User Value in Modern Payment Platforms A Graph Approach Laura Arditti

2025-05-06 0 0 1.35MB 8 页 10玖币

侵权投诉

User Value in Modern Payment Platforms:

A Graph Approach

Laura Arditti

Larus Business Automation

Venezia, Italy

laura.arditti@larus-ba.it

Martino Trevisan

University of Trieste

Trieste, Italy

martino.trevisan@dia.units.it

Luca Vassio

Politecnico di Torino

Torino, Italy

luca.vassio@polito.it

Alberto De Lazzari

Larus Business Automation

Venezia, Italy

alberto.delazzari@larus-ba.it

Alberto Danese

Nexi

Milano, Italy

alberto.danese@nexigroup.com

Abstract—Payment platforms have signiﬁcantly evolved in

recent years to keep pace with the proliferation of online and

cashless payments. These platforms are increasingly aligned with

online social networks, allowing users to interact with each

other and transfer small amounts of money in a Peer-to-Peer

fashion. This poses new challenges for analysing payment data,

as traditional methods are only user-centric or business-centric

and neglect the network users build during the interaction. This

paper proposes a ﬁrst methodology for measuring user value

in modern payment platforms. We combine quantitative user-

centric metrics with an analysis of the graph created by users’

activities and its topological features inspired by the evolution

of opinions in social networks. We showcase our approach using

a dataset from a large operational payment platform and show

how it can support business decisions and marketing campaign

design, e.g., by targeting speciﬁc users.

Index Terms—Payment Network, Consumer Behavior, Graph

Mining, User Value, Centrality

I. INTRODUCTION

Various forms of online payments have emerged in the

landscape of payment systems to support the rapid growth of

e-commerce. They are already widespread, often surpassing

more traditional payment methods [1], [2]. Online payments

provide users with a fast, direct, convenient, and secure way to

access their funds and conduct transactions, which is among

the main reasons for the success of e-payment technologies in

terms of their widespread adoption [3]. In particular, mobile

payment services are growing in popularity as they keep pace

with the shift toward massive mobile internet access [4]–[6].

At the same time, mobile payment platforms are becoming

more and more community-oriented: peer-to-peer (P2P) pay-

ment platforms [7] offer users new ways to interact with each

other to the point where they are evolving into a new form of

online social network [8].

Great efforts are made to ensure the success of a mobile

payment solution. In particular, companies aim to expand the

The research leading to these results has been funded by the European

Union’s funded Project INFINITECH under Grant Agreement No. 856632

and by the SmartData@PoliTO center for Data Science technologies.

user base and increase user engagement. Common approaches

to these tasks focus on designing incentives and reward mech-

anisms to keep existing users active and attract new valuable

users [9], [10]. When developing strategies to improve the

quality of the user base, one of the biggest challenges is

understanding users’ value and measuring it operationally, as

this is critical for business decisions such as designing mar-

keting campaigns. Classical user-centric approaches measure

the value of users only based on their individual activities.

However, they can hardly capture their quality in an environ-

ment characterized by a network of interactions, which poses

severe limitations to data analytics. Indeed, there is a lack of

methodologies to analyze user behavior in this new generation

of payment platforms, where interactions between users play

an essential role. It is therefore crucial to develop effective

metrics to guide business decisions and support the work of

marketing departments. Mobile payment platforms can be seen

as constantly expanding networks through a “member-get-

member” mechanism. Hence a good measure of users’ value

must capture not only their spending habits, but also how each

user contributes to the creation of a high-quality network.

In this paper, we propose a methodology that models

payment platforms as networks and exploits their structure

to guide business strategies. We overcome the limitation of

current metrics for quantifying user value in payment sys-

tems by leveraging relationships between users. Inspired by

graph mining methods, our approach combines user-centric

features with topological features extracted from the payment

graph. We also present a practical way to compute this new

metric using an iterative graph algorithm. To the best of our

knowledge, we are the ﬁrst to propose a practical method for

measuring user value in current payment platforms, which are

characterized not only by traditional activities (purchases from

merchants) but also by P2P interactions between users.

We present our methodology using a dataset of a real-world

payment network. More speciﬁcally, we present the business

case of YAP, Nexi’s mobile payment platform, which allows

users to perform payments at physical and online merchants

arXiv:2210.11168v1 [cs.SI] 20 Oct 2022

P2P

P2B

INVITES

P2P

P2B

P2P

P2B

INVITES

P2P

P2B

INVITES

P2P

INVITES

P2P

P2B

P2P

P2B

P2P

P2B

INVITES

P2P

INVITES

P2B

INVITES

P2P

Fig. 1: A small portion of the dataset. Users and merchants are

represented as orange and blue nodes, respectively. Invitations

among users are represented with bold orange links, P2B

transactions correspond to blue links, P2P money transfers are

displayed in green.

and to exchange money with each other. We applied our

methodology to analyze the YAP platform’s community, with

the goal of supporting business decisions. Our results show

that the approach is a practical tool to support marketing

campaigns and, more in general, business decisions.

