Design & Development

Product Concept & Fashion Design

Digital sketching, generative design and collection planning.

Advanced product design and concept work sits at the intersection of creative direction and operational feasibility: designers and technologists have to develop a coherent range across multiple digital tools (vector sketching, generative AI ideation, colourway and print systems) while keeping every option traceable to a price point, a fabric minimum order quantity and a delivery window. This means treating the design file itself as a managed asset — version-controlled, linked to the eventual tech pack, and reviewed through an approval workflow that captures every comment and sign-off rather than relying on email threads or verbal agreement that gets lost by production handoff.

The more advanced skill is knowing where generative and AI-assisted tools genuinely accelerate ideation versus where they introduce risk: these tools are strong at generating volume and variation quickly, but an experienced designer still has to filter outputs for manufacturability, brand identity consistency, and origin of training data before anything goes near a customer-facing collection. Range planning at this level also means actively managing option count and repeat rate against merchandising targets throughout the design process, not just at a final review, so the range stays commercially balanced rather than being cut down reactively once costs and lead times come back from sourcing.

How the work is done

1
Set the design brief from range plan and trend direction
Translate the merchandising plan's price architecture, option count and trend stories into a concrete brief per category before sketching begins.
2
Ideate in digital sketch and generative tools
Produce flats, technical illustrations and, where useful, AI-generated silhouette or print variations as a starting point rather than a final answer.
3
Filter concepts for feasibility
Cross-check each concept against known fabric availability, construction complexity and target cost before it advances past initial review.
4
Develop colourways and print repeats
Build colour combinations against the brand's palette library and engineer print repeats ready for the intended print method (screen, digital, jacquard).
5
Run structured design approval
Circulate options through a digital approval workflow with designer, merchandiser and buyer sign-off, keeping a version history of comments and changes.
6
Lock range and hand off to development
Finalise the approved option count and colourway plan, then pass specifications into the tech pack and 3D development stages with no open decisions left unresolved.

Decisions you have to make

How much to rely on AI-generated concepts versus original design work?: Use generative tools for volume ideation and early exploration, but treat outputs as raw material needing significant designer curation, not finished IP ready for approval.
How many colourways to develop per style?: Balance the commercial upside of choice against the cost and complexity of managing more SKUs; anchor the decision to the buyer's agreed plan and known reorder patterns rather than defaulting to a round number.
When to cut a design concept for feasibility versus push suppliers to accommodate it?: Cut early if the fabric or construction risk is high relative to the trend's expected life; push suppliers only for concepts central to the brand's positioning where the commercial upside justifies extra development cost.
How tightly to control brand consistency across AI-assisted outputs?: Set clear brand guardrails (proportions, signature details, colour rules) before generating concepts, since AI tools left unconstrained will drift toward generic trend outputs rather than distinctive brand identity.
How much option count to carry per category?: Weigh the merchandising benefit of range breadth against the dilution of buying depth per option; trim breadth first in categories with thin sell-through history.

Key metrics (indicative)

Design-to-approval cycle time

Indicative working range against the seasonal calendar

Long approval cycles compress downstream development and sourcing time, increasing pressure and risk later in the process.

Percentage of design concepts surviving feasibility review

Track against baseline per category

A very low survival rate suggests the brief and feasibility constraints aren't reaching designers early enough in ideation.

Option count versus range plan target

Per the buyer's agreed plan

Keeps design output aligned to what merchandising can actually buy and sell at planned depth.

Colourway sell-through variance across a single style

Track against baseline

Wide variance signals that colourway decisions should be more selective, reducing markdown exposure on underperforming colours.

Comment/revision cycles per style in the approval workflow

Track against baseline

A high number of revision cycles per style indicates the brief or trend direction wasn't clear enough at the outset.

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

Common pitfalls

Approving AI-generated concepts without checking whether similar imagery already exists elsewhere, risking originality and brand-identity issues later.
Designing options without checking fabric minimum order quantities first, leading to costly late-stage cuts once sourcing costs come back.
Letting colourway count grow late in the process without corresponding buy-depth checks, spreading demand too thin across too many SKUs.
Relying on email or verbal sign-off instead of a structured approval workflow, causing disputes later about what was actually approved.
Treating print repeat engineering as a final-stage task rather than checking it against the intended print method early, causing rework once production sampling starts.

Advanced notes and limits

Generative AI design tools are genuinely useful for rapid variation at the concept stage, but current tools still struggle with garment-specific construction logic (seams, grain lines, closures), so outputs typically need substantial technical redrawing before they are production-ready.
Brand consistency checks against AI-assisted design work are still largely manual; automated style-consistency scoring exists in pilot form at some tool vendors but is not yet reliable enough to replace a design director's review.
Range breadth versus depth trade-offs shift meaningfully by channel — a wholesale account may need narrower, deeper options than a direct-to-consumer digital range, and a single global range plan often needs channel-specific colourway or option splits rather than one-size-fits-all.
Digital approval workflows reduce miscommunication but do not remove the need for physical or 3D review at key gates; relying solely on flat digital sign-off for complex construction or fit-sensitive styles risks approving something that reads well on screen but fails in reality.

