since it cannot see a supplier's current capacity, energy costs or a genuinely novel construction.
Which cost driver to negotiate first?: Prioritise drivers with the largest cost share and the most controllable variance — commonly fabric consumption and labour minutes — over trims or packaging, which usually move cost less.
How to treat currency and duty risk in the cost sheet?: Model a currency and duty sensitivity range per the buyer's agreed plan rather than a single point estimate, since both can move materially between costing and shipment.
When is a marker-efficiency gain worth a pattern or construction change?: Compare the fabric saving against the cost and time of a pattern rework and any fit risk it introduces; a small yield gain rarely justifies re-opening an approved pattern late in development.
How to handle a supplier quote that beats the should-cost model significantly?: Investigate rather than accept, since an unusually low quote can signal a missing cost element, a quality trade-off, or a supplier under-pricing to win volume it cannot sustain.

Key metrics (indicative)

Should-cost vs. actual quote variance

indicative tolerance range, track against baseline

Consistent variance beyond tolerance points to a flawed should-cost model or systematic supplier margin padding.

Cost sheet accuracy at order vs. final invoice

track against baseline, minimise drift

Large drift between locked cost and final invoice indicates scope creep or uncontrolled substitutions during production.

Fabric consumption vs. marker-efficiency target

indicative working range per the buyer's agreed plan

Consumption above plan directly erodes margin and often signals a marker or grading inefficiency worth investigating.

Gross margin realised vs. planned

track against baseline by category

Realised margin below plan flags a costing, pricing or landed-cost assumption that did not hold through to delivery.

Cycle time to finalise a cost sheet

track against baseline per development stage

Slow costing turnaround delays sourcing decisions and compresses the negotiation window before order placement.

Metric targets are indicative working ranges, not standards or legal limits.

Common pitfalls

Accepting an AI-generated cost estimate as final without reconciling it against a real supplier quote, missing a novel construction the model has never seen.
Negotiating on total unit price alone, which hides which specific cost driver moved and makes future negotiations harder to anchor.
Ignoring marker-efficiency variance because fabric price looks favourable, letting a yield problem silently erode margin at volume.
Locking a cost sheet without a currency or duty sensitivity check, then absorbing an unplanned margin hit when exchange rates move before shipment.
Failing to track actual invoiced cost against the locked cost sheet, so unauthorised trim or fabric substitutions go unnoticed until margin reporting.

Advanced notes and limits

AI costing models are trained on historical style and price data, so they degrade in reliability for genuinely novel constructions, new supplier regions, or during periods of unusual input-cost volatility; treat their output as a prior, not a forecast.
Should-cost modelling assumes stable labour-minute standards, but actual sewing-line efficiency varies by operator skill and line balance, so the same cost sheet can understate cost at a less mature factory even with identical fabric and BOM.
Optimising for lowest landed cost can conflict with lead-time or quality risk; a cheaper freight mode or a lower-cost supplier region often trades directly against delivery reliability, and that trade-off needs to be made explicit in the costing decision, not left implicit.
Cost driver decomposition works cleanly for standard woven or knit styles but is harder to apply to highly engineered or technical fabrics, where fabric cost itself is volatile and not well represented by historical indices.

Worked example

Testing whether a marker-efficiency gain justifies a late pattern rework

Current fabric consumption per garment: 1.85 m
Fabric price: $4.20 per m
Proposed pattern rework improves marker efficiency, reducing consumption to: 1.79 m
Order quantity: 20,000 units
Estimated cost of pattern rework, re-grading and re-sampling: $3,800
Days remaining before PP cutoff: 9 days

1Fabric saving per unit = 1.85 m - 1.79 m = 0.06 m.
2Cost saving per unit = 0.06 m x $4.20/m = $0.252 per unit.
3Total saving across the order = $0.252 x 20,000 units = $5,040.
4Net saving after rework cost = $5,040 - $3,800 = $1,240.
5Net saving as a share of total fabric spend (1.85 m x $4.20 x 20,000 = $155,400) = $1,240 / $155,400 ≈ 0.8%.

The rework nets about $1,240, a marginal 0.8% gain on fabric spend, so with only 9 days to PP cutoff the fit and schedule risk of reopening the pattern likely outweighs the saving; proceed only if the rework can be validated without adding a fit-cycle delay.

Case study

Context

A denim brand's should-cost model consistently showed a 6-8% gap against supplier quotations on a new heavyweight jacket style, but the team had been accepting quotes near the top of that gap for two seasons without investigation.