II. DATASET AND ITS GRAPH REPRESENTATION

To develop and evaluate our methodology, we take a data-

driven approach and use as a reference a dataset collected

from an operational payment platform. The dataset comes

from the Italian app YAP1, a payment platform provided by

Nexi2, one of the biggest European players in digital payments.

YAP is based on a mobile application linked to a prepaid

card (accepted by online and physical stores) that also allows

its customers to exchange money with friends and contacts

without fees. In this paper, we use data from the production

databases of YAP, which include a set of transactions for the

years 2019, 2020 and 2021, as well as metadata about users

and merchants.

The dataset can be naturally represented in terms of a

heterogeneous graph, since there are entities that are related

to each other. In particular, we have identiﬁed three types of

relationships that reﬂect the three main types of interactions

between users and merchants.

1https://www.yap-app.it

2https://www.nexigroup.com/en/

1) Users are connected to merchants by “P2B” relation-

ships, representing monetary transactions characterized

by their date, amount and channel, which may be online

(i.e., e-shops) or ofﬂine (i.e., physical stores).

2) Users may transfer money to other users. This kind of

interaction is represented by “P2P” relationships among

users, which are characterized by their date and amount.

3) Finally, users may invite new users to join the platform.

This results in “Invite” relationships, whose tail and

head nodes correspond to users sending and accepting

the invitation, respectively. Note that we only model

invitations that resulted in the acquisition of a new users.

These relationships are characterized by a timestamp. Hence

we have a dynamic graph, with edges appearing and disap-

pearing over different time windows.

We sketch a small portion on this heterogeneous graph

in Figure 1, where users and merchants are connected with

three types of edges. For privacy reasons, we anonymize the

dataset by removing personally identiﬁable information. As a

result, users and merchants are identiﬁed by unique numeric

identiﬁers. Each user is associated with some personal details

(age, gender, place of residence, occupation), while merchants

are characterized by a category indicating the type of activity

and the province of their retail store.

We store our dataset in the graph database Neo4j3, which

provides a native representation of graph data, so we could

efﬁciently traverse the graph, query it for patterns and visualize

the resulting information. The dataset is quite large and

includes a number of nodes in the range (106,107)and a

number of relationships in the range (107,108).4

For our methodology, the “Invitation Network” plays a rele-

vant role. It simply represents the network of users connected

by the “Invite” relationships. Formally, we deﬁne it as the

subgraph G= (V,E)of our dataset comprising all users Vand

the invitation relationships among them.5The edges Ethen

represent the “Invite” relationships among couples of nodes

(u, v)∈ E ⊂ V × V. The Invitation Network Gplays a key

role in the development of our methodology, as it captures the

temporal evolution of the YAP network in terms of new users

acquired through accepted invitations. We therefore brieﬂy

characterize its main topological features. First, we note that

the invitation graph has a special structure: Gis a forest, i.e.,

each weakly connected component (WCC) of Gis a directed

tree, since each user can send many invitations but he can

accept only one. An example of a WCC from the dataset is

shown in Figure 2. The top user sent several invitations, 8 of

which were accepted. Some users in turn invited other users,

forming a WCC with a total of 34 users. The size of the WCCs

varies from small single-user or two-user components (none

or a single accepted invitation) to subgraphs with hundreds

of users. In Figure 3, we show the distribution of WCC size

in terms of a complementary cumulative distribution function

3https://neo4j.com

4We cannot disclose the exact numbers and ranges as they represent trade

secrets.

5Merchants cannot invite neither users or other merchants to the platform.

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

UserValueinModernPaymentPlatforms:AGraphApproachLauraArdittiLarusBusinessAutomationVenezia,Italylaura.arditti@larus-ba.itMartinoTrevisanUniversityofTriesteTrieste,Italymartino.trevisan@dia.units.itLucaVassioPolitecnicodiTorinoTorino,Italyluca.vassio@polito.itAlbertoDeLazzariLarusBusinessAutomationVe...

展开>> 收起<<

User Value in Modern Payment Platforms A Graph Approach Laura Arditti.pdf

共8页,预览2页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

User Value in Modern Payment Platforms A Graph Approach Laura Arditti

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: