Restricted · Internal
demo-token
The Magnum Ice Cream Company's cultural-moment engine: it finds the moments where a brand has an earned, ownable right to play — across the portfolio's 9 brands, 23 products, 80 markets and 65 moments — then closes the loop by turning the highest-scoring ones straight into a creative brief, routed into MRCL MAKE, where a chosen moment becomes a made asset. Not just where to show up: what to make, and how to get it moving.
Every product, against every moment, in every market — the whole board, not just the highlights. Slice it by brand, product or market, or set a floor and see what's actually worth a conversation.
| Product | Moment | Market | RTP | Earned | Whitespace | Composite | Flags |
|---|
The whole cultural year as one living spine you scroll — Cannes, Coachella, Diwali, the World Cup drawing themselves in month by month, so you feel how close each moment really is long before it needs a plan.
Nine brands, twenty-three products, nine different cultural identities across The Magnum Ice Cream Company portfolio. Here's where each brand's codes actually land — the moments that fit, ranked, not guessed at.
This isn't a new tool built for the pitch — it's a proven one, pointed at a new client. Occasionist is the Groupe's own cultural-moment engine, re-applied here to The Magnum Ice Cream Company's nine-brand portfolio. Same three-dimension read on a moment — right-to-play, earned attention, whitespace — same eye for patterns that repeat across markets. Different client, same rigour.
Ported: the recommender algorithm, score definitions, and Shell chrome pattern from the source engine's own internal documentation. Not ported: live Gemini scoring/chat, the real 75MB data snapshot, PP Engines platform auth, the shader gate — all out of scope for a same-morning static demo.
Every brief Occasionist generates routes into MRCL MAKE — the Human x AI craft system that turns a chosen cultural moment into a made asset.
Why Art, Fashion & Festival — straight from the client's own brand book
Where Art Meets Innovation
Flexing Our DBAs
Crack — DBA
Needs ContentMagnum's own brand book names Art, Fashion and Festival as where it plays beyond the ice-cream cabinet — the same three pillars this tool's THEME_VOCAB and moment library are built around. [NEEDS-DAVE] Source: Magnum 2024 Brand Book Lite — marked "confidential Unilever document, internal use only; check usage rights with the Global Magnum Brand team prior to using" on its own disclaimer page. Used here on the working assumption that referencing a client's own supplied brand assets back to them in a pitch response is standard practice — not yet confirmed with Unilever/Magnum brand team. Flag before this demo link goes beyond the immediate pitch team.
90-second orientation
You give it three things — brand & product, market, window — and it returns ranked moments, scored across three independent dimensions, with a cultural rationale grounded in each product's actual codes. Ported from Occasionist's own framing. The brand row filters the product chips and doubles as a select-all toggle per brand — same grammar as the region buttons on the market picker.
Brief — the strategic recommender, start here. Playbook — the full grid, every product × moment × market. Horizon — the calendar as a timeline. Portfolio — nine brand tiles, each with its products + top moments. About — the one-pager.
Every (moment × product × market) triple is scored on three independent 0–100 dimensions. A high score on one doesn't imply a high score on the others.
Does this product have credible permission to show up at this moment in this market? Driven by brand codes, audience, category logic, and any regulatory or cultural blocks.
If the brand commits 12+ weeks ahead — briefing creative, lining up partnerships — can it actually own the conversation, not just show up?
How empty is the moment of category competitors? High = uncrowded territory. Low = everyone else is already there.
RTP + Earned + 0.5 × Whitespace + theme bonus. The single number rows are sorted by.
A red flag like category-crowded, regulatory-check or away-market caps the score and shows in the FLAGS column. The regulatory-check flag is a live link — click it to open the Food Checker (compliance.html), which preselects that row's market, runs the check and scrolls to the results.
Most of the moment library is tagged to a single home market (e.g. Notting Hill Carnival → UK, Oktoberfest → DE, Vivid Sydney → AU) rather than left global. Scored in that home market, RTP and Earned get a real credibility boost — it's the brand's own turf. Scored anywhere else, the row is deliberately capped below the recommendation floor and marked away-market: activating someone else's local festival from outside is a stretch, not a strength.
When a product × moment scores RTP ≥ 70 in three or more markets at once, it surfaces as a cross-market pattern at the top of the Brief — a signal this isn't a one-market fluke.
The optional Layer in a theme row on the Brief tab lets you overlay a specific campaign idea on top of the product × market × window query — pick theme tags directly (e.g. fashion, sustainability) and the engine intersects them against every moment's own theme tags.
In boost mode (default) every theme overlap adds up to +60 to the composite, so themed moments rise without other strong matches falling out. In strict mode only matching moments are returned. Matched rows show a ★ N tag next to the rationale.
The Ask the Guide button (bottom-right, every tab) opens a chat grounded in two sources: the live SCORES grid (product/moment/market/theme questions) and a corpus of real Magnum Brand Book + Occasionist-mechanic facts (CORPUS_FACTS). It runs a ranked, tokenised retrieval — candidates are scored by evidence rather than first keyword hit, entities are matched on token overlap (so "Milan fashion" resolves to Milan Fashion Week), colloquial phrasing is mapped to the mechanic vocabulary ("how is this ranked?" → the composite formula), and a near-miss falls back to the closest real entities. Timing questions ("what's coming up this summer?", "which fashion moments are in September?") filter the moment calendar the Horizon tab draws and list what lands in the window, soonest first. Try "Top cross-market moments for Magnum Remix", "Why Art, Fashion and Festival?" or "What's on next quarter?". Press ESC to close.
Wiring in a genuinely live model is code-ready but switched off (LIVE_MODEL_ENDPOINT = '') — this is a static site with no backend, so a real connection needs a Cloud Function/Cloud Run endpoint routing to Mars Ollama or the Mercury Claude-Max proxy (never a direct model call from the browser, and never api.anthropic.com, per platform routing rules). Standing that up is new GCP infrastructure and needs sign-off, not a same-session change — until then, every answer comes from the two sources above.
Every row on Brief and Playbook has a Why? button — it opens the full score breakdown, matched themes, home/away-market context and the rationale in one place, no scrolling a wide table. Rows scoring composite ≥ 150 additionally get Create Brief, which drafts a mini creative brief from that row's real data (objective, product fit, local context, suggested activation, KPIs) and a Send step that captures a recipient's name and email.
The recommender math, score bands, theme-bonus formula and cross-market threshold are ported verbatim from the real engine. The data feeding it is a deterministic synthetic generator (not live Gemini scoring), and the chat is keyword retrieval, not a live model call — see the notes on the Brief tab and in the chat panel.
Routed to MRCL MAKE