We Are Launching our Own NFT! Characterizing Fashion NFT Transactions - Preliminary Results

Full text

We Are Launching our own NFT! Characterizing
Fashion NFT Transactions - Preliminary Results
Sarah Bouraga
Namur Digital Institute (NADI)
University of Namur, Belgium
sarah.bouraga@unamur.be
Abstract—Blockchain technology and Non Fungible Tokens
(NFTs) have been a hot topic for several years now, as proven by
the multitude of brands launching their own NFT projects. In this
paper, we will consider some popular fashion NFT collections,
namely: adidas Originals Into the Metaverse, AMBUSH OFFI-
CIAL POW! REBOOT, Azuki x AMBUSH IKZ, CULT & RAIN
- The Genesis Collection, Dolce& Gabbana: DGFamily, Dolce&
Gabbana: DGFamily Glass Box, Chito x Givenchy NFT, MUGLER
- We Are All Angel, RTFKT x Nike Dunk Genesis CRYPTOKICK,
Prada Timecapsule.
First, we will analyze and examine if we can find salient
characteristics of transactions pertaining to these collections.
Second, we will attempt to propose a first taxonomy of fashion
NFT transactions. From the results, we can state that most
transactions occur at the NFT launch and that they belong to
the Memberships category. Secondly, the results show that we
can propose a taxonomy of four transaction groups or clusters.
The findings can have practical implications for both re-
searchers and practitioners, indeed the results: (i) can be a
stepping stone for future research on (fashion) NFTs, (ii) can
help practitioners analyze transactions using our preliminary
taxonomy.
Index Terms—Blockchain, Non Fungible Token, Fashion, Ad-
dress Classification, Data Analytics
I. INTRODUCTION
Blockchain technology has been a hot topic for several years
now. This technology enables various types of tokens, among
others cryptocurrencies and Non Fungible Tokens (NFTs).
NFTs have been pretty popular in recent years. At first, these
tokens were popular among artists. With time, many brands
have tried launching their own NFTs with more or less success.
Multiple sectors could benefit from this new technology. One
of these sectors is the fashion sector, which could use NFTs to
(i) propose augmented reality clothing, (ii) launch Metaverse
fashion, (iii) provide their customers with digital twins, or (iv)
even open virtual stores 1 2. Case in point, brands like Nike or
Gucci have been able to generate millions of dollars in revenue
using NFTs.
The goals and corresponding contributions of this paper are
twofold. Firstly, we want to identify the salient characteristics
of transactions related to fashion NFTs. Secondly, we want to
propose a first organization of these transactions into distinct
groups. Specifically, this paper will examine the following
research questions:
1https://nftevening.com/luxury-fashion-brands-nft/
2https://tinyurl.com/4e49rb3w
1) RQ1 - What are the characteristics of the transactions
pertaining to NFT released by major fashion brands?
2) RQ2 - How can we organize these transactions in
consistent groups and thereby proposing a preliminary
taxonomy?
In order to address these questions, we will scrape and
collect transactions and their corresponding characteristics
from the launch of the fashion brand’s collection of NFTs to
today. We will then conduct an exploratory data analysis and
clustering. This will allow us to paint a picture of one of the
use cases of Web 3.0 outside the regular “tech” or finance-
related use cases. We will focus on ten popular collections
available on OpenSea: adidas Originals Into the Metaverse,
AMBUSH OFFICIAL POW! REBOOT, Azuki x AMBUSH IKZ,
CULT & RAIN - The Genesis Collection, Dolce& Gabbana:
DGFamily, Dolce& Gabbana: DGFamily Glass Box, Chito x
Givenchy NFT, MUGLER - We Are All Angel, RTFKT x Nike
Dunk Genesis CRYPTOKICK, Prada Timecapsule.
The results will be a first step in addressing fashion NFTs
from a data analysis perspective. Consequently, the findings
will allow us to provide a potential stepping stone for future
research directions regarding fashion NFTs or even NFTs
launched by any type of brands in general. Finally, the results
can be used to analyze past, present and future transactions
related to fashion NFTs.
The rest of the paper is structured as follows. Section II
presents the related work. Section III introduces the method-
ology while Section IV-A describes the sample. Section IV-B
presents our clustering. Finally Sections V and VI discusses
the results and concludes this paper respectively.
