Surface Inventory¶
The canonical definition of the designed surface - the unit Design effort and price are computed from - with the counting rules and sizing guidance, calibrated against five shipped MAJOR sites.
Definition¶
A surface is one designed composition: roughly one screen's worth of stacked Roost Sections, framed and composed by a designer. Everything a project designs resolves to surfaces through the decision table below - nothing prices any other way.
Counting Rules (the Decision Table)¶
| The thing | Surfaces | Rule |
|---|---|---|
| Template at normal density (up to ~10–12 Sections) | 1 | One per distinct page-level design. Content types sharing a layout collapse to one (a news post and an insight article on the same layout are one surface); listing and detail are separate; home counts as its own (a distinctive composition even on the generic template); 404 counts once. |
| Dense or long-scroll template | 1 per ~10 Sections, rounded | A 25-Section page is two-to-three screens of composition work, and counts that way. This is also how a single-page campaign site prices honestly. |
| Named variant (Section level and above) | 1 | The named-variant test: if it ships as a named variant of a Section or template - a variant file, a CMS variation, a distinct frame - it was designed, reviewed, and played back: count 1, even when it looks like "just" a media or layout tweak. Prop-level permutations (a background-colour field, a token theme) count 0. Variants of Elements and Atoms (cards, buttons, list items) count 0. |
| Bespoke form flow | 1 | A form with its own designed layout beyond the pattern kit (a multi-step application, a distinct enquiry flow). Forms assembled from the system's form patterns count 0. |
| Distinct interactive tool | 1–2 | Calculator, map experience, configurator, search experience - by its screens and states. |
| Commissioned library components (the "component menu") | 1 per ~5 Components | Roost Components designed for the library without appearing on any launch page design - editorial-flexibility or roadmap components the client explicitly buys. Strict test: if it appears in a launch page design, it is already inside that surface and counts 0. |
| Components, Elements, and Atoms on designed pages | 0 | Extracted from surface design - inside the surface rate. |
| States, breakpoints, prop-level permutations | 0 | Inside the surface rate. |
| Illustration and animation systems | 0 | What the ambition band's rate carries - mascot sets, animated brand-pattern families, scroll choreography. Two calibrated sites proved this (eight animated illustrations inside an Expressive rate; ten choreographed mascots inside a Premium rate). |
| Design system formalisation | 0 - the 20% line | Documented tokens, variant matrices, usage docs, governance handover: the named design system line, not surfaces. |
Variants are where footprints grow - the largest calibrated site tripled its template count through its variant estate - so the inventory names them explicitly rather than discovering them in production.
Why Pages, Not Sections or Components¶
The tempting alternative - counting Sections or Components directly - is what the portfolio data rejects: component registries varied 21–65 bloks across sites of near-identical design effort (95h vs 120h), because registries accumulate cheap permutations while effort tracks the compositions a designer actually frames. Pages predict; part inventories don't. The two density rules above (dense templates, the component menu) catch the honest exceptions - the 25-Section template and the off-template library - without reopening per-component pricing, which the calibration killed.
The Process¶
- Draft at 2.1 (Architect). The template list falls out of the IA and content model: every page-rendering content type, listing and detail pairs, one-offs, system pages. The principal drafts the variant plan against the 2.3 direction - a brand-led direction signals variant appetite, and pricing it here is what prevents the ten-template site quietly becoming a forty-surface one.
- Add the tools from scope. 2.4's in-scope workstreams and 2.2's integration map surface the interactive tools.
- Price at 2.5. The Cornerstone prompt's Part A audits the inventory for completeness against the IA and scope; the inventory × the band rate is the Design price the Cornerstone fixes. The surface list appears next to the price on the estimate, always. An inventory that lands under the sold assumption redeploys the delta by default (the future-planning list, an added round, the design system line); a lower price is the honest fallback.
- Confirm at chapter start (3.1). Design's first act is confirming the inventory against the Cornerstone. Growth beyond it is a priced variation with the Decider; a surface swap at equal weight is just work.
- Capture actuals per surface - the list is also the calibration instrument (see the effort model).
Estimating Before 2.1 Exists (the Sales Moment)¶
The inventory proper arrives at 2.1, but sales conversations need a number earlier. The five-step estimate:
- Templates from the nav or sitemap (the existing site for a
replatform; the IA sketch for greenfield): home - always its own
surface - plus one per listing/detail pair, distinct fixed pages,
- Deep catalogues collapse hard: hundreds of URLs are usually a dozen templates. This is the floor.
- Four variant-appetite questions, each adding surfaces:
- Audiences - each distinct audience beyond the first tends to demand its own hero and section treatments: +2–4 each.
- Distinct worlds - divisions, sub-brands, careers microsites, IR sections (a structurally separate frame): +3–5 each.
- Future flex - how much gets built beyond the page list the client can point at today, which is what the named variants buy: the treatments that let the site flex later without coming back for new design and build. Close to the list runs ~1.3–1.6× templates in total; room to move ~1.8–2.2×; open-ended 2.5×+. Ask it as headroom, never as taste - "how brand-led is it" invites a compliment rather than a count, and brand expression is the ambition band's axis, not this one.
