FraudFinder Assesment
Illya Mowerman
5/23/2019
Outline
- Overall Summary: Revenue and Customer Base
- New vs. Lost Customers
- Product Summary & Switching to FraudFinder 2.0
- Volume Based Pricing
- Health Conclusion
- Appendix
Overall Summary: Revenue and Customer Base
Revenue is growing, and the customer base is shrinking. This due to price increases and switching to FraudFinder 2.0
Overall Summary
|
Year
|
Revenue
|
Revenue per
Customer
|
Customer
Count
|
|
2016
|
$879,686
|
$239
|
3675
|
|
2017
|
$1,034,273
|
$310
|
3333
|
|
2018
|
$1,179,254
|
$387
|
3047
|
|
2019
|
$1,272,550
|
$442
|
2880
|
Overall Growth
|
Year
|
Revenue Growth
|
Revenue per
Customer Growth
|
Customer
Growth
|
|
2017
|
17.57%
|
29.64%
|
-9.31%
|
|
2018
|
14.02%
|
24.72%
|
-8.58%
|
|
2019
|
7.91%
|
14.17%
|
-5.48%
|
New vs. Lost Customers
The revenue from new customers is not compensating for the lost revenue of those that left.
Product Summary and Switching to FraudFinder 2.0
Likely within two years most customers will have switched to FraudFinder 2.0
Switching by Year: FraudFinder to FraudFinder 2.0
|
Year
|
Switch Count
|
Switch
Proportion
|
|
2016
|
190
|
7.61%
|
|
2017
|
345
|
17.13%
|
|
2018
|
367
|
25.70%
|
|
2019
|
140
|
14.86%
|
Recommendation: Perhaps volume based pricing
- Many high value customers have low to no usage
- Several low price customers have high volume
- Correlation between Price and volume is 0.35
- A simulation of Volume based pricing is merited
##
## Pearson's product-moment correlation
##
## data: for_corr$total_revenue and for_corr$Transactions
## t = 25.837, df = 4672, p-value < 2.2e-16
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.3282333 0.3784097
## sample estimates:
## cor
## 0.3535758
Health Wrap-up
- Growth in revenue will likely fall in approximately two years, as most customers will have switched to FraudFinder 2.0. In addition, current Annual increases may not be sustainable
- Revenue is estimated to be 20% higher in two years, slightly lower than current growth
- After two years, it is likely that revenue growth will follow the same growth rate as other company segments
- Recommendation 1: Simulate volume based pricing
- Recommendation 2: Recover lost customers in either volume based pricing or with FraudFinder1.0
Appendix: Data Cleaning Process
- Checked for duplicates: 68 dups found un usage table with Transactions = NA
- Transposed Revenue table
- Left Joined revenue with usage
- Check if there is any usage without revenue: Yes there are, but all are zero
- Removed $0 revenue with transactions NA
- Removed no revenue and transactions > 0: 260 records
- Constrained the data to 99th percentile in FF, FF2.0, and other revenue. See next slide EDA
- Added new variables: switching, attrition, new customer
EDA for Outliers
## Cust_ID Year FraudFinder FraudFinder20
## Length:13009 Min. :2016 Min. : 0.0 Min. : 0.00
## Class :character 1st Qu.:2016 1st Qu.: 0.0 1st Qu.: 0.00
## Mode :character Median :2017 Median : 0.0 Median : 74.35
## Mean :2017 Mean : 116.8 Mean : 248.44
## 3rd Qu.:2018 3rd Qu.: 118.5 3rd Qu.: 177.17
## Max. :2019 Max. :8676.6 Max. :64358.59
##
## Other Transactions total_revenue
## Min. : 0.00 Min. : 0 Min. : 0.02
## 1st Qu.: 0.00 1st Qu.: 124 1st Qu.: 95.14
## Median : 0.00 Median : 21239 Median : 153.38
## Mean : 26.74 Mean : 971985 Mean : 391.97
## 3rd Qu.: 0.00 3rd Qu.: 212706 3rd Qu.: 278.91
## Max. :32808.40 Max. :802117023 Max. :71084.31
## NA's :8246
## Warning: attributes are not identical across measure variables; they will
## be dropped
EDA for Outliers Continued
## variable mean std_dev variation_coef p_01 p_05 p_25 p_50 p_75
## 1 Year 2017 1.1e+00 5.5e-04 2016 2016 2016 2017 2018
## 2 Transactions 971985 1.3e+07 1.3e+01 0 0 124 21239 212706
## p_95 p_99 skewness kurtosis iqr range_98 range_80
## 1 2019 2.0e+03 0.13 1.6 2 [2016, 2019] [2016, 2019]
## 2 2861836 1.5e+07 51.34 3081.7 212582 [0, 14858604.16] [0, 1076403.2]