II. RE LATE D WORK
The potential of NFTs, and blockchain in general, for
brands has been recognized in the literature, as proven by
the various surveys on the topic. For instance, various studies
explored the potential of NFTs in marketing. Examples include
Hofstetter et al. [1] recognized the potential of NFTs to
challenge traditional marketing, which led to the field of
“crypto-marketing”. The authors discussed the opportunities
offered by NFTs, and highlighted multiple use cases such as
digital ownership, uniqueness and economic value. Also, Peres
et al. [2] discussed the opportunities and threats related to
blockchain use in marketing. Next, Malik et al. [3] focused on
the opportunity blockchain offers for creative industries. The

impact the Metaverse could have on retail [4], on the fashion
value chain [5], and on marketing [6] has been explored.
Finally, Joy et al. [7] provided an overview of how luxury
brands can leverage new technologies, in this case blockchain
(and specifically NFTs and Metaverse), along with artificial
intelligence, machine learning, and virtual reality. The authors
in [1]–[4] also pointed out multiple research directions.
We can find in the literature some papers discussing and
assessing the impact NFTs can have on how brands can
interact with their customers. For example, Belk et al. [8]
discussed consumer perceptions of ownership and of digital art
with a focus on Metaverse. Chalmers et al. [9] discussed the
limitations of traditional marketplaces where creative industry
entrepreneurs operate, and how NFTs can address these lim-
itations. Specifically, the authors emphasized (i) how digital
goods can be copied effortlessly, and (ii) the difficulties for
creators to capture value of their own digital goods. The
authors then identified various opportunities and threats, both
in the long-term and short-term, for the creative industry
entrepreneurs pertaining to NFTs. Lee et al. [10] studied
the role of branded NFTs in building brand engagement in
the Metaverse. The authors used text mining based on data
collected from forums, blogs and communities. This allowed
them to identify six candidate constructs or attributes of
branded NFTs, namely: scarcity, financial value, uniqueness,
prestige, originality, and communication consistency. They
validated these attributes and the impact they have on brand
engagement through a partial least squares-structural equation
modeling. Their findings suggested that branded NFTs’ at-
tributes promote a positive brand attitude, enhancing brand
commitment, NFT purchase intention, and, in turn, active
engagement with NFTs.
Various authors explored or provided frameworks or so-
lutions to take advantage of the potential of NFTs in retail.
Alnuaimi et al. [11] studied the use of NFTs and Ethereum in
the context of fine jewelry and gemstones, to mitigate the vul-
nerability of paper-based certificates. The authors developed
and made available smart contracts for digital certification,
proof-of-ownership, sale history, quality, and proof-of-delivery
in the context of precious gemstones. Using a Design Science
Research methodology, Udokwu et al. [12] proposed a DApp
solution for the authentication of luxury accessories. Pirnay
et al. [13] analyzed the value of NFT from the customers’
perspective. Drawing on Holbrook’s framework for value
conceptualization, the authors provided a way to organize NFT
initiatives into eight types: consumables, speculations, play,
art and memories, status, possessions of rarity, charity, and
community belonging.
Furthermore, some papers explored and provided an anal-
ysis of the NFT landscape. Nadini et al. [14] studied the
NFT market: they identified its statistical properties, they
constructed the network of interactions, clustered objects asso-
ciated with NFTs, and studied the predictability of NFT sales.
The findings show that agents specialize on NFTs associated
with similar objects; and that collections are composed of
visually similar objects. Regarding the predictability of NFT
sales, the results show that sale history and visual features
constitute good predictors for price. Kapoor et al. [15] studied
the factors influencing NFT valuation, including the influ-
ence of social media. Specifically, the authors modeled the
question as binary and as an ordinal classification problem.
The binary classification problem explored whether the NFT
sale is profitable or not, leading to the following two classes:
loss bearing NFT and profitable NFT. The multiclass ordinal
classification problem, on the other hand, focused on the profit
bearing class and explored the extent to which a NFT was
profitable, leading to the following classes: (i) assets selling
between $10 and $100, (ii) assets selling between $100 and
$1,000, (iii) assets selling between $1,000 and $10,000, (iv)
assets selling between $10,000 and $100,000, and (v) assets
selling between $100,000 and $1,000,000. The results show
that the top 10 most important features for the prediction are
originating from Twitter and OpenSea, proving the relevance
of considering both social media and transaction data.
Finally, some authors focused on the use of NFTs in the
fashion sector. Alexander & Bellandi [16] explored the poten-
tial of NFTs on value creation in luxury fashion. Applying
a qualitative methodology, the authors reported the following
findings. First of all, six use cases of NFTs for luxury brands
were identified, namely: (i) to implement a digital identity
in the Metaverse, (ii) to generate hype, (iii) metadata utility,
(iv) to generate new collaborations, (v) attract new consumers,
and (vi) explore new revenue streams. Pandey et al. [17],
discussed blockchain technology and how it can be used in
the fashion industry. The authors highlighted various use cases:
transparency and traceability on the supply chain, fight against
counterfeiting, proof of digital ownership, warranties and
coupons, decrease in operational cost, monitoring of deliveries,
and finally, privacy regarding customers’ or retailers’ data.
