Fragrance development produces experience. Very little of it becomes knowledge.
Every project runs through cycles of formulation, evaluation and revision — and most of what is learned slips away when the brief closes. ORIZOMIA builds the data, language and AI infrastructure that lets it accumulate.
Infrastructure
SystemsPublications
ResearchAdvisory
ServicesHow much of what the industry learns becomes durable knowledge — and how much evaporates with each “finished” fragrance?
The missing foundations of the olfactory domain, built as systems.
Perfumery has vocabularies, standards and ontologies. None was written to hold a development record. The ORIZOMIA framework — an eight-layer descriptor and governance architecture — is.
Olfactory database
OperationalA versioned relational database that connects the commercial, chemical and regulatory names of ingredients across three languages, with semantic and chemical embeddings. It carries a maintained canon of the 81 labelled fragrance allergens of Annex III to Regulation (EU) 1223/2009, as amended by (EU) 2023/1545. Every design decision is written down.
How a descriptor is decomposedDocument extraction engine
OperationalThe engine reads supplier safety data sheets, allergen declarations and IFRA certificates into verified, structured data. Each field stays anchored to its source passage. Deterministic validators check CAS numbers, composition sums and allergens against the closed EU list. The output flags what needs review; it never declares compliance.
Retrieval over the literature
Internal useA retrieval engine tracks usage rights source by source across the olfactory literature. It grounds answers only in sources the user is entitled to use and refuses to answer when the evidence is thin.
CPA · Process observability
In serviceCycle Project Analysis is a quantitative diagnostic of how development projects move through documents, tracking repeated versions, reopened decisions and informal handovers that leave no trace. The diagnostic yields two independent indices for documentary expansion and the informal load of decisions taken outside the documents, with a technical report and an executive summary. A front-end prototype shows the intended workbench.
Client formulas are not required, and are never added to the database. Documents shared for an engagement are processed in a segregated environment and returned or deleted when it ends.
Sources without a licence remain bibliographic references only.
Artificial intelligence proposes; a human confirms. Every automated step keeps an audit trail.
No compliance verdicts — review signals only. Regulatory responsibility stays where the law places it.
A descriptor is not a family. It is a set of weighted axes.
A material has a CAS number and a regulatory entry. Words describing its smell, including fresh, soapy, leathery and gourmand, have never had a shared, checkable definition. ORIZOMIA is the reference layer underneath that vocabulary.
Different houses mean different things by the same word. The same house means different things in a brief, a panel report and a safety dossier. Nothing translates reliably between English, Italian and French.
Each descriptor is decomposed into weighted sensory axes, linked to materials and to regulatory entries, recorded in three languages, and carried by a database row that names its source and the date it was last verified.
How a descriptor is written down
A descriptor is not filed under a family. It is written as a combination of sensory axes. Each axis has a weight between −1 and 1. The equiv role marks identity; related marks a contribution without defining it. A negative weight encodes an absence: a quality the descriptor requires to be low.
H axes are objective dimensions — edibility, sweetness, temperature, citric character, salinity, diffusion. S axes are perceptual qualities, such as wateriness. P axes are spatial phenomenology, such as airiness.
gourmand — equiv:H20@0.6; related:H05@0.4
Two axes, both positive — a simple decomposition.
marine — equiv:S05@0.7; related:H21@0.5; related:H06@0.4
Three axes around one identity axis — a multi-axis synthesis.
misty — related:H02@-0.4; related:S05@0.4; related:P02@0.4
Inverse coupling: diffusion runs to the left of the origin, because the constraint encodes an absence.
What travels is the decomposition, not the word.
fresco (it), fresh (en) and frais (fr) are surface forms of one canonical descriptor. They map onto an overlapping subset of axes: H06 temperature, H13 citric, S05 wateriness. The word changes at the border; the axes do not.
Harmonising the weights across the three languages is current work, not a finished result.
The allergen layer carries 135 INCI names and 193 constraints in four rule families, covering the whole annex. Each of the 30 regulatory acts records the date it was last checked.
Two asymmetries.
-
The regulatory layer is complete; the lexicon is about half defined.
686 of the 1,328 terms carry a definition in their own language and an active vector. The remaining terms are recorded and not yet defined, and the page above counts them as such.
-
The canonical core converges with two independent references.
