CoChat
Sign Up
Cited by 108Customer churn and segmentationCustomer Service Quality and LoyaltyAI and HR Technologies

The DOI matches its Crossref record; 142 of the 157 references it cites were found in a registry or an index; 3 of the 54 sources its claims cite were checked for support.

A Survey on Churn Analysis in Various Business Domains

Open access
Ahn, Jaehyun, Hwang, Junsik, Kim, Doyoung, Choi, Hyukgeun, Kang, Shinjin·2020·IEEE Access
DOI 10.1109/access.2020.3042657
Open this paper in CoChat — free accountOpen PDF

Abstract

In this paper, we present churn prediction techniques that have been released so far. Churn prediction is used in the fields of Internet services, games, insurance, and management. However, since it has been used intensively to increase the predictability of various industry/academic fields, there is a big difference in its definition and utilization. In this paper, we collected the definitions of churn used in the fields of business administration, marketing, IT, telecommunications, newspapers, insurance and psychology, and described their differences. Based on this, we classified and explained churn loss, feature engineering, and prediction models. Our study can be used to select the definition of churn and its associated models suitable for the service field that researchers are most interested in by integrating fragmented churn studies in industry/academic fields.

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

This survey reviews churn analysis across business domains, collecting definitions and comparing churn loss, feature engineering, and prediction models. It classifies churn as contractual or non-contractual and reports that preferred modeling techniques vary by industry, with deep learning more common where dense log data exist.

Key findings

  • The survey identifies 12 business fields with churn prediction and reports model-type counts: telecommunications has 26 traditional machine learning, 3 statistics, 3 graph theory, and 1 deep learning study; games has 16 traditional machine learning, 7 statistics, and 2 deep learning; finance has 10 traditional machine learning, 2 statistics, and 1 deep learning (Table 2).
  • Churn definitions split into contractual churn (non-renewal at contract date) and non-contractual churn, where a time window defines inactivity; Yang et al. (2019) found more than 95% of mobile-game customers did not return after 3 consecutive absent days and set 3 days as the window, while Lee et al. (2018) set 14 weeks as the period when 75% or more customers were continuously unconnected.
  • Loss is framed via customer acquisition cost (CAC) and customer lifetime value (CLV); the paper cites Reichheld et al. (1996) that a 5% increase in retention rate achieved 35% and 95% increases in net present value for a software company and an advertising agency, respectively.
  • Feature engineering includes one-hot encoding, bucketing, data imputation, normalization, feature embedding, and imbalanced-data handling; for imbalance, Amin et al. (2016) found MTDF highest accuracy versus SMOTE, ADASYN, MWMOTE, ICOTE, and TRkNN, while Gui (2017) reported SMOTE best in telecom. Performance evaluation commonly uses AUC/ROC or lift/top-decile lift.

Methods

  • Design: literature survey integrating fragmented churn studies across industry and academic fields; no new empirical experiment is reported.
  • Data/sample: collected definitions of churn from business administration, marketing, IT, telecommunications, newspapers, insurance, and psychology, and cases from 12 business fields.
  • Analysis: thematic classification and comparison of churn definitions, churn loss (CAC/CLV), feature engineering, prediction models (traditional machine learning, statistics, graph theory, deep learning), and performance evaluation.

Limitations

  • The paper states that because each study uses different sizes and types of features, the set of studies has a limitation in comparing a common performance; readers cannot determine which model is suitable and has the best performance for their study.
  • The paper states future work will combine feature engineering with open churn datasets and construct various churn prediction models, including a deep learning model, to compare and evaluate suitability.
  • Evident from the text: the survey is narrative and does not run a benchmark or meta-analysis; claims such as deep learning outperforming other algorithms are based on cited studies rather than independent validation.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1109/access.2020.3042657

Cites

142 of 142 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

3 of 54 cited sources checked · 0 supported · 0 flagged · accuracy not yet measured

Show all 29 claims

  • Probably not a substantial claim. 130
  • It may state two separate claims as one. 33
140 of 140

  • The paper relies on this without support.

    Customer churn behavior is quantitative.

    “Customer churn behavior is quantitative.”

  • The paper relies on this without support.

    It is difficult to directly relate customer churn to a decrease in sales.

    “However, it is difficult to directly relate customer churn to a decrease in sales.”

  • The paper relies on this without support.

    Communication costs arising from different churn criteria between service personnel in the process of fusing heterogeneous services have been sharply increasing.