III. METHODOLOGY
A. Selected Brands
Various (luxury) fashion brands have launched their own
NFTs since 2021, and some of them have been able to generate
millions of dollars in revenue off their entry in the Web 3.0.
As mentioned in the Introduction, fashion brands are seeking
to capitalize on the NFT trend by proposing: (i) augmented
reality clothing, (ii) Metaverse fashion, (iii) Digital Twins, and
(iv) Virtual stores. The fashion NFT projects are organized into
collections, which contain a number of NFTs.
In this paper, we focus on some of the “Top” fashion
NFT collections, in terms of revenues. We also focus on the
collections available on the marketplace OpenSea 3. Hence,
we are interested in the following collections (note that the
Description was retrieved from OpenSea or the brand website
itself):
•Collection: adidas Originals Into the Metaverse
– Brand: Adidas
– Description: “Into the Metaverse is a collaborative,
co-created NFT project between adidas Originals and
3https://opensea.io

NFT pioneers gmoney, Bored Ape Yacht Club and
PUNKS Comic” 4
•Collection: AMBUSH OFFICIAL POW! REBOOT
– Brand: Ambush
– Description: “The collection of 2,022 POW!® Re-
boot Rings merge virtual and real, granting access to
exclusive drops, experiences, and more” 5
•Collection: Azuki x AMBUSH IKZ
– Brand: Team Azuki
– Description: “The Azuki x AMBUSH tokens can be
redeemed for the physical item” 6
•Collection: CULT & RAIN - The Genesis Collection
– Brand: Cult and Rain
– Description: “The collection comprises of 1406 4K
animated NFT’s matched with identical luxury phys-
ical sneakers” 7
•Collection: Dolce & Gabbana: DGFamily
– Brand: Dolce & Gabbana
– Description: “DGFamily Boxes are digital col-
lectibles (NFTs) that also serve as tiered membership
to the exclusive Dolce & Gabbana universe. DG-
Family members receive exclusive digital, physical,
and experiential privileges as Dolce & Gabbana
pushes the boundaries of luxury and culture in the
metaverse” 8
•Collection: Dolce & Gabbana: DGFamily Glass Box
– Brand: Dolce & Gabbana
– Description: “DGFamily Glass Boxes are redeemed
for a tiered DGFamily Box during an on-chain reveal
accessible” 9
•Collection: Chito x Givenchy NFT
– Brand: Givenchy
– Description: A rare set of 15 unique NFT graphic
designs 10
•Collection: MUGLER - We Are All Angel
– Brand: Mugler
– Description: “A collection of 30 crypto-art col-
lectible, limited to 300 pieces in collaboration with
Marc Tudisco” 11
•Collection: RTFKT x Nike Dunk Genesis CRYPTO-
KICKS
– Brand: Nike
– Description: “RTFKT creates unique experiences
with phygital fashion, sneakers, and digital artifacts”
12
4https://opensea.io/collection/adidasoriginals
5https://opensea.io/collection/ambushofficialpowreboo
6https://opensea.io/collection/ambushikz
7https://opensea.io/collection/cultandrain-genesis
8https://opensea.io/collection/dolce-gabbana-dgfamily
9https://opensea.io/collection/dolce-gabbana-dgfamily-glass
10https://opensea.io/collection/chitogivenchynft
11https://opensea.io/collection/chitogivenchynft
12https://rtfkt.com/wtf
•Collection: Prada Timecapsule
– Brand: Prada
– Description: “Each limited edition physical product
will be accompanied by a gifted NFT” 13
Some NFTs that can be considered more successful - or
that generated more revenue for the brand - did not make the
list because they were not available on OpenSea, but on other
marketplaces. This is the case for instance for Jimmy Choo
and its NFT available via Binance marketplace.
B. Data Collection and Feature Engineering
On each collection page on OpenSea, the platform offers a
link to Etherscan or Polygonscan which allows the reader to
collect data about the given collection.
In May 2023, we collected the following features on
Etherscan14 and Polygonscan 15:Transaction Hash,Block
Number,Unix Timestamp,Datetime,Address From,Address
To,Contract Address,Value In (ETH),Value Out (ETH),Cur-
rent Value,Transaction Fee (ETH),Transaction Fee (USD),
Historical Price ETH (USD),Status,Error Code, and Method.