The 141 canonical descriptors overlap with two external reference sets built without contact with this one. That convergence is a reason to triage the core rather than enlarge it.
Every row names its source and the date it was verified. A statement with no source does not enter the database.
No source is used without an explicit licence decision on record. Sources whose terms do not permit use are marked and excluded, not quietly ingested; licences incompatible with the project are refused by a constraint in the schema itself.
The system triggers review; it never concludes anything about a regulatory obligation. The decision stays with the person who holds it.
Client formulas are not required, and are never added to the database.
Every change enters an append-only trail. The architecture is documented in 205 decision records, including reversals of earlier decisions.
What it is not.
-
Not the largest odour dataset.
Larger public collections exist on molecule counts, on ingredient catalogues and on descriptor counts. What is claimed here is the typed relations between them.
-
Not a regulatory calculation engine.
No verdict, no certification, no legal arithmetic. The output points at a passage and a date; it decides nothing.
-
Not generally available.
It runs as a working instance, not a product: no public interface, no hosted service, no self-serve access.
-
Not externally reviewed.
No peer review, no outside audit. Where a result rests on an internal check rather than an independent one, the documentation says so.
Four situations this was built for.
-
Houses and suppliers
Whose supplier documentation and internal vocabulary do not line up.
-
Evaluation panels
That need one word to mean one thing across sessions and across languages.
-
Regulatory teams
That need a statement traceable to the source it came from and the date it was checked.
-
Researchers on olfactory language
Who need a structure explicit enough to argue with.
Early access is by conversation, not a sign-up form. Tell us which words and which languages you need to line up.
Thirty-two working papers on why the industry's knowledge stays tacit.
The ORIZOMIA research series examines the olfactory domain as an economic and linguistic system. The abstracts of all thirty-two papers are published below. Full papers are available on request while public release is in progress.
Every white line is a curated, typed relation from the ORIZOMIA knowledge graph. Hover to read it; click a paper to highlight its connections. The graph draws the 115 relations between papers. A further 1,297 typed relations run inside them. Both cover WP 01–16, and the formalisation of WP 17–32 is in progress.
Structure and costs
Structural constraints of the domain, its cost structure, coordination costs and the functional coincidences that shape how fragrance work is organised.
Read the four abstracts WP 01–04
Structural constraints of the olfactory domain
A perfume cannot travel as a file. Its stimulus coincides with the matter that carries it, a difference the paper treats as one of kind rather than degree. Four layers of constraint follow: physical, systemic, socio-cognitive and regulatory-institutional. Recognition and playback are treated as distinct trajectories. ORIZOMIA enters only at the close, as a neutral meta-institutional space for standards, language and pre-competitive infrastructure — together with the conditions under which the whole construction would fail.
The cost structure of the olfactory domain
Why can a perfume not be multiplied at zero cost, like a file? Because matter that enters the process must leave it in some form. The paper separates incompressible material costs, fixed by physics, chemistry and regulation, from compressible systemic costs born of informational friction. Its conclusion, argued in falsifiable terms, places the sector closer to structured chemical manufacturing than to any information industry — with a precise consequence for where the remaining margins lie.
Reducing coordination costs: a structural analysis
Not every cost in the olfactory domain comes from matter: a large share arises from how information is produced, exchanged and validated, and is in principle compressible without touching physical or regulatory constraints. The paper models these informational frictions and sets three criteria for recognising compressible areas. The analysis is logico-structural, not empirical; it states its KPIs and falsification criteria. Eleven operational fields, from formula development to audit, then take the test.
When grammar becomes law: functional coincidence in technical-institutional systems
In highly formalised technical-institutional systems, the line between producing a valid description and issuing a binding rule tends to collapse. The paper formalises functional coincidence, where semantic, epistemic and decisional authority cease to be distinct. It separates Domain-Specific Language from Domain-Specific Authority. The analysis is conceptual-structural: it clarifies, it does not prescribe. The central question is whether a description can keep the first two authorities without acquiring the third.
Taxonomies and language
Olfactory taxonomies and their genealogy, fragrance as social practice, brief ambiguity and vocabulary as shared infrastructure.
