PUIG · CANVAS
Consumer Intelligence Platformv2.0 · Vision Proposal · Apr 2026
Concept Direction · AI Studio

Canvas.

A consumer intelligence OS for every Puig brand team.
Editorial on the surface. Agentic underneath.
PUIG / AI STUDIO01 / 22
PUIG · CANVAS
Situation02
The Brief, in one line

Brand teams are drowning in dashboards and starving for decisions.

The current prototype proves the ambition — ten feature tiles, one analyst, a roadmap. But it reads as an archive of tools, not a partner that thinks with you.

This proposal reframes the product around one question a brand team actually asks: “What should we do next, and why?”

PUIG / AI STUDIO02 / 22
PUIG · CANVAS
Vision03
Vision statement
Canvas is the thinking surface for Puig — where signal, synthesis and story meet, so every brand team moves from reading the market to composing it.
PUIG / AI STUDIO03 / 22
PUIG · CANVAS
Shift04
From today’s prototype, to Canvas
From — v1 Prototype
A catalog of features.
  • 10 independent modules in a left rail
  • Role banner on every screen (“As a Brand Manager…”)
  • Primary-color data viz — rainbow effect
  • Dashboards that report; the analyst chat is a sidebar
  • Navigation assumes the user already knows the answer they want
To — Canvas
An answer surface.
  • 3 rooms (Listen, Personas, Predict) — plus a Home feed and a Briefs workspace
  • Ambient role context via workspace, never a label
  • Warm-paper palette, single editorial accent, Lettera throughout
  • Agentic layer runs in the background, drafts narratives, flags anomalies
  • ⌘K command palette — ask in words, get a composed answer
PUIG / AI STUDIO04 / 22
PUIG · CANVAS
Contents05
What’s in this document
I
Vision statement
p.03
II
Information architecture & sitemap
p.06
III
UX principles
p.08
IV
Listen — social listening
p.10
V
Personas — synthetic consumer lab
p.13
VI
Predict — Spate-powered forecasting
p.15
VII
Visual design direction
p.18
VIII
Home / dashboard concept
p.20
IX
Scalability & next steps
p.21
PUIG / AI STUDIO05 / 22
PUIG · CANVAS
Chapter II06
CHAPTER · II

Architecture.

Three rooms. One home. One line of words.

PUIG / AI STUDIO06 / 22
PUIG · CANVAS
Sitemap07
Platform structure

Five
destinations.

Collapse 10 feature tiles into 3 “rooms” of work, one ambient Home, and a workspace where decisions get assembled. Every side-trip (Pricing, Ingredients, Competitive) lives as a lens inside a room, not a separate tile.

This is not a rename. It is a mental model: what am I trying to know today?

HOMEtoday’s signals, your agenda
·Brief-worthy moments
·What changed overnight
LISTENwhat the world is saying
·Pulse · Sentiment · Topics · Creators
·Crisis Radar (beta)
PERSONASwho they are — simulated
·Audience Builder · Persona Cards
·Simulation Room (concept testing)
PREDICTwhat’s coming next — via Spate
·Query Console · Trend Radar · Scenario Sandbox
BRIEFSwhere work gets shipped
·Project workspace · Auto-drafted narratives
⌘Kask in words, skip the nav
PUIG / AI STUDIO07 / 22
PUIG · CANVAS
Chapter III08
CHAPTER · III

Principles.

Five rules that decide what we build — and what we don’t.

PUIG / AI STUDIO08 / 22
PUIG · CANVAS
UX principles09
How Canvas behaves
01
Answers, not dashboards
Every screen opens with a one-line reading of what the data means, drafted by the agent. Charts support the sentence — they don’t replace it.
02
Words before clicks
⌘K is the fastest path to anything. “Compare CH Good Girl sentiment in Mexico vs Spain” should resolve without a single menu.
03
Context over configuration
Brand, market, category live in the workspace header. Filters inherit; users never rebuild the same context twice.
04
Surface the anomaly, hide the rest
Stable metrics collapse to a single line. Only what moved gets real estate. Proactive AI alerts replace the “scan the dashboard” ritual.
05
Every insight is shippable
One click turns any finding into a brief page, a Keynote slide, a Slack card, or an email. Intelligence earns its name only when it becomes action.
PUIG / AI STUDIO09 / 22
PUIG · CANVAS
Chapter IV · Listen10
FEATURE · 01 — LISTEN

Listen.

The world says a lot. Canvas tells you which sentences you need to act on this week.

PUIG / AI STUDIO10 / 22
PUIG · CANVAS · LISTEN
Role & journey11
Feature 01 — role in the product

From sentiment score to conversation to own.