We also added the following features:
•Days between launch and NFT transaction. This feature
was created by computing the difference between the date
of the NFT launch (based on the date the contract was
deployed on the chain) and the date of the transaction.
•Chain. This feature indicates the chain the NFT was
available on.
•Collection. We added the name of the collection the NFT
traded on the transaction belongs to.
•Category. This column was based on the information
available on the OpenSea collection page. It indicates the
category the NFT belongs to. The potential values are:
Art, Memberships, and PFPs (i.e. Profile Pictures). An
Art NFT represents a piece of digital art. A Memberships
NFT gives access to a physical or digital space. And
a PFP NFT aims to be used as a social media profile
picture, or avatar.
IV. DATA ANALYSIS
A. Sample and Exploratory Data Analysis
Our sample consists of more than 100,000 transactions
related to fashion NFT sales. As mentioned in Section III-A,
we gathered data about ten different collections. For each
collection, we collected the following number of transactions:
adidas Originals Into the Metaverse (60,033 transactions),
AMBUSH OFFICIAL POW! REBOOT (7,554 transactions),
Azuki x AMBUSH IKZ (1,249 transactions), CULT & RAIN -
The Genesis Collection (2,722 transactions), Dolce & Gab-
bana: DGFamily (2,807 transactions), Dolce & Gabbana:
DGFamily Glass Box (10,242 transactions), Chito x Givenchy
NFT (30 transactions), MUGLER - We Are All Angel (441
transactions), RTFKT x Nike Dunk Genesis CRYPTOKICKS
13https://opensea.io/collection/pradatimecapsule
14https://etherscan.io
15https://polygonscan.com

TABLE I: Fashion Brands NFT - Statistics (Retrieved from OpenSea)
Brand # Items Created Creator Earnings Chain Category Total Volume Owners
Adidas 10.9K December 2021 10% Ethereum Memberships 49,208 ETH 8,440
Ambush 2,022 February 2022 7.5% Ethereum Memberships 1,038 ETH 1,202
Team Azuki 40 November 2022 0% Ethereum - 191 ETH 14
Cult and Rain 1,406 February 2022 10% Ethereum Art 343 ETH 643
Dolce & Gabbana 3,879 April 2022 10% Ethereum Art 905 ETH 1,616
Dolce & Gabbana (Glass Box) 1,121 April 2022 10% Ethereum Memberships 3,871 ETH 720
Givenchy 15 November 2021 0% Polygon PFPs 49 ETH 8
Mugler 300 February 2023 10% Ethereum - 105 ETH 164
Nike 14.8K April 2022 10% Ethereum Art 9,068 ETH 7,135
Prada 727 May 2022 10% Ethereum Art 40 ETH 212
(14,938 transactions), Prada Timecapsule (1,216 transac-
tions).
Table I shows some basic information about the collections
that we directly retrieved from OpenSea. We can see that most
collections were created in 2022, that the creator’s earnings
are usually around 10 %, and the total volume in Ether ranges
from 40 ETH to close to 50,000 ETH. The collections are
all deployed on Ethereum except for Chito x Givenchy NFT
which is available on Polygon. The number of items available
range from 15 to close to 15,000, and the number of owners
from 8 to around 8,500.
Table II shows the summary statistics for the collections
deployed on Ethereum. We can see that Value IN is close
to 0.0 ETH and does not go over 0.5 ETH for 75% of the
transactions, this trend is also reflected in the column Current
Value $/ETH. The transaction fees have a mean of around 50
USD (around 0.0272 ETH). Finally, we can also note that on
average, the transactions were recorded the day of the NFT
launch; and 75% of the transactions were made 4 days after
the launch. We can see the same trend for the transactions on
Polygon, i.e. for the Chito x Givenchy collection on Table III,
while the sample is much smaller than the first one.
Figures 1a to 2e show the evolution of the number of
transactions over time by collection. We can see that usually,
the number of transactions is at the highest close to the launch
date and drops dramatically afterwards (Figures 1a, 1c, 1d, 1e,
2b and 2c). Some exceptions can be found in Figure 1b, where
the transactions follow the reverse trend; and in Figures 2a,
2d, and 2e where we can see some more oscillations after the
launch.
Finally, Figures 3a and 3b show the distribution of NFT
categories and methods in our sample. We can see that the
Category “Membership” is the most popular in our dataset,
followed by the Category “Art”. Only the collection “Chito x
Givenchy” proposed Profile Picture (PFPs) NFT. For a number
of transactions, no Category was documented. As far as the
Methods are concerned, we only displayed the methods for
which we had at least 5 transactions, given that 58 different
methods were found in the dataset. The top 5 methods are:
Set Approval For All (n= 31,981), Purchase (n= 30,101),
Early Access Sale (n= 13,323), Transfer From (n= 7,730),
Safe Transfer From (n= 7,518), and Mint (n= 6,035).