Read the seven abstracts WP 05–11
The ORIZOMIA taxonomy: a descriptive, relational and governance system
'Fresh' means one thing in a fine-fragrance brief, another in a panel report, a third in a safety dossier. A dictionary cannot carry that load. ORIZOMIA answers with an eight-layer architecture, from core ontology to ambiguity control, under explicit governance. Two moves set it apart: the reversed sequence Semantics, Ontology, Pragmatics places language before ontology, while ambiguity becomes a predictive object, not an error to correct, on the wager that divergence can be anticipated.
ORIZOMIA and the existing classification systems
Perfumery already has classification systems, arguably enough. The paper compares six systems along seven axes, from functional mandate to governance of change: Edwards' Fragrance Wheel, the French SFP families, houses' internal vocabularies, IFRA categories, Dravnieks' ASTM Atlas and ISO 13299. Each fulfils its own mandate correctly, while the finding lies in the space none covers and in what coexistence with six working systems would mean for import, mapping and dialogue.
Fragrance as social practice
Perfume taxonomies describe their object as a chemical-sensory entity. This paper turns to the substrate of social practices on which each rests. Five dimensions are developed: lexicon as cultural practice, scent as identity marker, embodied perception, ritual traditions and material labour behind the formula. Supply-chain sustainability emerges as a sixth. The thesis is not revisionist: technical taxonomies remain valid in their mandate. What none documents is what it presupposes — the substrate ORIZOMIA proposes to record.
A genealogy of perfume classifications
The classifications perfumery treats as natural orderings of smell are sediments of specific choices, made in specific contexts for specific purposes. The paper traces four phases, from pre-industrial manuals (1709–1860) through the synthetic revolution's named accords (1860–1973) to regulatory institutionalisation (1973 to today). The method is genealogy in Foucault's sense, not revisionism. Each institution answered real problems, and its classifications remain valid within their mandate. The live question is the structural space they leave open.
Pre-competitive coordination in the olfactory domain
Wine, whisky, cocoa, olive oil and coffee have all developed shared descriptor systems; perfume alone has none. The contrast grounds a hypothesis about a sector knowledge commons in the Ostrom-Hess sense, offered as theoretically plausible rather than empirically established because Ostrom's principles transfer only in part and primary research remains future work. The conclusion is methodological: right or wrong, depositing the argument first lets others contest it.
Irreducible ambiguity in fragrance briefs
The ambiguity of a perfume brief, the paper wagers, is not a drafting fault but an intrinsic feature of creative communication where shared vocabulary is structurally poor. Four types operate: lexical, pragmatic, contextual and intentional. Competence lies in knowing which to clarify and which to preserve as interpretive space. ORIZOMIA's L4 and L8 layers offer light scaffolding. The paper declares itself the first deposit in a field no peer-reviewed study has yet examined.
Formalised olfactory vocabulary as company infrastructure
The paper argues for formalised olfactory vocabulary as company infrastructure rather than a composer's tool, with value accruing across R&D, marketing, compliance, training, commercial relations and AI integration, largely through cross-functional effects rather than an additive sum. The supporting theory is verified; its application to perfumery is the paper's proposal. A dedicated section weighs what adopting AI before this linguistic anchor exists would cost — from semantic contamination and technical debt to supplier lock-in.
AI, regulation and limits
Structured olfactory coding for AI, regulatory compliance and machine assistance, contextual olfaction and the limits of AI in perfumery.
Read the seven abstracts WP 12–18
Olfactory AI as a structured-coding problem
No olfactory AI performs beyond its data pipeline, whether graph networks predicting perception from structure, generative language models or recommendation systems. The paper follows the chain from coding quality to model performance to operational decisions, and names the tetrad conditioning all three families: structured coding, calibrated panels, data governance, standardised documentation. Written for CTOs and R&D leaders, it fixes the technology's operating conditions. Each house must still build or recruit specialist competence.
Regulatory compliance and AI in the fragrance industry
AI in the European fragrance industry operates inside four regulatory frames: Cosmetics Regulation 1223/2009, Regulation 2023/1545 on allergens, the IFRA Standards with QRA2, and the AI Act (2024/1689). Their intersection is where an AI-assisted decision carries regulatory weight; the paper maps it and builds the operating framework, completing the triad of WP11's business case and WP12's technical feasibility. Written for compliance directors and CTOs, it precedes the technical investment without replacing it.