Problem

Margin reporting later showed the jacket category underperforming plan, and a review found the should-cost model had used an outdated labour-minute standard that did not reflect the extra operations needed for the heavier fabric and additional topstitching.

Action

The costing team rebuilt the labour-minute standard from actual time-and-motion data for the heavyweight construction, separated it from the standard denim labour assumption, and re-ran the should-cost model against the current quote.

Outcome

The revised should-cost model narrowed the gap to within an acceptable tolerance, confirming the supplier quote was justified rather than inflated, and the corrected labour standard was carried forward to prevent the same false-alarm gap recurring on future heavyweight styles.

Audit checklist

Should-cost model is built independently before supplier quotations are requested, not reverse-engineered from a received quote.
Labour-minute standards used in the should-cost model reflect the actual construction complexity, not a generic category default.
AI or historical-benchmark cost estimates are reconciled against a real supplier quote before being used in negotiation.
Cost sheet separates fabric, trims, CM/CMT labour, packaging, freight and duty as distinct, individually testable drivers.
Marker-efficiency assumptions are checked against actual achieved consumption before locking the cost sheet.
Currency and duty sensitivity has been modelled as a range, not a single point estimate.
Cost sheet is formally locked at order confirmation with a defined change-control process for later revisions.
Actual invoiced cost is tracked against the locked cost sheet through production to detect scope creep or substitutions.

Glossary

Should-cost model: An independent cost estimate built from fabric consumption, labour minutes, trims and overhead assumptions, used to evaluate supplier quotations rather than accept them at face value.
CM/CMT: Cut-Make (or Cut-Make-Trim) cost, representing the labour and associated overhead charged for constructing a garment from cut components.
Landed cost: The total cost of a unit delivered to the buyer's destination, including factory cost, freight, duty, insurance and related charges.
Cost driver decomposition: Breaking a total unit cost into separately analysable components such as fabric, labour and freight so each can be negotiated or modelled independently.
Marker efficiency: The percentage of fabric width and length actually used by pattern pieces in a cutting marker, with the remainder lost as waste; small efficiency gains can materially affect fabric cost.
Cost sheet variance: The difference between a locked cost sheet at order confirmation and the actual invoiced cost realised through production.
Duty classification: The tariff code and associated duty rate assigned to a garment based on its construction and fibre content, which materially affects landed cost.
Currency exposure: The risk that a cost or price agreed in one currency changes in effective value due to exchange-rate movement before payment is settled.
Should-cost vs. quote reconciliation: The line-by-line comparison of an independent cost estimate against a supplier's quotation to identify and question deviating cost categories.
Change-control on a locked cost sheet: A defined process requiring formal approval before any fabric, trim or construction substitution is allowed to alter a cost sheet already locked at order confirmation.

Practice questions

1. A supplier quote for CM is 12% below the should-cost model's labour-minute estimate. What should the costing team do?
2. Fabric consumption is 2.1 m/unit at $5.00/m, and a marker rework would reduce it to 2.05 m/unit at a one-time cost of $2,500 for an order of 15,000 units. Is the rework worth it purely on cost?
3. Why should currency and duty be modelled as a sensitivity range rather than a single point estimate in the cost sheet?
4. What is the main limitation of an AI-generated cost estimate for a genuinely novel construction?
5. How should a team prioritise which cost driver to negotiate first on a new style?
6. A cost sheet was locked at order confirmation but the final invoice came in 9% higher. What is the appropriate first diagnostic step?

Sub-topics in this chapter

Product costing: Roll-up of fabric, trim, labour, overhead and margin into a target and actual cost per style.
Fabric consumption: Calculated or marker-driven fabric use per garment, the largest variable in most cost sheets.
Wash costing: Cost of laundry, dyeing or garment finishing operations added per SKU.
Overhead allocation: Assigning factory overheads to styles using SAM, machine time or labour cost.
AI costing: Models trained on historic quotes to predict cost from tech-pack features in seconds.
Quotation management: Systems that manage RFQs, supplier responses and negotiation history in one place.

Lessons that teach this chapter

Where this chapter is applied

The value chain stages that use this chapter's skills — chapter to stage to skill.

Study tools for this chapter

Saved on this device only

Study without an account, or sign in to save progress across devices. See the full chapter syllabus.

All chapters and the 42-stage course