B. Clustering
Clustering analysis can be used to address three research
questions: (i) the definition of a taxonomy, (ii) data simplifi-
cation, and (iii) the identification of relationships among the
objects in the data [18].
In this paper, we are aiming for an exploratory study of
transactions related to fashion NFT sales, i.e. we are trying to
propose a preliminary taxonomy or classification of objects.
In order to carry out this analysis, we will use the following
features:
1) Unix Timestamp.
2) Current Value.
3) Method. It is defined on Etherscan as “the function
executed based on decoded input data. For unidentified
functions, method ID is displayed instead.” We believe
this information can be interesting to distinguish be-
tween NFT transactions.
4) Transaction fee in ETH. It is defined as the Gas Price
multiplied by the Gas used in Transaction. It is expressed
in Ether.
5) Historical price of ETH. This feature represents the price
of Ether in dollars.
6) Time between NFT launch and transaction. This feature
was created by computing the delta between the times-
tamp of the transaction and the day the NFT contract was
deployed. We believe this feature can be of interest to
distinguish between “early innovators” and other types
of profiles.
7) Category. The possible values are: Art, Memberships,
and PFPs.
8) Chain. The possible values are: Ethereum or Polygon.
Before applying the clustering algorithm, we standardized
the data and transformed the categorical variables via a One
Hot Encoding, using respectively the StandardScaler and
OneHotEncoding of ScikitLearn [19].
Next, in order to identify the ideal number of clusters, we
computed the inertia for a number of clusters ranging from 1 to
10. We plotted the results and using the Elbow technique, we
concluded that we should continue with 4 clusters. Addition-
ally, we computed the Silhouette Score, defined as “a function
of the intracluster distance of a sample in the dataset, aand
the nearest- cluster distance, bfor each sample” [20]. The
Silhouette score for our chosen configuration, i.e. 4 clusters,

TABLE II: Summary Statistics for the Collections available on Ethereum
Value IN Value OUT Current Value Transaction Fee Transaction Fee Historical Days Since
(ETH) (ETH) $/ETH (ETH) (USD) $ Price/ETH Launch
Count 101,193 101,193 101,193 101,193 101,193 101,193 101,193
Mean 0.1550 0.0 281.4016 0.0272 49.3664 3225.98 28.1379
Std 0.1782 0.0 323.3039 0.0525 95.1539 946.3842 70.7868
Min 0.0 0.0 0.0 0.1 0.1804 994.41 0.0
25% 0.0 0.0 0.0 0.0022 3.9258 2889.72 0.0
50% 0.15 0.0 271.5075 0.0054 9.7087 3876.37 1.0
75% 0.4 0.0 724.856 0.0304 55.0169 3876.37 4.0
Max 5.0 0.0 9064.6 0.9786 1779.0205 4112.35 472.00
TABLE III: Summary Statistics for the Collection available on Polygon
Value IN Value OUT Current Value Transaction Fee Transaction Fee Historical Days Since
(MATIC) (MATIC) $/MATIC (MATIC) (USD) $ Price/MATIC Launch
Count 30 30 30 30 30 30 30
Mean 0.0 0.0 0.0 0.0078 0.0072 1.6133 60.6
Std 0.0 0.0 0.0 0.0150 0.0138 0.4428 124.9526
Min 0.0 0.0 0.0 0.0001 0.0001 0.89 0.0
25% 0.0 0.0 0.0 0.0014 0.0013 1.5 0.0
50% 0.0 0.0 0.0 0.0039 0.0036 1.5 0.0
75% 0.0 0.0 0.0 0.0114 0.0105 1.78 41.25
Max 0.0 0.0 0.0 0.08336 0.0767 2.73 423.0
is 0.4469; while for 3 and 5 clusters, the Silhouette scores are
0.4106 and 0.4378 respectively, validating our decision of 4
clusters.
Finally, we chose the k-means clustering algorithm, we fit
our data and predicted the cluster each transaction belongs to.
We show the salient characteristics of the clusters in Figures
4a to 4d.