Standards, regulations and institutional gaps
The olfactory domain is among Europe's most regulated spaces and poorest in semantic infrastructure, with standards that measure odour and govern its safety and environmental impact but build no ontology, controlled lexicon or governed vocabulary. The paper reconstructs the landscape and positions ORIZOMIA as meta-institutional architecture that integrates and translates rather than competes — a map, for the custodians of olfactory governance, of the space their own mandates leave uncovered.
Contextual olfaction
ISO sensory standards treat all five senses through one device: the neutral booth, calibrated on industrial taste analysis. The paper argues uniformity fails olfaction. Of Stevenson's three functional classes, ingestion, hazard avoidance and social communication, the booth suppresses two. It serves panel calibration but not social practice. The proposed olfactory living lab has three domain profiles and an education pilot. The evidence is exploratory, not causal, pending confirmatory RCTs.
E = F(M, B, P, C): why olfactory experience is not additive
The same molecule yields different experiences depending on biology, memory and context; no single explanatory axis will do. E = F(M, B, P, C) treats olfactory experience as the non-additive interaction of stimulus, biology, psycho-mnestic profile and context, in nine conditional, falsifiable hypotheses tested across three scenarios that expose where single-component models break. The model claims no proven correctness. It states what it asserts and ignores, making it falsifiable.
E = F(M, B, P, C) beyond the checklist
The literature lists molecular structure, receptors, psychology and context without specifying how they combine. It offers a checklist, not a model. Within E = F(M, B, P, C), the paper separates checklist from model using some nonzero mixed partial derivatives for biologically motivated pairs and maps the criterion onto five implementation classes and an empirical agenda. The contribution remains formal. No dataset yet measures all four components at adequate N; building one is the bottleneck.
The architectural limits of AI in perfumery
Debate on AI in perfumery swings between announced revolution and principled scepticism; both treat the limit as one threshold. The paper distinguishes three strata — architectural limits that data can dissolve, epistemic-structural limits that persist however much data arrives, institutional limits revisable by law. An explicit, falsifiable demarcation criterion assigns each constraint, from the roughly 5,000-molecule data ceiling to the perfumer's tacit knowledge — because confusing the strata, the paper argues, misallocates research and distorts expectations.
Aesthetics and knowledge
The autographic paradox, beauty without provenance, categories without foundations and the olfactory lekton — what fragrance language actually refers to.
Read the eight abstracts WP 19–26
The autographic paradox of olfactory art
Olfactory art is not transmitted but performed — on someone's body, at an unrepeatable moment, with a chemistry set by skin, temperature and hour. When panels rate AI-generated formulae level with a senior perfumer's, the anomaly becomes an analytical lever. The essay argues perfumery exceeds Goodman's autographic/allographic divide and needs a third category, the ephemero-corporeal work. Algorithmic generation splits the aura of composing from the aura of experiencing. What remains irreducibly autographic — and for how long?
Without yesterday
A perfumer judges a fragrance on skin over hours; in that span a machine generates more formulae than a lifetime could assess. The essay names what generation does not contain. Creative rupture requires selective forgetting (Nietzsche). Embodied competence takes time speed cannot shorten (Anders). Incorporating practices are transmitted only in co-presence (Connerton). All three presuppose a subject constituted in time. They bear on the conditions of tomorrow's quality, not today's.
Without where
Calabrian bergamot and Mysore sandalwood owe their value to a causal tie of place and chemistry: the molecule carries its terroir's trace. The essay reads this as Peircean indexicality, which generative perfume lacks, as generalisation converts indices into symbols, leaving 'icons of indices', formulae resembling structures once bound to the world without being so. The loss is structural, not a data shortage. The resulting category rests on declared absence and awaits a critical lexicon.
Free beauty, adherent beauty
The trade separates author perfume from market perfume without naming the frame beneath. The essay recovers Roudnitska's 1977 application of Kant's free and adherent beauty. It draws the consequence Roudnitska left undrawn: the evocative brief yields adherent beauty by structure of process, not by aesthetic inferiority. Production confuses categories under shared words. Panels disagree because they never settle which beauty they judge. Beyond the framework lies territory it does not cover, the olfactory sublime.
Apollo in the laboratory
Smell is immediate and name-resistant. A perfumer's training builds the opposite: classification, nomenclature, scales, families. The essay reads the opposition through an Apollonian and Dionysiac correspondence proposed without Nietzsche's authority, then uses genealogy to reopen three hierarchies, natural over synthetic, French over functional and niche over mass, as historical without being arbitrary. The tension, the essay wagers, is irresolvable — and productive precisely because, as in tragedy, it generates a third thing.
