"That Is Not What the Design Looked Like": Drawing the Line at the Design-to-Development Handoff, With Delivery Specs and a 12-Dimension Scorecard

We inherited a mess last year. A health food brand, NT$600,000 for a website plus member system, with design at one studio and development at another. Design delivered 42 Figma artboards, development finished the build, and at acceptance the client said "this does not match the design." Digging in, there were 63 disputed items, and 51 of them came down to the same thing: the design never drew that state — the pressed button, the input field in error, the list with no data, the label with text too long, the third breakpoint on mobile.

The design shop said "that is common sense." The dev shop said "it was not in the spec." Neither was wrong, because no line in either contract defined what "design delivered" means. Twenty-seven working days of delay, and the bill landed on the client.

Industry Myths Worth Breaking

Myth 1: "Handing over the design file is delivery"

Reality: A Figma file is a medium, not a deliverable. The actual deliverable is a specification that can be implemented and accepted. A beautiful artboard without a spacing system, state variations, or rules for extreme content leaves engineers guessing — and guesses never match what the client pictured. Complete design delivery includes five things: artboards, a component library, a state inventory, breakpoint rules, and content-extreme examples.

Myth 2: "Development should just follow the design without asking"

Reality: No design file can enumerate every state. A moderately complex page has 30 to 60 possible states (loading, empty, error, insufficient permission, overflowing content, offline, partially loaded). Even a detailed file covers 8 to 15 of them. The remainder must come from pre-agreed default rules, not developer improvisation and not post-hoc argument.

Myth 3: "Using one vendor for both prevents this"

Reality: it reduces friction but does not eliminate it. Single-vendor lowers communication cost and clarifies accountability, but disputes become internal and invisible to the client — you never learn which details an engineer decided alone. The real fix is not "one vendor," it is "handoff specification regardless of vendor count."

Myth 4: "Figma has Dev Mode, engineers can read it themselves"

Reality: Dev Mode gives values (spacing, colour, type size), not intent. It tells you a margin is 24px. It does not tell you whether that is the site-wide card gap that should become a variable, or a one-off that happened to look right. Miss that distinction and changing one spacing value touches 40 files three months later.

The Core Framework: Handoff Accountability Quadrants

Drop every disputed item into this table and ninety percent resolve on the spot.

	Drawn in the design	Not drawn
Defined in spec	Quadrant A — Build as drawn Any deviation is a development defect, fixed free.	Quadrant B — Build to rule Implement per agreed defaults (design system, component library); deviation is a development defect.
Not defined in spec	Quadrant C — Design governs visuals Match the artboard, but interaction and edge cases need confirmation; falls under change request.	Quadrant D — New requirement Neither side defined it; goes through change control and pricing.

The table's real value is not assigning blame but turning arguments into classification. Debating "is this a bug" is unresolvable. Debating "which cell does this belong in" is resolvable, because the criteria are factual: was it drawn, was it specified.

The pricing rule is simple: Quadrants A and B are fixed free and count toward acceptance; Quadrant C items are estimated at 0.5 to 2 hours each with a change request once the running total exceeds 8 hours; Quadrant D always takes a change request at standard rates. Writing those three sentences into a contract would have saved our client 27 days.

Three Typical Scenarios

Small projects (NT$150k–400k, single vendor): usually no formal annotation. Skip the full design system but require a state inventory — pressed, disabled, loading and error states for every interactive component, plus three breakpoint sketches. Adds 8 to 12 hours and prevents around 70% of disputes.
Mid-size projects (NT$400k–1.2M, split vendors): the most failure-prone size. Put a design delivery acceptance checklist into the contract and require a handoff meeting before development starts, confirming component naming, spacing system, colour variables and icon specs. Two hours typically saves twenty days.
Large or retained projects (above NT$1.2M): these need a real design system with token mechanics. Maintain a mapping table between Figma Variables and code-side variables, with design-side changes synced within one week. The investment here belongs in process, not more artboards.

Hidden Costs

Filling in missing states: 3 to 6 design hours per moderately complex page. At NT$1,500/hour across 10 pages, NT$45,000–90,000 upfront. After the argument it costs triple.
Naming refactors: design says "primary button," code says btn-primary, the client says "the orange one." Each mismatch adds 15 to 30 minutes per conversation, accumulating 20 to 40 hours.
Font licensing: unlicensed fonts discovered pre-launch. Traditional Chinese commercial licences commonly run NT$8,000–60,000 annually, and a swap forces re-tuning line height and line breaking site-wide.
Image and icon specs: without agreed export formats developers export their own, producing blurry or oversized assets. Rebuilding an icon set takes 6 to 15 hours and directly affects largest contentful paint.
Extreme content testing: names from 2 to 15 characters, titles from 5 to 60, amounts into nine figures. Skip it and layouts break in week one; remediation runs 10 to 20 hours.
The acceptance meetings themselves: two hours with three people is six labour hours, and dispute-heavy projects hold 8 to 12 of them.
The second design pass: the client says "it does not look like what I imagined" and design redoes a round. Nobody wants to pay, so the prime contractor absorbs it or it becomes delay.

In total, handoff hidden costs on a NT$600,000 project run NT$120,000 to NT$250,000 — 20% to 40% of project value. Writing the spec properly upfront usually costs NT$30,000 to NT$60,000.