Worked example

Deciding colourway count against SKU cost and reorder pattern

Base style target price: $48 retail
Incremental cost per additional colourway (setup, sampling, minimums): $1,850 per colourway
Projected incremental units sold from adding one more colourway: 620 units
Gross margin per unit at full price: $19.20
Historical reorder rate on 4th-and-beyond colourway in this category: 9% of first buy

1Gross margin contribution from the additional colourway = 620 units x $19.20 = $11,904
2Net contribution after incremental cost = $11,904 - $1,850 = $10,054
3Historical 4th-colourway reorder rate of 9% is low, signalling weak sell-through tail beyond the third colourway in this category
4Compare the $10,054 projected contribution against the category's known pattern of low reorder past colourway 3
5Flag the projection as optimistic unless supported by a specific trend or customer research reason for this style

Approve up to three colourways as the commercially safe default for this style family, and only add a fourth if there is a specific trend-driven or pre-order signal justifying the projected 620-unit uplift, since the category's historical pattern suggests colourway four typically underperforms this forecast.

Case study

Context

A denim brand's design team began using a generative AI tool to rapidly produce dozens of wash and stitch-detail concept variations per style meeting, replacing what had been a slower manual sketching process.

Problem

Within two seasons, a noticeable share of the AI-generated concepts that reached sample stage were rejected late in development because the generated stitch details or pocket constructions were not achievable within the factory's existing machinery without costly retooling, and some proposed washes were visually derived from imagery whose rights were unclear.

Action

The team introduced a mandatory feasibility filter immediately after AI ideation, requiring a technical designer to flag construction and machinery constraints before any concept moved to sampling, and added a review step specifically for print and wash sourcing to confirm the AI outputs were not closely reproducing identifiable third-party artwork.

Outcome

Late-stage sample rejections attributable to infeasible construction dropped substantially the following season, and the added review step became a standing gate in the approval workflow rather than an ad hoc check.

Audit checklist

Every design concept is checked against known fabric availability and construction complexity before advancing past initial review.
AI-generated concepts are treated as raw ideation material requiring designer curation, not as finished, approval-ready output.
Colourway and option counts per style are actively tracked against the merchandising plan throughout design, not only at final review.
The design file and its approval history are version-controlled with all comments and sign-offs traceable.
Print repeats are engineered for the specific intended print method before being handed to development.
Concepts using AI-generated imagery have been checked for potential reliance on identifiable third-party source material.
Design decisions to cut or push a concept for feasibility are documented with the reasoning, not left as verbal agreements.
The range hand-off to tech pack and 3D development contains no open or unresolved design decisions.

Glossary

Range architecture: The planned structure of a collection by price point, category and option count, used to keep design output aligned to commercial targets.
Generative ideation: The use of AI tools to rapidly produce large volumes of design variations as a starting point for human curation, rather than as finished output.
Colourway: A specific combination of colours applied to a single style, treated as a distinct option for costing, sampling and buying purposes.
Print repeat: The engineered unit of a printed design that tiles seamlessly across fabric width, sized and adjusted for the specific print method being used.
Design approval workflow: A structured, typically digital process by which design options are reviewed and signed off by designer, merchandiser and buyer with a retained history of decisions.
Feasibility filter: A review step, ideally early in design, that checks a concept against known fabric, construction and cost constraints before further development investment.
Reorder rate: The proportion of an initial buy quantity that is subsequently reordered, used here as a signal of a colourway or option's ongoing sell-through strength.
Option count: The total number of distinct styles, or style-colour combinations, planned within a category or range, managed as a commercial lever alongside depth per option.
Version control (design): The practice of tracking successive iterations of a design file with a retrievable history, preventing loss of decisions or reversion to superseded versions.
Training data provenance: The origin of the images or designs an AI generative tool was trained on, relevant to assessing originality and rights risk in its outputs.

Practice questions

1. A brand is deciding whether to add a fourth colourway to a style. Incremental cost is $2,000, projected incremental units are 500 at a margin of $15/unit. Should they proceed, using only this data?
2. What is the main practical risk of letting AI-generated concepts go straight into sample development without a feasibility filter?
3. Why should option count be tracked throughout the design process rather than only at final range review?
4. A designer wants to push a supplier hard to accommodate a technically difficult concept. What should determine whether this is justified?
5. What are two distinct risks of relying heavily on AI-generated imagery for print or wash concepts?
6. Explain why a design approval workflow needs a retained version history rather than relying on the latest file only.

Sub-topics in this chapter

Digital sketching: Vector and raster tools (Illustrator, Photoshop, Procreate) used for flats, artwork and technical illustrations.
AI-assisted design: Generative models that produce silhouette, print or colour variations from a brief, used as ideation not final artwork.
Colourway development: Building multiple colour combinations from a single style, checked against palette libraries and brand rules.
Print development: Creation, repeat engineering and colour separation of prints ready for screen or digital textile printing.
Range planning: Structuring the collection into options, price points and drops that meet the merchandising plan.
Design approval workflows: Digital sign-off flows across designer, merchandiser and buyer with comment history and versioning.

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