Three names for the same shelf
One line of perfumes has three names: 'artistic' in the boutique, 'niche' in distribution, 'fine fragrance' in sales reports. All mean the serious perfume. The essay argues they answer independent questions of application, position and status, while synonymy works as a circular loan: each term covers the gap the others leave, lending a guarantee none grants alone. The three referents still behave as distinct; it is that persistence that ends their tenure as synonyms.
Categories without foundations
Perfumery treats chypre, fougère, 'fresh' and 'woody' as if their families carved an order already present in things. Those families hold up an entire industry while resting on nothing, the essay argues; the fault lies in presenting conventional categories as discoveries when they remain choices, although every trade uses them. Are the families real or invented? The essay's ask is not abolition — only that the trade hold them for what they are.
The olfactory lekton
Naming a smell involves the molecule, its source and a third, elusive thing: what is said of it. The essay places olfactory meaning in the Stoics' lekton, sustained only by collective agreement, so that 'this is a chypre' is judged right or wrong against a norm re-granted at every use, a truth with an expiry date. For digitisation and archiving, a molecule can be fixed. The meaning a community gives it cannot.
Training, ownership, price
How a nose is trained, who owns tacit expertise, the comparative anatomy of the ineffable and the economics of the unsaid in luxury fragrance.
Read the six abstracts WP 27–32
The word on two levels
Terms governing the olfactory product are born in definition and end in positioning: 'alcohol-free' or 'hypoallergenic' carries two truth conditions at once. The paper names this slippage, classifies its forms by the definitional anchor's status, and adds a diachronic form where regulation empties the anchor and the name survives. Elsewhere gaps close with a measurement. The nose cannot verify the claim, so slippage turns permanent and an honest taxonomy inherits the accounting.
The lexicon of ghosts
Words in perfumery's lexicon outlive their referents uncorrected. 'Musk' covers synthetics owing nothing to the deer; 'Mysore' promises an Indian wood on an Australian oil. The essay finds the cause in the sense rather than weak vocabulary. Olfaction offers no authority with a mandate over names, so no word is withdrawn when it turns false. A name is a position materials occupy in time, not a label — which changes what a faithful taxonomy must record.
Training a nose
Perfumery's core competence is partly, irreducibly tacit; the sector teaches the chemistry, which can be written, and leaves the rest to apprenticeship. Although this knowledge resists writing, the essay locates the fault in its lack of scaffolding. A sample-anchored lexicon, a cross-cultural frame and calibrated individual differences would make it examinable and more open, the tacit residue intact. The declared risk: scaffolding built inside proprietary house tools, not as public knowledge.
Who owns the nose
In autumn 2022 the Pantone colours went behind a subscription: the colour survived; its name acquired a price. Colour rested on a public standard. Smell has no shared base of 'primaries'. The new 'odour map' can thus be licensed with nothing open to resist it, while secrecy, data and supply-chain control meet no counterweight. The best reading of odour is already someone's property. What, the essay asks, would a common floor beneath it require?
A comparative anatomy of the ineffable
Every expert trade knows more than it can say. How much becomes public depends on notation and sample-anchored lexicons. Olfaction, the essay argues, comes last on both. Nature bars notation: no handful of 'primaries' composes all odours. Nothing bars the lexicon: wine, mostly smell itself, built its aroma wheel and calibrated panels. The gap is partly institutional, compressible but not eliminable. Scaffolding moves the threshold from secret to code without democratising.
The price of the unsaid
A luxury perfume costs as much as a small appliance; the liquid inside is worth a few euros. The essay rejects the claim that buyers pay for an illusion, treating ineffability as a defended asset comparable on the hedonic plane but kept beyond comparison on rare materials, author-nose and story by the known mechanism of obfuscation. As dupes make the hedonic comparable, where the value sat starts to show — and whom the silence served.
Essays & articles
An eleven-essay series in Italian for a broader industry readership, alongside articles in English written for the international trade press.
Request →Working documents and detailed specifications are available on request.
Measured services for fragrance and cosmetics companies.
Every engagement reads documents a company already has and returns findings anchored to them. Each asks a different question. CPA, below, stands for Cycle Project Analysis. Scope and price are agreed before the work starts, and every engagement produces a documented result. Three things are excluded: judgement about your people, compliance verdicts, and building or running your software.
