This page describes how DomiDo scores and ranks designs in the gallery and how it evaluates the use-case categories the platform can enable. DomiDo is built by Avvyland Limited (UK) and sells universal blocks and fasteners only; every construction shown on the platform is a user-generated design built from those blocks. Scoring serves two related purposes that work together: it ranks listings in the gallery for discovery, and it helps interpret which use-case categories the universal block system serves best as supply grows. The scored items are user-generated use-case categories that the platform enables, never finished products: DomiDo does not sell finished constructions and there are no "hero products" at launch or ever — only universal blocks and fasteners. The scoring framework combines a portfolio-evaluation model that grades a category against fourteen weighted criteria with a per-listing gallery ranker that combines listing-quality, engagement, recency, diversity, moderation, and category-fit signals into a single ranking score.
Scoring is the analytical layer that turns a large catalogue of user-generated designs into a coherent gallery. It supports gallery ranking — given a query or a default browse, which listings come first? — and it supports merchandising decisions about which categories deserve curated landing pages or featured slots. It also supports portfolio understanding by surfacing which use-case categories the universal block system serves best from a geometry, market, and manufacturability perspective, and it supports demand-signal interpretation: when interest reservations or pre-orders concentrate around certain categories, the scoring framework helps explain which features of those categories drive the concentration. Scoring is not a marketing claim about specific products, and the scored items here are use-case categories that the user-generated content (UGC) ecosystem can populate rather than products the platform sells.
The portfolio-evaluation scoring framework uses fourteen weighted criteria. The weights are set so they sum to one hundred percent, and the full catalogue with weights and descriptions is below.
| Criterion | Weight | Description |
|---|---|---|
| Custom-fit necessity | Fourteen percent | How much does each installation require unique dimensions or configuration? Higher means the AI design pipeline adds more value. |
| Market-demand volume | Ten percent | Search volume, market size, and growth trajectory for the category. |
| Competition intensity | Ten percent | A higher score means the category is fragmented or has no dominant incumbent. |
| Price-gap opportunity | Eight percent | The gap between cheap-and-poor alternatives and expensive-and-good alternatives that the DomiDo proposition can fill. |
| Block-system superiority | Eight percent | How well the DomiDo block system outperforms alternative materials (wood, metal, concrete, fabric). |
| AI design value-add | Ten percent | How much the AI-assisted design experience improves the buyer's purchase decision. |
| Repeat and expansion potential | Seven percent | The likelihood of follow-on purchases (more kits, expansions, related categories). |
| Visual and viral appeal | Six percent | The shareability and social-media potential of finished installations. |
| Assembly feasibility | Four percent | Whether a non-technical builder can assemble it in a few hours. |
| Block-count sweet spot | Three percent | The category sits in the middle of the kit-size distribution rather than far above or below it. |
| Shipping feasibility | Three percent | Weight, dimensions, and logistics practicality. |
| Regulatory simplicity | Two percent | Freedom from permits, certifications, and compliance burdens. |
| Self-containment | Eight percent | Whether the category can be built entirely from DomiDo blocks and connectors without external parts. |
| Geometric compatibility | Seven percent | Whether the category's natural form works well with blocky, voxelised geometry. |
Two criteria deserve a closer look because they materially shape which categories belong on DomiDo at all. Self-containment is critical because the DomiDo system is universal blocks and fasteners: a category built almost entirely from blocks scores high, while a category that needs significant external parts — pond liners, plumbing, electrical, wire mesh, glass — scores lower because the kit is then a frame for something else rather than the thing itself. A score of ten means one hundred percent DomiDo with no external parts, eight means ninety-five percent and above with a small consumable like paint or a soil liner, six means seventy to eighty percent with some external parts, four means fifty to seventy percent with significant external parts, two means thirty to fifty percent, and one means blocks are a minor component. Geometric compatibility captures how naturally a category's form maps to blocky geometry: a score of ten means naturally rectangular shapes such as walls, boxes, and borders, eight means it works well as blocky with a slight compromise, six means a blocky look is unusual but marketable, four means it looks wrong when blocky and would need curves or smooth surfaces, two means thin or fine elements that blocks cannot achieve, and one means fundamentally incompatible.