12-Dimension Partner Scorecard

Ask both the design and development vendors these before signing. Score 0 to 3 each, 36 maximum. Below 24 means fix the process before starting.

☐ 1. Can they show a design delivery acceptance checklist from a past project?
☐ 2. Do design files use components and variables rather than loose layers?
☐ 3. Is there a state inventory (pressed, disabled, loading, error, empty)?
☐ 4. Are spacing and type scales systematised rather than per-page freehand?
☐ 5. Are breakpoint rules explicit, including min and max width behaviour?
☐ 6. Are extreme content examples provided (longest, shortest, empty)?
☐ 7. Are font and image commercial licences confirmed in writing?
☐ 8. Does component naming match between design and code?
☐ 9. Will they attend a handoff meeting before development starts?
☐ 10. Can development produce a clickable front-end skeleton within two weeks for design feedback?
☐ 11. Is there a defined change request process and rate card?
☐ 12. Are acceptance disputes resolved by classifiable criteria (like the quadrants) rather than case-by-case negotiation?

How ScriptWalker Works — and When Not to Hire Us

Three situations bring people to us: the client has designs and needs a builder, wants design and build together, or — like the story above — two vendors are already fighting and someone must close it out.

For the first we run a design implementability review (6 to 10 hours, from NT$12,000), flagging missing states, inconsistent spacing and unbuildable effects, then hand a remediation list back to the design side; that list typically cuts downstream disputes below 20%. For the second, handoff specs go into the contract annex. For the third, we classify disputes through the quadrants so both sides agree on free versus billable before anyone discusses schedule.

When we are the wrong fit:

You want pixel-perfect Figma replication with zero implementation feedback — a pure markup shop will do that far cheaper.
Designs are still changing substantially but you want development to start. Parallel work is possible but requires accepting rework and moving to retainer instead of fixed price.
You want us to arbitrate between design and development without adjusting the contract or process. Arbitration without a specification is just a new person in the same argument.

Kickoff Playbook

Week 1: three-party handoff meeting. Confirm the component naming map, spacing and type scale, colour variable list and icon export specs. Deliverable is a two-page handoff memo, signed by all three parties.
Week 2: design completes the state inventory and extreme content examples; development produces an unstyled clickable skeleton so the client sees flow early.
Weeks 3–4: fully implement 2 to 3 representative pages and hold the first handoff review. The purpose is not verifying features but calibrating standards — run the disputes through the quadrants once and confirm all three parties read the criteria the same way.
Week 5 onward: regular development cadence with a handoff check every two weeks; accumulated Quadrant C and D items batched into change requests.
Two weeks before acceptance: freeze design changes, handle Quadrant A and B fixes only. This line must be in the contract or the project never closes.

Decision Checklist

☐ Does the contract define acceptance criteria for "design delivered"?
☐ Is there a state inventory?
☐ Are spacing and type systematised or freehand per page?
☐ Are breakpoint rules written down?
☐ Are extreme content examples provided?
☐ Are font and image commercial licences confirmed?
☐ Does component naming align between vendors?
☐ Is a handoff meeting scheduled before development starts?
☐ Are dispute criteria classifiable, or negotiated case by case?
☐ Is the change request rate card in the contract?
☐ Is the pre-acceptance design freeze in the contract?
☐ Who makes the final call, and is it one person?
☐ Is a small calibration review scheduled in the first four weeks?

Fewer than eight ticks and your project will very likely stall two to four weeks at acceptance.

FAQ

Should design and development be split across vendors?

Splitting is not the problem; the absence of a handoff spec is. Splitting lets you pick the strongest team for each side and improves negotiating room, at the cost of diffused accountability. The test: if you or a named person can chair the handoff meeting and rule on criteria, split freely. If nobody can, paying a prime contractor an extra 10% to 20% integration premium is worth it.

Is "pixel perfect" a reasonable requirement?

For static screens, yes. For responsive and interactive behaviour, no. The same file necessarily differs across devices, font rendering engines and content lengths. The reasonable standard is design system consistency — the same spacing values, the same colour variables, consistent component behaviour — not page-by-page pixel comparison.

Who pays when the build does not match the design?

Use the quadrants. Drawn and specified is a development defect, fixed free. Not drawn but covered by the design system is also development's responsibility. Drawn but with undefined interaction is absorbed in small volume and billed above threshold. Undefined by both sides is a new requirement, priced separately. Put those four sentences in the contract and the argument shifts from "whose fault" to "which cell."

Do we need a full design system? We are a small company.

Not full, but a minimum viable one: a spacing scale (4, 8, 12, 16, 24, 32, 48), a type scale of 5 to 7 levels, a colour variable set with semantic names like danger and success, and four states for buttons and inputs. An afternoon defines all four and it prevents the most disputes.

Acceptance is already stalled. How do we stop the bleeding?

Three steps. First, list every disputed item without debating any of them. Second, classify through the quadrants and immediately execute the uncontested A and B items to restart momentum. Third, bundle C and D items into one change request and negotiate price and schedule once, not item by item. In practice this compresses a two-to-four-week deadlock into about five working days.

Stuck Between Design and Development?

ScriptWalker offers a design implementability review (from NT$12,000, 6 to 10 hours) and handoff dispute classification and closeout. The first happens before development; the second after the argument starts. The first is much cheaper. We also take on combined design and build projects with handoff specs written into the contract annex.