Purpose
Convert the firehose of social and search into a short, ranked list of conversations the brand should influence, ignore, or worry about.
Primary user
Comms, Brand, Digital — but designed to be skimmable in 90 seconds by anyone.
Signal surface
Canvas Agent summarizes sentiment shifts, crisis risk, and emerging topics into a morning digest — before anyone opens a chart.
Ideal journey
01 · Morning
Canvas emails 3 sentences
“Sentiment on CH Good Girl rose +0.4 in Spain; ‘refill’ is the driver; two creators worth briefing.”
02 · Open
Pulse strip at the top
Volume, sentiment, crisis risk — three numbers, three arrows. If they’re flat, Canvas says so.
03 · Drill
Topic, not channel
Trending conversations rank by momentum, each with a one-sentence “why.” Posts sit inside each topic.
04 · Act
Send to Brief
Any topic, post or creator becomes a card in a live Brief — with the agent’s draft already inside.
PUIG / AI STUDIO11 / 22
PUIG · CANVAS · LISTEN
Interface concept12
Listen — concept layout (low-fi)
Workspace · CH Good Girl · EU5 · Q1 ‘26
▲ Canvas — this week in Listen
Refill packaging became the dominant positive driver (+41% volume, +0.5 sentiment). Two mid-tier creators on TikTok Spain have outsized engagement; worth briefing.
Mentions, 7d
2.4K
Sentiment
+7.2
Crisis risk
Low
Trending topics
8
Trending conversations · ranked by momentum
Refillable packaging+41%
Fragrance longevity+12%
Clean ingredients+6%
Oud cultureflat
Post stream — per-topic drill-in
PUIG / AI STUDIO12 / 22
PUIG · CANVAS
Chapter V · Personas13
FEATURE · 02 — PERSONAS

Personas.

Audiences you can speak to, not just segment.

PUIG / AI STUDIO13 / 22
PUIG · CANVAS · PERSONAS
Role, journey & concept14
Feature 02 — Persona Lab redesigned

From audience filters to a room of people who answer you.

Purpose
Validate concepts, claims and packaging in hours instead of weeks — by letting brand teams interview synthetic audiences grounded in real panel data.
Shift
The v1 screen is a form + a radar chart. Canvas inverts it: personas are cards you greet, behavior profiles are a consequence, and the simulation is the main surface.
Journey
① Define audience in one sentence  →  ② Meet the 4–6 generated personas  →  ③ Ask them something (claim, price, pack)  →  ④ Compare their answers  →  ⑤ Ship into a Brief.
PERSONAS · CH GOOD GIRL · EU5 · FRAGRANCE
Portrait
Persona 01
Eco-Conscious Mara
28 · Mid-high income · Barcelona
SustainabilityClean beauty
Portrait
Persona 02
Luxury Traditionalist Inés
52 · High income · Madrid
HeritageQuality
Portrait
Persona 03
Gen-Z Trendsetter Zoé
23 · Mid income · Paris
TikTokValue
Ask all 6
“Would you pay €95 for recycled-glass packaging on this fragrance?”
PUIG / AI STUDIO14 / 22
PUIG · CANVAS
Chapter VI · Predict15
FEATURE · 03 — PREDICT (SPATE)

Predict.

Spate’s signal. Canvas’s voice.

PUIG / AI STUDIO15 / 22
PUIG · CANVAS · PREDICT
Role & integration16
Feature 03 — Spate as the engine, Canvas as the cockpit

Don’t rebuild the telescope. Build the observatory.

Spate already covers 20B searches, 60M TikTok videos, and a trend lifecycle model. Canvas should integrate via API — and focus on what Spate doesn’t do well: tying predictions to Puig brands, to a workspace, and to a decision.

Inside Canvas, Predict appears as one module with three modes. Under the hood, they’re the same Spate data surfaced for three different kinds of question.

Mode 01 · Console
Ask about one thing. “Is oud cologne still rising in Italy?” → lifecycle verdict, 12-mo forecast, similar trends.
Mode 02 · Radar
See the map. 2-axis bubble chart (momentum × maturity), filtered to Puig categories. Drill any bubble for Spate detail.
Mode 03 · Sandbox
Combine signals. Stack ingredient + claim + demographic, get a predicted trajectory and a list of competitor moves Spate has detected.
Canvas adds
Brand lens (auto-filters to current workspace) · Narrative draft (agent explains the chart in plain language) · Ship-to-Brief button.
PUIG / AI STUDIO16 / 22
PUIG · CANVAS · PREDICT
Interface concept17
Predict — query console + radar (low-fi)
Workspace · Rabanne · Global · Fragrance · Powered by Spate API
Ask Predict — ⌘K
“How mature is the refillable luxury trend in fragrance, EU5?”
Verdict — Canvas narrative
Growth stage, likely to continue. Momentum +38% YoY. Two competitor filings since Dec. Strongest in Spain, slower in Italy.
12-month forecast · Spate
Lifecycle curve — Spate data surface
Send to BriefOpen in SpateSimulate on Personas
Similar trends Spate has detected
Refillable luxury
+38%
Olfactive layering
+22%
Gender-fluid EDP
+14%
Niche-to-mass
+9%
Solid perfume
−3%
Competitor moves · last 90 days
Chanel — refill launch (Jan). Dior — patent filing (Feb). YSL — packaging test (Mar).
PUIG / AI STUDIO17 / 22
PUIG · CANVAS
Chapter VII18
CHAPTER · VII

Language.