The first cluster consists of 36,352 transactions: 36,333 on
Ethereum, and 19 on Polygon. These transactions can also be
broken down according to their category: Art (n= 6,144),
Memberships (n= 30,189), and PFPs (n= 19). The
average duration between the NFT launch and the date of
the transaction is around 5 days (5.1317). Transactions in this
cluster report an average transaction fee of around 10 USD
(10.245). As far as the Methods are concerned, we can find
a majority of Set Approval For All (n= 21061), Transfer
From (n= 5107), Safe Transfer From (n= 4860), and Mint
(n= 3382).
The second cluster consists of 17,819 transactions: 17,809
on Ethereum, and 10 on Polygon; and 14,577 belong to the
Art category, 1,545 belong to the Memberships category, 10
to the PFPs, and 1,687 do not document any category. The
average duration between the NFT launch and the date of the
transaction is around 150 days (148.8362). Transactions in this
cluster report an average transaction fee of around 2.5USD
(2.6896). As far as the Methods are concerned, we can find a
majority of Set Approval For All (n= 10917), Safe Transfer
From (n= 2654), and Transfer From (n= 2616).
The third cluster consists of 7,474 Ethereum transactions:
1in the Art category, 1in no category, and the rest (7,472)
in the Memberships category. The average duration in days
between the NFT launch and the date of the transaction is close
to 0 (0.3613). Transactions in this cluster report an average
transaction fee of around 350 USD (348.8699). As far as the
Methods are concerned, we can find a majority of Purchase
(n= 4786), and Mint (n= 2648).
Finally, the fourth cluster consists of 39,577 Ethereum
transactions. These transactions can be broken down accord-
ing to their category: Art (n= 957), and Memberships
(n= 38,620). The average duration in days between the NFT
launch and the date of the transaction is close to 0 (0.1974).
Transactions in this cluster report an average transaction fee of
around 50 USD (49.719). As far as the Methods are concerned,
we can find a majority of Purchase (n= 25293) and Early
Access Sale (n= 13305).
V. DISCUSSION
A. RQ1 - What are the Characteristics of the Transactions
Pertaining to NFT Released by Major Fashion Brands?
From our sample, we can state that (i) many transactions
belong to the Memberships category, and (ii) most transactions
occur at the NFT launch. Based on the data we collected,
and based on the NFT collections we studied, we can state
that the goals pursued by the fashion brands, when launch-
ing their NFT collections, include: the implementation of a
digital identity in the Metaverse, the generation of a hype
around the NFT, the generation of new collaborations, explore
new revenue streams, and the creation of a proof of digital
ownership. This is in line with the results reported in [16],
[17]. Following the framework proposed in [13], we can say
that the collections we analyzed here belong to three types:
art and memories (CULT & RAIN - The Genesis Collection,
Dolce & Gabbana: DGFamily, Chito x Givenchy NFT, RTFKT
x Nike Dunk Genesis CRYPTOKICK, Prada Timecapsule), and
community belonging (adidas Originals Into the Metaverse,
AMBUSH OFFICIAL POW! REBOOT, Dolce & Gabbana:

(a) Evolution of Transactions over Time - Adidas
(b) Evolution of Transactions over Time - Ambush
(c) Evolution of Transactions over Time - Azuki
(d) Evolution of Transactions over Time - Cult and Rain
(e) Evolution of Transactions over Time - Dolce &
Gabbana Glass Box
Fig. 1: Evolution of Transactions over Time
(a) Evolution of Transactions over Time - Dolce &
Gabbana
(b) Evolution of Transactions over Time - Givenchy
(c) Evolution of Transactions over Time - Mugler
(d) Evolution of Transactions over Time - Nike
(e) Evolution of Transactions over Time - Prada Time-
Capsule
Fig. 2: Evolution of Transactions over Time

(a) Distribution of NFT Categories
(b) Distribution of Methods
Fig. 3: Distribution of NFT Categories and Methods (for
Methods with at least 5 occurrences)
(a) Cluster Analysis - Average Historical Price
(b) Cluster Analysis - Average Transaction Fees (ETH)
(c) Cluster Analysis - Average Current Value $Price/Eth
(d) Cluster Analysis - Average Duration between Launch
and Transaction
Fig. 4: Cluster Analysis
DGFamily Glass Box). We believe that we could consider all
the collections belonging to the possessions of rarity type.
B. RQ2 - How can we Organize these Transactions in Con-
sistent Groups?
As mentioned in Section IV-B, the results of the clustering
indicate that, optimally, the transactions should be grouped
into four clusters, where their salient characteristics are dis-
played on Figures 4a to 4d. In the previous section, we
highlighted the relevant features of each cluster. With these
information, we could characterize these four clusters as:
•Cluster 1 - Early Adopters. Groups transactions
recorded shortly after the NFT launch date, with a rela-
tively low average transaction fee. The transactions here
belong to three NFT categories. We could hypothesize
that these transactions are made by users interested in
the NFT hype but who are also more careful than the
ones present in Clusters 3 and 4.
