Hezi Industrial Design — Dongguan, China. 320+ products shipped.
A client approves a beautiful sketch. Three weeks later the mold is cut, the prototype lands, and the thickness is wrong and the button can't be pressed. The sketch was the problem — not because it was ugly, but because it was drawn to impress instead of drawn to decide.
At our studio, product design sketching runs two to three rounds. Here is what each round is actually for, and the three checks a sketch must pass before we let it anywhere near a 3D model.
The round structure: each round has one job
| Round | Output | Real purpose |
|---|---|---|
| Round 1 | 5–8 form directions | Elimination — confirm what the client does not want |
| Round 2 | Narrowed to 2–3 | Converge: one shared perspective and line language |
| Round 3 | Detail fixes only | Buttons, ports, parting lines — no new ideas |
Round one is not a beauty contest. The direction a client finally picks is rarely the one we liked most internally — it's the one closest to the vague picture in their head. Redrawing a sketch costs a fraction of rebuilding a 3D model, so the elimination work belongs here. Every direction must carry its main dimensions and its parting direction; a proposal missing either doesn't advance, because it will be redrawn anyway once it reaches engineering.
Paper or tablet? The structure team doesn't care
Structural engineers read three things: orthographic views, key dimensions, parting direction. The medium is irrelevant. What differs is cost and communication:
- Paper: fastest for round-one divergence — seven or eight directions in one review. Scanned pages keep line weight and hesitation visible; remote clients can see which stroke was uncertain.
- Tablet: better for round two — layers make color and overlay iteration cheap. But a clean digital sheet can falsely signal the design is already resolved.
- Quick 3D rough model: slowest (half a day per form), but the only way to honestly judge volume. Use it after the form is basically settled.
For remote clients we'd rather send paper originals plus a voice note walking through each direction than a polished render. It produces better decisions.
The three checks before a sketch moves forward
- Side profile at near 1:1. We control length-to-height proportion on a near-orthographic side view and refuse dramatic perspective. The harder the perspective, the faster the form collapses in 3D — client signs, model arrives, thickness is wrong, buttons don't fit, full rework.
- The hand is in the drawing. For handheld products we sketch the grip holding it. Button positions are marked against an adult index fingertip — roughly 20 mm; a smaller contact area is hard to press accurately.
- Parting direction marked in round one. A smooth sweeping surface with no draft angle or ejector-pin space is the most expensive mistake to fix later. This one is written from 320+ shipped products.
When the client rejects three rounds in a row
It's usually not the sketching. At least one of three inputs was never locked: target user, retail price band, or benchmark competitor. The missing one keeps resurfacing as "it just feels off." We once had two rounds rejected, asked about retail price, and the answer was "not calculated yet" — once the price band landed, most of the facet details were scrapped. That rework is free of charge but eats the schedule: a 7–15 working-day appearance phase becomes a month.
When to stop sketching
Not when the client nods. When a structural engineer can pull orthographic views straight from the sheet. If they can't, keep drawing — don't rush into 3D. Sketching is the cheapest round in any engagement tier (appearance 5,000–12,000 RMB; with structure 15,000–40,000; full case 30,000–80,000), so it deserves the extra two days.
Originally published on the Hezi Industrial Design website. Hezi is an industrial design studio in Dongguan specializing in product appearance, structural, and CMF design with 320+ products in mass production.
Top comments (0)