Using this csv with hypothetical transactional data, imagine that you are trying to understand if there is any kind of suspicious behavior.
3.2.1 - Analyze the data provided and present your conclusions (consider that all transactions are made using a mobile device).
- Peak hours: The highest number of chargebacks occur between 19:00 and 22:00, with a peak of 44 chargebacks from 19:00 to 20:00. This suggests increased fraudulent activity during evening hours.
- Low activity hours: Very few or no chargebacks occur between 3:00 and 11:00, indicating lower risk during early morning hours.
- The chargeback transactions amounts are usually 2x higher than average transaction amounts.
- The highest disparity is during 3:00 - 4:00, where chargeback amounts are 310% higher than average transactions, despite having only 4 chargebacks.
- The average transaction amount (non chargeback) drops after 00:00.
+---------------+-------------+--------------+--------------------------------+--------------------+----------------------------------------+
| Hour | Chargebacks | Transactions | Avg Transaction Amt (excl. CB) | Avg Chargeback Amt | Comparison |
+---------------+-------------+--------------+--------------------------------+--------------------+----------------------------------------+
| 0:00 - 1:00 | 24 | 109 | $499.1 | $1090.1 | 118.41% higher than avg transactions |
| 1:00 - 2:00 | 14 | 97 | $637.45 | $1201.6 | 88.5% higher than avg transactions |
| 2:00 - 3:00 | 18 | 43 | $636.38 | $1501.61 | 135.96% higher than avg transactions |
| 3:00 - 4:00 | 4 | 26 | $354.72 | $1455.43 | 310.3% higher than avg transactions |
| 4:00 - 5:00 | 0 | 7 | $370.75 | $N/A | N/A |
| 5:00 - 6:00 | 0 | 4 | $425.06 | $N/A | N/A |
| 6:00 - 7:00 | 1 | 1 | $1.3 | $263.93 | 20202.31% higher than avg transactions |
| 7:00 - 8:00 | 0 | 0 | $N/A | $N/A | N/A |
| 8:00 - 9:00 | 0 | 3 | $558.93 | $N/A | N/A |
| 9:00 - 10:00 | 0 | 7 | $247.08 | $N/A | N/A |
| 10:00 - 11:00 | 0 | 29 | $544.99 | $N/A | N/A |
| 11:00 - 12:00 | 5 | 88 | $789.34 | $1618.57 | 105.05% higher than avg transactions |
| 12:00 - 13:00 | 9 | 102 | $672.08 | $1619.23 | 140.93% higher than avg transactions |
| 13:00 - 14:00 | 11 | 184 | $712.42 | $1458.47 | 104.72% higher than avg transactions |
| 14:00 - 15:00 | 16 | 231 | $637.13 | $1138.51 | 78.69% higher than avg transactions |
| 15:00 - 16:00 | 18 | 221 | $767.95 | $1122.72 | 46.2% higher than avg transactions |
| 16:00 - 17:00 | 30 | 248 | $674.11 | $1390.48 | 106.27% higher than avg transactions |
| 17:00 - 18:00 | 31 | 240 | $793.65 | $1702.95 | 114.57% higher than avg transactions |
| 18:00 - 19:00 | 23 | 255 | $725.62 | $1743.69 | 140.3% higher than avg transactions |
| 19:00 - 20:00 | 44 | 218 | $572.88 | $1691.87 | 195.33% higher than avg transactions |
| 20:00 - 21:00 | 37 | 235 | $607.99 | $1310.45 | 115.54% higher than avg transactions |
| 21:00 - 22:00 | 42 | 185 | $739.39 | $1232.22 | 66.65% higher than avg transactions |
| 22:00 - 23:00 | 29 | 134 | $608.79 | $1765.7 | 190.03% higher than avg transactions |
| 23:00 - 24:00 | 35 | 141 | $686.13 | $1560.43 | 127.42% higher than avg transactions |
+---------------+-------------+--------------+--------------------------------+--------------------+----------------------------------------+- There are 10 devices that represents more than 30% of the total chargeback transactions. These devices may be compromised or used by fraudsters repeatedly.
+-----------+-------------------+
| Device ID | Total Chargebacks |
+-----------+-------------------+
| 563499 | 19 |
| 101848 | 15 |
| 342890 | 15 |
| 438940 | 13 |
| 547440 | 12 |
| 357277 | 6 |
| 960729 | 6 |
| 542535 | 6 |
| 308950 | 5 |
| 174844 | 5 |
+-------------------------------+- Isolating the top 10 merchants with chargebacks, we can see that they represent more than 30% of the total chargeback transactions. These merchants might be targets for fraudulent activities or have weak security measures.
+-------------+-------------------+
| Merchant ID | Total Chargebacks |
+-------------+-------------------+
| 17275 | 22 |
| 4705 | 19 |
| 1308 | 15 |
| 53041 | 14 |
| 77130 | 13 |
| 91972 | 11 |
| 44927 | 11 |
| 73271 | 10 |
| 55854 | 9 |
| 29214 | 9 |
+---------------------------------+3.2.2 - In addition to the spreadsheet data, what other data would you look at to try to find patterns of possible frauds?
-
Device fingerprints, IP addresses, browser information, mobile network operator data. Enables us to identify suspicious devices, detect impossible travel scenarios, and recognize patterns of device sharing among multiple accounts.
-
Transaction History: Long-term detailed transaction logs including dates, amounts, merchants, transaction types, and results (approved, declined, chargebacked). Allows us to build user behavior profiles, identify sudden changes in spending patterns, and recognize unusual transaction sequences.