Scored categories fall into tiers that are used to plan curation and seeding, not to make any "hero product" claim. The full tier model is summarised below.
| Tier | Description |
|---|---|
| Tier A: launch priority | The categories with the strongest portfolio scores. These are the natural seed targets for the public gallery so visitors can find inspiration that matches a strong fit. |
| Tier B: strong candidates | Categories with good scores that benefit from designer creativity. These are useful seeding targets after Tier A is established. |
| Tier C: viable | Categories that fit the block system but are not the strongest fits. They are part of the broader gallery rather than the curated entry points. |
| Tier D: marginal | Categories with significant external-part dependencies or geometric mismatch. They appear in the gallery if designers create them, but the platform does not curate around them. |
| Tier E: weak or eliminated | Categories that score poorly enough that the platform actively discourages them in safety and use-case copy. |
| Tier F: not viable | Categories the platform does not support, typically due to safety, regulatory, or fundamental incompatibility issues. |
Categories that score in the top tier share a recurring pattern: they are predominantly rectangular, predominantly self-contained, have meaningful custom-fit demand, and have an active competitor landscape where DomiDo can offer a clear differentiator. Examples include raised garden beds, heat-pump and air-conditioning covers, privacy screens, decorative walls, modular play structures, wheelie-bin shelters, branded retail displays for events, balcony privacy screens, garden borders, restaurant patio dividers, modular feature walls, seating walls, and step risers. Categories that score lower typically need external parts (water features, outdoor kitchens, chicken coops with mesh, hot-tub surrounds) or are dominated by curves or thin elements (wedding arches, ornamental bridges, garden stools, tomato cages). The scoring framework is dynamic — categories rise and fall as the platform learns more about geometry, market, manufacturability, and the user-generated designs being created — and it is extensible: new criteria can be added when the team observes a pattern that the current criteria miss.
Gallery ranking layers per-listing scoring on top of the portfolio-evaluation framework. Per-listing scoring combines listing-quality signals — completeness, image and 3D-viewer fidelity, description quality, and the owner's trust signal — with engagement signals such as views, saves, shares, time-on-listing, and conversion to interest reservation or pre-order. A freshness multiplier lifts new listings while preventing them from drowning older quality, a diversity penalty stops the gallery from becoming dominated by a single designer or category, moderation status gates default ranking so that only listings that pass moderation appear and reported listings are re-scored according to the disposition, and the portfolio category score is one input to listing ranking but is not the only one.
Freshness handling gives a new listing a temporary boost so it can accumulate engagement signals; the multiplier decays exponentially over weeks, and without freshness handling the gallery would ossify around the earliest listings. Anti-gaming protections include detection of view inflation, save-and-share spam, coordinated engagement, and self-engagement by the listing owner; suspicious signals are discounted in scoring and surfaced to moderation. Personalisation is light at launch and tasteful at scale: the active signals are jurisdiction (filter to legal use cases), saved preferences (favourite categories), and prior session behaviour (categories already viewed), and personalisation never overrides moderation, safety, or category-fit constraints. Diversification ensures that a single designer or category does not dominate the default browse, with a diversification rule that re-orders the top results so that the same designer or category does not appear too close together in the ranking.
The scoring model is a weighted linear combination of normalised signals that produces a single ranking score per listing. The model is easy to inspect, easy to debug, and easy to evolve; a future move to a learned-ranking model is possible without rewriting the application, and the abstraction layer between signals and the ranker is explicit.
The diagram shows the two parallel paths into the ranker. A listing flows through signal extraction — quality, engagement, recency, diversity, and moderation — while its category is graded against the portfolio framework to produce a portfolio score; both inputs feed the ranker, which combines them as a weighted combination, and the ranker's output drives the listing's position in the gallery. Inputs to the ranker are the signals listed above, and outputs are the gallery rank, the per-category curation set, and an explanation trace for any listing that lets the operator understand why it ranks where it ranks.
Scoring evolves through three channels. Criterion weights shift as the team learns which criteria predict user behaviour; new criteria are added as new patterns emerge, such as seasonal demand or category bundling; and as the scale of the data justifies it the ranker can move from a weighted linear combination to a learned model, with the same input shape and the same explanation requirements. Merchandising is the operational expression of scoring: Tier A categories drive curated landing pages and featured slots, Tier B and Tier C categories appear in the default browse, Tier D and Tier E categories are still browsable but not promoted, and Tier F is excluded by safety or regulatory copy. Phase A interest reservations and Phase A.5 no-capture pre-orders flow into the scoring model as engagement signals; their weight is significant because they are the strongest demand signals the platform has at this stage, and the conversion-rate aggregate per category — listing view to reservation, reservation to pre-order — feeds the merchandising decisions of the operations team.