    “In the industry, communication costs arising from different churn criteria between service personnel in the process of fusing heterogeneous services (e.g., vehicle sharing service/insurance, online music service/department store) have been sharply increasing.”

  • The paper relies on this without support.

    There are many difficulties for researchers to launch new research on churn because of fragmented previous studies.

    “However, because of fragmented previous studies, there are many difficulties for researchers to launch new research on churn.”

  • The paper relies on this without support.

    The trend of building churn prediction models is changing, and performance is rapidly improving.

    “The trend of building churn prediction models is changing, and performance is rapidly improving.”

  • The paper relies on this without support.

    The CAC cost is the minimum value to operate service makes from a customer with Return on Investing (ROI).

    “The company measures the cost of acquiring a customer with the CAC, and that CAC cost is the minimum value to operate service makes from a customer with Return on Investing (ROI).”

  • The paper relies on this without support.

    Pure log data includes all truthful customer information.

    “Pure log data includes all truthful customer information.”

  • The paper relies on this without support.

    Service indicators are generally sparse and have many outliers, and the data distribution is skewed to one side.

    “This is because service indicators are generally sparse and have many outliers, and the data distribution is skewed to one side.”

  • The paper relies on this without support.

    The distribution of service data is generally skewed to one side.

    “The distribution of service data is generally skewed to one side.”

  • The paper relies on this without support.

    In a stable service, the number of churned customers generally account for a small proportion compared to the number of retained customers.

    “In a stable service, the number of churned customers generally account for a small proportion compared to the number of retained customers.”

  • The paper relies on this without support.

    The types and cycles of log data used for each business are different.

    “The reason why the preferred model for each business field is different is that the types and cycles of log data used for each business are different.”

  • The paper relies on this without support.

    Most churn prediction models are related to customer relation management.

    “Most churn prediction models are related to customer relation management.”
12 of 12

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • The paper says there are 12 business fields but lists 14

    Minor

    The text states that churn prediction was performed in 12 business fields, while Table 2 lists 14 distinct business fields.

1 of 1

Citations

Cites

Flagged first, then by citations

15 of the 157 references couldn’t be matched to a known paper.

Research works printed but not linked: 10

  1. komenar, 1996, Electronic Market
  2. nielsen, 2001, Major study to track store switching, Retail World
  3. almana, 2014, A survey on data mining techniques in customer churn analysis for telecom industry, Int J Eng Res Appl, 4, 165
  4. 2020, Investopedia
  5. lewis-evans, 2012, Finding out what they think: A rough primer to user research, Part 2, Gamasutra
  6. umayaparvathi, 2017, Automated feature selection and churn prediction using deep learning models, Int Res J Eng Technol, 4, 1846
  7. nozhnin, 2012, Predicting churn: When do veterans quit, Gamasutra
  8. stewart, 2019, The Actual Difference Between Statistics and Machine Learning
  9. chandar, 2006, Modeling churn behavior of bank customers using predictive data mining techniques, Proc Nat Conf Soft Comput Techn Eng Appl (SCT), 24
  10. nath, 2003, Customer churn analysis in the wireless industry: A data mining approach, Proc Annu Meet Decis Sci Inst, 561, 505

Cited by

Sorted by citations

See all 108

Related

See all

Metadata

Authors
Ahn, J., Hwang, J., Kim, D., Choi, H., & Kang, S.
ORCID
Venue
IEEE Access
Publisher
Institute of Electrical and Electronics Engineers (IEEE)
Published
2020
Volume
8
Pages
220816-220839
ISSN
2169-3536 (ISSN-L)
DOI
10.1109/access.2020.3042657
OpenAlex
W3112348427
Semantic Scholar
c4b810efe7f18ea3735744fb7c8fc768e7bb6e9e
Cited by
108
References
157
Citations by year
2026: 12 · 2025: 18 · 2024: 27 · 2023: 23 · 2022: 19 · 2021: 8 · 2020: 1
FWCI
13.15
Type
journal article
Language
en
Open access
Licence
Locations
  • IEEE Access (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • National Research Foundation of Korea, Korean government (NRF-2019R1A2C1002525)
  • Nexon Korea
Topics
  • Social Sciences › Business, Management and Accounting › Marketing › Customer churn and segmentation
  • Social Sciences › Business, Management and Accounting › Marketing › Customer Service Quality and Loyalty
  • Social Sciences › Business, Management and Accounting › Organizational Behavior and Human Resource Management › AI and HR Technologies
Keywords
customer churn prediction, feature engineering, prediction model, telecommunications, insurance, online services, business domain

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-10-12, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-12 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review