Flash sale merupakan strategi pemasaran populer di e-commerce yang sering memanfaatkan kesan kelangkaan untuk mendorong keputusan pembelian cepat. Namun, praktik ini dapat mengarah pada fake scarcity atau kelangkaan palsu, yaitu ketika kelangkaan diciptakan secara manipulatif. Penelitian ini bertujuan untuk mendeteksi indikasi fake scarcity pada flash sale dengan menggunakan algoritma Isolation Forest. Data penelitian menggunakan enam fitur numerik yang merepresentasikan aktivitas penjualan dan promosi. Analisis dilakukan melalui tiga tahap: (1) pengukuran Mutual Information Score untuk menilai kontribusi fitur, (2) evaluasi rasio anomali menggunakan pie chart berdasarkan kombinasi fitur, serta (3) analisis distribusi anomaly score melalui histogram.Hasil penelitian menunjukkan bahwa ketika seluruh fitur digunakan, rasio anomali mencapai 74,1%. Fitur transaksi seperti Clicks, Discount_Level, dan Units_Sold lebih dominan, sedangkan Conversions, Bundle_Price, dan Customer_Satisfaction_Post_Refund memperkuat hasil dengan memberikan konteks tambahan.Pola yang terdeteksi konsisten menunjukkan adanya ketidakwajaran dalam perilaku transaksi. Kesimpulannya, algoritma Isolation Forest dapat digunakan secara efektif untuk mendeteksi indikasi fake scarcity dalam flash sale tanpa memerlukan data berlabel. Temuan ini diharapkan menjadi kontribusi awal dalam pengembangan sistem deteksi otomatis yang mendukung praktik perdagangan online yang lebih transparan.

Flash sale is a popular marketing strategy in e-commerce that often leverages the impression of scarcity to encourage quick purchasing decisions. However, this practice can lead to fake scarcity, which occurs when scarcity is created manipulatively. This study aims to detect indications of fake scarcity in flash sales using the Isolation Forest algorithm. The research data employs six numerical features representing sales and promotional activities. The analysis was conducted in three stages: (1) measuring the Mutual Information Score to assess feature contribution, (2) evaluating the anomaly ratio using a pie chart based on feature combinations, and (3) analyzing the distribution of anomaly scores through a histogram. The results show that when all features are used, the anomaly ratio reaches 74.1%. Transaction-related features such as Clicks, Discount_Level, and Units_Sold are more dominant, while Conversions, Bundle_Price, and Customer_Satisfaction_Post_Refund strengthen the findings by providing additional context. The detected patterns consistently indicate irregularities in transaction behavior. In conclusion, the Isolation Forest algorithm can be effectively used to detect indications of fake scarcity in flash sales without requiring labeled data. These findings are expected to serve as an initial contribution to the development of automated detection systems that support more transparent online trading practices.