The visual system. Paper, ink, and a single pigment.

PUIG / AI STUDIO18 / 22
PUIG · CANVAS
Visual design direction19
Design language
Canvas
Warm paper tones (PMS 7527 → 419). No white, no pure black. Broad negative space. Editorial margins. The screen should feel closer to a Kvadrat swatch book than a SaaS app.
Type
Lettera for UI (already in the brand). Paralelo / editorial serif for display — headlines, persona names, verdicts. Type size carries hierarchy; color almost never does.
Accent
One pigment, never two. Swapped via Tweaks for brand themes (Rabanne, JPG, CH, Byredo). Accent marks only what matters: the one number that moved, the active tab, the AI’s draft.
Cards & rules
Square corners, 1px hairlines, no drop shadows. A card is just an area of paper defined by its border. Tables over cards wherever possible.
Data viz
Single-hue accent + grayscale. Bubble charts on a printed-grid paper. Sparklines inline with numbers. No rainbow categories.
Motion
Slow, editorial (200-400ms ease-out). Agent messages type in one line at a time — the “intelligence” is felt in the pace, not in spinners.
Icons
Minimal set, 1px stroke, geometric. Used sparingly; labels do most of the work.
AI affordances
A single visual vocabulary for anything the agent authored: a hairline left-rule in accent, a mono eyebrow ▲ CANVAS, italic serif body. Never generative gradients.
Palette · Puig warm grays + pigment8 tokens
Paper 00 · 10 · 20 · 30 · Ink 90 · 100 · Accent · Tint PMS 7527 / 7534 / 7536 / 419
Type · Lettera + editorial serif2 families
Canvas.
The quick brown fragrance jumps over the lazy brief.
MONO / LETTERA · TABULAR NUMS · CAPS
PUIG / AI STUDIO19 / 22
PUIG · CANVAS
Home / landing20
Home concept — “today in your brand”

A morning paper, written overnight by the agent.

The home page is not a tile grid. It is a single-column feed, composed by Canvas each night. Three things, in order:

01 · Today
One short paragraph — what changed, what to do. Written in the voice of the agent. Dated & scoped to the workspace.
02 · Signals
Three cards — one from each room (Listen / Personas / Predict). Each is a headline + sparkline + a single verb (“Investigate”, “Simulate”, “Forecast”).
03 · Your briefs
In-flight projects as a short list — with the agent’s last contribution to each.

The three priority features are the only call-to-actions on home. Everything else is one ⌘K away.

PUIG · CANVAS / HOME · CH GOOD GIRL · EU5 · THU 19 APR
▲ Today · Canvas
Two quiet weeks just ended. Sentiment for Good Girl in Spain rose overnight on refill news, while Dior Sauvage’s price moved +6% in Italy. One creator brief is waiting for you.
Listen
Refill beats longevity as top driver
→ Investigate
Personas
Eco-Mara would pay €95 with recycled glass
→ Simulate again
Predict · Spate
Olfactive layering enters growth stage
→ Forecast
Your briefs
CH Good Girl Refill · launch narrativeCanvas drafted intro
JPG Scandal · Q2 creatorsAwaiting personas
Rabanne Fame Refill · pricing3 comments
PUIG / AI STUDIO20 / 22
PUIG · CANVAS
Scalability21
From brand teams to markets — same product, different workspace

One product. A workspace per team.

Canvas is the same app for a CH global brand lead and a Rabanne market director in Italy — only the workspace in the header changes. Brand ↔ Market is a scope, not a separate product.

Rooms are composable: a market team sees Listen auto-filtered to its geography, Personas seeded from the local panel, and Predict querying Spate’s local data. No market-specific build required.

Module design
Every room is a primitive: input, signal, narrative, action. The same four slots work for Pricing, Ingredients, Comms. New verticals = new templates, not new screens.
Agent memory
Canvas remembers the brand + market + category context per user. A CH Italy lead never has to re-set filters.
Content model
Briefs, persona sets, forecasts, creator lists are all blocks. Any block can be shared, duplicated, versioned across markets.
Role, not title
Permissions and defaults are attached to workspaces. The UI never changes by job title; it changes by what’s in your workspace.
PUIG / AI STUDIO21 / 22
PUIG · CANVAS
Next22
Strategic recommendations
Next 2 weeks
Validate the mental model
6 interviews with brand and market leads against the 3-room structure; test the agent’s morning paper as the home page.
Weeks 3 – 6
Ship the clickable
Prototype Home + Listen + Personas + Predict at hi-fi. Spate API integrated behind the three modes. Measure: time-to-insight, brief cycle time.
Weeks 7 – 12
One brand, end to end
Pilot with a single brand (CH or Rabanne) through launch. Instrument ⌘K and “Ship to Brief” as the two north-star interactions.
Q3 ‘26+
Extend, don’t expand
Add markets, then templates (Pricing, Ingredients, Campaign Debrief) inside the existing rooms — never as new tiles.
Thank you.
PUIG / AI STUDIO · CANVAS · APR 2026
PUIG / AI STUDIO22 / 22