•Cluster 2 - Crypto Art Aficionados Early Major-
ity. Groups transactions recorded long after the NFT
launch date, with a low average transaction fee. We
could hypothesize that these transactions are made by
users interested in crypto-art (Art is the prominent NFT
category and PFPs is also present).
•Cluster 3 - Price Insensitive Innovators. Groups trans-
actions recorded around the NFT launch date, with a
relatively high average transaction fee. These transactions
pertain almost exclusively to Memberships NFTs. We
could hypothesize that these transactions are made by
users seeking a sense of community.
•Cluster 4 - Price Sensitive Innovators. Groups trans-
actions recorded around the NFT launch date, but with
an average transaction fee lower than Cluster 3’s. These
transactions are made by users interested in Art and
Community.
We should note that we borrowed the terminology Innovators,
Early Adopters, Early Majority from the Innovation literature
(and more specifically the Diffusion of Innovations curve).
However, we do not claim that the groups actually fit the
classical typology found in the innovation literature.
C. Limitations
Although our analysis offers some interesting results, this
work suffers from two main limitations.
Firstly, we deliberately restricted our analysis to OpenSea
and only analyzed 10 NFT collections. It would be ideal to
extend the analysis to a greater number of collections. While
we believe we have a fair view of the current situation with
our sample and that the results will not change dramatically
when expanding the analysis to other platforms, it would be
interesting to test whether our intuition is validated or not.
Secondly, we only evaluated one clustering algorithm and
we optimized only one parameter, namely the number of
clusters. In the future, we should test different models of
clustering and also various configurations of parameters in
order to provide better results. Nevertheless, the application of

the k-means algorithm allowed us to propose a first taxonomy
of transactions.
VI. CONCLUSION
In this paper, we aimed to study transactions pertaining to
fashion NFTs. We collected transaction data about 10 NFT
collections: RTFKT x Nike Dunk Genesis CRYPTOKICKS,
Prada Timecapsule, Chito x Givenchy NFT, Dolce & Gab-
bana: DGFamily Glass Box, Dolce & Gabbana: DGFamily,
adidas Originals Into the Metaverse, MUGLER - We Are
All Angel, AMBUSH OFFICIAL POW! REBOOT, and Azuki
x AMBUSH IKZ. We then tried to paint a picture of these
transactions and tried to propose a preliminary grouping of
these transactions.
Specifically, we addressed two research questions:
1) RQ1 - What are the characteristics of the transactions
pertaining to NFT released by major fashion brands?
2) RQ2 - How can we organize these transactions in
consistent groups and thereby proposing a preliminary
taxonomy?
Firstly, we proposed an overview of the transactions related
to our topic here, namely fashion NFT. We reported, among
other features, the transaction fees (both in ETH and in USD),
the chain the NFT was available on, the category of NFT,
and the number of days between the NFT launch and the
transaction date (RQ1).
Secondly, the results show that we can propose a taxonomy
of four groups or clusters. We labeled these four groups: (i)
Early Adopters, (ii) Crypto Art Aficionados Early Majority,
(iii) Price Insensitive Innovators, and (iv) Price Sensitive
Innovators. We described the characteristics of each group
(RQ2).
The practical implications of this paper are twofold. On
the one hand, these results can be a stepping stone for future
research on fashion NFTs or on NFTs in general. We believe
that this work contributes to the body of knowledge on
blockchain. On the other hand, the findings can help practition-
ers understand the landscape of top fashion NFT by reading
the past and upcoming transactions from the perspective of
our proposed taxonomy. For example, the results show that the
level of interest - or at least the level of transaction activity -
decreases quickly after the launch. This insight can be useful
for brands, when they reflect on the strategy behind their NFT.
Future work will consist in addressing the limitations men-
tioned in Section V. Additionally, we believe an analysis of the
transactions per token could also generate interesting results.
In this work, we purposefully merged all the transactions in
order to propose a fashion NFTs taxonomy, and thereby have
an overview of these NFTs as a whole. A finer-grained analysis
could also provide another insightful perspective.
Data Availability. The data are available upon request.
REFERENCES
[1] R. Hofstetter, E. de Bellis, L. Brandes, M. Clegg, C. Lamberton, D. Reib-
stein, F. Rohlfsen, B. Schmitt, and J. Z. Zhang, “Crypto-marketing: How
non-fungible tokens (nfts) challenge traditional marketing,” Marketing
Letters, vol. 33, no. 4, pp. 705–711, 2022.
