We reveal
what matters.

GOLDEN CASE 01

From study
to machine.

A TikTok Shop Beauty benchmark that became a reproducible analysis system — not a disposable deliverable.

Deck commercial · Kalodata / GMV data

THE PROBLEM

A traditional study
dies on delivery.

Source → Excel → a few charts → slides → done. The answer can be right. The asset still disappears.

The format isn’t the problem. The lack of capitalization is.

INSIGHT / CONCENTRATION

Ranking
isn’t enough.

In the Top 200 Beauty France panel, the Top 10 concentrates about 41.7% of observed revenue, the Top 50 about 75.7%.

Observed panel · a ranking presence is not an economic weight.

INSIGHT / CREATORS

5.3% of relationships.
25.3% of GMV.

Creator relationships tied to the seller make up 5.3% of volume but 25.3% of observed creator GMV.

Observed panel · descriptive finding, not a universal law.

BEYOND SCROLLS

Change the question.
Keep the machine.

The TikTok Shop demonstrator opens a family of cases: market studies, pricing, portfolio, market entry, strategy.

Same architecture; new sources, entities, metrics and decision rules.

GOLDEN CASE / 01

Got a market
question?
Let’s build the system.

Today, to answer it. Tomorrow, to answer the next ones.

HOCUS / GOLDEN CASE

Before AI,
look at
the terrain.

How a Data & AI diagnostic turns a vague ambition into an executable path.

Deck commercial · decision-oriented diagnostic

THE PROBLEM

Many AI ideas.
Few trajectories.

An organization wants to automate or understand its customers better. But data, process and constraints aren’t aligned yet.

The diagnostic makes prerequisites visible before picking a technology.

THE DIAGNOSTIC

Seven axes to
assume nothing.

Strategy, data estate, infrastructure, skills, existing AI usage, governance and execution capacity.

Maturity is a decision tool, not a decorative score.

USE CASES

Prioritize
what can survive.

Each case is scored on value, data readiness, risk and adoption path.

Quick win, structuring project or no-go: all three are valid outcomes.

THE ROADMAP

The diagnostic ends
with an order of things.

A reliable near-term scope, then a data foundation, governance and recurring use cases to structure.

The roadmap ties actions, owners, prerequisites and success criteria together.

HOCUS / DIAGNOSTIC DATA & AI

Before building
the model,
let’s build the path.

A data or AI question first deserves readable terrain, verifiable evidence and a possible decision.

HOCUS / USE CASE / 001

A card
is a system.

91,515 horse-racing profile cards, one entity graph, one bounded narrative layer — the vertical is just the demonstrator.

Case: Entity-360 · production data, aggregated & anonymized

THE PROBLEM

Rich data.
A hard read.

A racing catalogue holds history, form, rivalries, comments and pedigree. The challenge isn’t writing more text — it’s making the signal legible at the right moment.

Content must follow business reality, not the other way round.

THE METHOD

The model doesn’t calculate.
It verbalizes.

Facts are prepared upstream in a structured base. The model receives a bounded context and produces a controllable narrative — the card is then stored, not regenerated on every read.

Fidelity is a property of the architecture, not a prompt setting.

ENTITY-360

The card is only
the visible surface.

One pivot table gathers current state, aggregates and proximities. It becomes a contract shared by recommendations, the API and the content layer.

One entity. Several readings. Zero duplicated logic.

AT SCALE

Industrialized.
Still under control.

91,515 cards generated and cached, refreshed only when a business event justifies it — expert-validated on sampled review.

Figures cover content generation only, not total operating cost.

ATELIER / LORE

One vertical.
One general pattern.

The same engine becomes Customer-360, Company-360 or Asset-360 elsewhere. This is the proof behind Hocus Lore — contextual content at scale.