- Campaign cadence - a marketing team that runs landing programmes: +2–4.
- Enumerate the tools from scope signals - search, locator, calculator, configurator, gated library, portal: 1–2 each.
- Cross-check against the portfolio outline below ("about a CDSDS", "KIMS-scale") and the typical counts.
- The principal owns and approves the count and band on every proposal - sales runs the method; the principal signs the number before it goes out.
- Sell the number as an assumption. It goes in the handover's assumptions as "assumes ~N surfaces at [band]"; 2.1 trues it up, and the delta is a Cornerstone conversation, never absorbed drift. The early number only needs to land in the right band - the computed price corrects at confirmation.
Watch for appetite signals that move the answer: a recent rebrand, a new marketing leadership, or "our sectors all look the same" are cues that a site sitting close to its page list today will not stay there - ask the variant-appetite questions against the site they want, not the site they have.
Suitable Sizings¶
Band boundaries (the size axis, validated against fifteen sized MAJOR sites - see the portfolio outline below):
| Band | Surfaces | Reads as |
|---|---|---|
| S | ≤5 | The smallest site we build: a landing presence, a single-purpose microsite |
| M | 6–12 | Campaign site, microsite, focused brochure |
| L | 13–25 | The full marketing site - the MAJOR default |
| XL | 26+ | The large build: membership, multi-audience, or variant-rich programmes. Priced at a 35-surface representative; materially larger is a bespoke quote |
Typical counts for early conversations (before 2.1 has run, e.g. indicative build estimates at the sales moment):
- Campaign / microsite: 5–8
- Brochure / focused SME site: 8–12
- Full marketing site: 14–20 - the portfolio's most common shape
- Large build (membership, multi-audience, variant-rich): 26+ - observed at 31 and 49
- Multi-site / multi-region estate: 50+ - beyond anything yet shipped. Not a priced tier: there is no honest representative for a shape MAJOR has never delivered. Quote bespoke, off the grid entirely.
Rule of thumb: surfaces ≈ 1.5–2× the template count for most sites (observed: 1.4–2.2×). Challenge that upward when the concepts conversation is brand-led and variant-hungry - the ~5× outlier in the portfolio was a system-led site whose appetite for variants, not its ambition band, drove the footprint.
The Portfolio Outline¶
| Band | Sites (approx. surfaces · production mode) |
|---|---|
| S | The LUC Christmas card (1 · premium), at the micro end - nothing else has come in this small |
| M | MCL (7 · system-led) · Canny Comms (8 · system-led) · Fering (8 · expressive) · MNF (9 · system-led) · Quorum (12 · system-led) · Theatre Workshop (12 · system-led) |
| L | MAJOR's own site (13 · system-led) · Concurrent (15 · expressive) · Into Games (16 · premium) · CDSDS (16 · system-led) · Realis (16 · system-led) · LUC (17 · expressive) · UHV Design (18 · system-led) · Source EV (20 · expressive high) · KIMS (23 · system-led) |
| XL | Musicians' Union (31 · system-led) · Tracsis (49 · system-led) |
Two readings worth keeping: system-led is the house default production mode (eleven of fifteen sites), with Expressive and Premium as the deliberate step up; and every large build to date has been system-led - scale and ambition have not yet arrived on the same project, the combination to price carefully when it does.
Calibration Notes¶
- The named-variant test is a revalidation result, not a preference. A stricter compose test (count only structurally recomposed frames; zero the restyles) was trialled and rejected: a full re-audit of the five calibration sites under it shrank counts 20–40% and smeared the band rates together (system-led observations diverged to 7.3–9.0h; expressive-high became indistinguishable from premium), because shipped restyle variants carry real design time. The named-variant test is what holds the fit: all five sites land at 5.5–10.6h/surface, which the 5.5/7/10.5 rates sit on directly (System-led on Tracsis, Premium on Into Games, Expressive between Concurrent's two readings). Do not re-litigate toward "structural-only" variant counting.
- The two density rules are rate-coherent (the component menu at one surface per five Components implies 1.2h per commissioned component at System-led and 2.4h at Premium, matching market benchmarks) - and both survived the re-audit unfired: exact section-band counts on the static builds top out at 12–13 bands (the ~12-Section ceiling is real), and the largest library-only component set found anywhere was 2, far under the five-per-surface threshold. They exist for the honest exceptions, not the portfolio norm.
- XL is 26+ and is observed - at 31 and 49 surfaces, priced at a 35-surface representative. Nothing is priced above it: a multi-site estate describes scale beyond anything yet shipped, so it carries no tier. Above the representative, XL is quoted bespoke from the actual inventory.
- When the inventory and the concepts disagree (a Premium direction on a 40-surface build), the price conversation happens at 2.3/2.5, where trade-offs are cheap - never discovered in Design.