CPA Screening
A first measurement of one development cycle from its documents counts structural events and reopenings, and shows how the load falls across functions. It returns two indices for documentary expansion and the informal load of decisions taken outside the documents, with a technical report and an executive summary. The two are never added together. It runs on existing material, under NDA.
CPA Pilot
A full measurement of one development cycle locates time and money in repeated versions, reopened decisions and untracked handovers. The result is a technical report and an executive summary.
Allergen readiness audit
Where your allergen declarations and supplier documentation stand against Regulation (EU) 2023/1545 and the applicable IFRA Standards. The extraction engine reads each document and anchors every value to its source passage. A person verifies the findings. What comes back is a list of findings, each tied to its document and passage, and the documents that are missing, expired or unreadable.
Record structure audit
A measurement of the record itself, not of a cycle, and against no regulation. It returns what is recorded, what is not, where the same thing carries three names and what no system could read today, plus a target schema and a gap list. Restructuring the record, and specifying the layer that would connect it to your systems, are scoped from it.
For CPA, a screening before a pilot. For the audits, a scoped first pass before anything larger.
If the measurements do not justify the next step, the report says so.
Terms of engagement Five terms
Scope and price
Scope and price are agreed before the work starts, itemised by hours, people and tools, and there are no success fees. A mutual NDA is signed before any document exchange.
Your documents
Screenings run on ORIZOMIA's tooling and are reviewed by hand. Your documents are processed in a segregated environment, never used to train models, and returned or deleted when the engagement ends.
What is retained
What is kept is the measurement — event counts and index values, without your name on them — so that the next project can be read against something. You can opt out of that in writing.
Contested findings
If you contest a finding, it is registered and answered within three working days. The frozen dataset is never rewritten after the fact. Changes go into the next version of the method.
No contingent pricing
A measurement is bought and delivered on its own. Its price does not change if you commission what it recommends. Its findings do not depend on your doing so.
Start with a conversation.
Enquiries about the systems, the research series or advisory engagements are welcome. We reply within two working days with a concrete next step.
ORIZOMIA is the work of one person. Daniele Saro is an independent researcher and consultant in data and AI for perfumery and beauty; he trained in fragrance evaluation at the Italian Perfumery Institute and came to the sector from finance, with degrees in financial economics (Trieste) and philosophy (Vita-Salute San Raffaele).
To make the first exchange useful
- Your role, and the company or project you represent.
- Which documents you handle — safety data sheets, allergen declarations, IFRA certificates, briefs.
- The deadline you are facing, if there is one.
Everything you share is treated as confidential. A mutual NDA is signed before any document exchange.
Privacy notice.
This site uses lightweight, first-party, cookieless analytics to count page views. We do not use third-party trackers or advertising cookies. The only personal data we process beyond that is what you choose to send us by email.
Data controller
Daniele Saro — ORIZOMIA, Milan, Italy · P.IVA 14630540962 · contact@orizomia.com.
What this site does not collect
The site sets no cookies and runs no third-party analytics, advertising or tracking. The only analytics in use is a lightweight, first-party, cookieless page-view counter, described above. There are no contact forms or embedded third-party content, and fonts and images are served with the page itself, so visiting this site sends your data only to its hosting server.
Hosting logs
The server hosting these pages may record standard technical logs, including IP address, time of access and resource requested, for the security and operation of the service. The legal basis is legitimate interest (art. 6(1)(f) GDPR); logs are kept only as long as those purposes require.
Email correspondence
If you write to contact@orizomia.com, we process your address and the content of your message solely to reply and to manage the exchange (art. 6(1)(b) and (f) GDPR). Messages are kept for as long as the correspondence is relevant, are never added to marketing lists, and are never shared with third parties. Anything you share is treated as confidential from the first message.
Documents shared in engagements
Documents exchanged during an advisory engagement, including safety data sheets, declarations and certificates, are handled under the engagement contract and a mutual NDA. They are kept in a segregated working environment, never used to train models, and returned or deleted when the engagement ends.
Your rights
You may exercise your rights of access, rectification, erasure, restriction, objection and portability (arts. 15–22 GDPR) by writing to the address above. You may also lodge a complaint with the Italian supervisory authority, the Garante per la protezione dei dati personali (garanteprivacy.it).