[2] R. Peres, M. Schreier, D. A. Schweidel, and A. Sorescu, “Blockchain
meets marketing: Opportunities, threats, and avenues for future re-
search,” 2022.
[3] N. Malik, M. Y. Wei, G. Appel, and L. Luo, “Blockchain technology for
creative industry: Current state and research opportunities,” International
Journal of Research in Marketing, 2022.
[4] K. Yoo, R. Welden, K. Hewett, and M. Haenlein, “The merchants of
meta: A research agenda to understand the future of retailing in the
metaverse,” Journal of Retailing, 2023.
[5] S. Periyasami and A. P. Periyasamy, “Metaverse as future promising plat-
form business model: Case study on fashion value chain,” Businesses,
vol. 2, no. 4, pp. 527–545, 2022.
[6] M. Khatri, “Revamping the marketing world with metaverse–the future
of marketing,” Int. J. Comput. Appl, vol. 975, p. 8887, 2022.
[7] A. Joy, Y. Zhu, C. Pe˜
na, and M. Brouard, “Digital future of luxury
brands: Metaverse, digital fashion, and non-fungible tokens,” Strategic
change, vol. 31, no. 3, pp. 337–343, 2022.
[8] R. Belk, M. Humayun, and M. Brouard, “Money, possessions, and
ownership in the metaverse: Nfts, cryptocurrencies, web3 and wild
markets,” Journal of Business Research, vol. 153, pp. 198–205, 2022.
[9] D. Chalmers, C. Fisch, R. Matthews, W. Quinn, and J. Recker, “Beyond
the bubble: Will nfts and digital proof of ownership empower creative
industry entrepreneurs?” Journal of Business Venturing Insights, vol. 17,
p. e00309, 2022.
[10] C. T. Lee, T.-Y. Ho, and H.-H. Xie, “Building brand engagement in
metaverse commerce: The role of branded non-fungible toekns (bnfts),”
Electronic Commerce Research and Applications, vol. 58, p. 101248,
2023.
[11] N. Alnuaimi, A. Almemari, M. Madine, K. Salah, H. Al Breiki, and
R. Jayaraman, “Nft certificates and proof of delivery for fine jewelry
and gemstones,” IEEE Access, vol. 10, pp. 101263–101 275, 2022.
[12] C. Udokwu, R. Zimmermann, A. Norta, P. Brandtner, A. Kormiltsyn, and
S. M. Aroh, “Exerting qualitative analytics and blockchain requirement-
engineering in designing and implementing a luxury products authenti-
cation system,” Inventions, vol. 8, no. 1, p. 49, 2023.
[13] L. Pirnay, C. Deventer, and V. A. de Sousa, “Providing customer
value through non-fungible tokens: A preliminary study,” in Hawaii
International Conference on System Sciences (HICSS), 2023.
[14] M. Nadini, L. Alessandretti, F. Di Giacinto, M. Martino, L. M. Aiello,
and A. Baronchelli, “Mapping the nft revolution: market trends, trade
networks, and visual features,” Scientific reports, vol. 11, no. 1, p. 20902,
2021.
[15] A. Kapoor, D. Guhathakurta, M. Mathur, R. Yadav, M. Gupta, and
P. Kumaraguru, “Tweetboost: Influence of social media on nft valuation,”
in Companion Proceedings of the Web Conference 2022, 2022, pp. 621–
629.
[16] B. Alexander and N. Bellandi, “Limited or limitless? exploring the
potential of nfts on value creation in luxury fashion,” Fashion Practice,
vol. 14, no. 3, pp. 376–400, 2022.
[17] R. Pandey, A. Harbola, D. Negi, P. Verma, and A. S. KUMAR,
“Blockchain and its relevance in fashion industry,” Ilkogretim Online,
vol. 20, no. 3, pp. 3198–3205, 2021.
[18] J. F. Hair, “Multivariate data analysis,” 2009.
[19] L. Buitinck, G. Louppe, M. Blondel, F. Pedregosa, A. Mueller, O. Grisel,
V. Niculae, P. Prettenhofer, A. Gramfort, J. Grobler, R. Layton, J. Van-
derPlas, A. Joly, B. Holt, and G. Varoquaux, “API design for machine
learning software: experiences from the scikit-learn project,” in ECML
PKDD Workshop: Languages for Data Mining and Machine Learning,
2013, pp. 108–122.
[20] L. Igual, S. Segu´
ı, L. Igual, and S. Segu´
ı, Introduction to data science.
Springer, 2017.