Customer comfort limit utilisation: Management tool informing credit limit-setting strategy decisions to improve profitability

Figures & data

Figure 1. Top: The effect of a credit limit increase on utilisation rate (UT).

Table 1. Illustration of the interpretation of the CCLU measure

Old balance	R10 000	New balance	R12 000
Old limit	R20 000	New limit	R25 000
Old limit utilisation	50%	New limit utilisation	48%
Balance growth = (R12 000-R10 000)/R10 000 = 20%
Limit growth = (R25 000- R20 000)/R20 000 = 25%
CCLU = balance growth/limit growth = 80
CCLU % = balance growth/limit growth = 80%

Figure 2. Distribution of CCLU values when the definition is applied to real data.

Table 2. Variables selected for the different modelling techniques

	Utilisation rate models			CCLU model
	Model 1	Model 2	Model 3	Model 4
Variables	Ordinary linear regression	Weighted logistic regression	Neural network	Neural network
Average debit balance (scaled & binned)		X		X
Three-month average of fees on bank’s card due to transactions (binned)	X	X	X	X
Three-month average of the maximum utilisation of an outside card (binned)	X	X	X	X
Three-month average of the value of payment on the bank’s own cards before CLI (scaled & binned)	X	X	X	X
Three-month average of the value of the turnover on bank’s own cards before CLI (scaled & binned)	X	X	X	X
Three-month average of the volume of cash transactions on bank’s own cards before CLI (binned)	X	X	X	X
CLI amount (scaled & binned)	X	X	X	X
Current limit after CLI (scaled & binned)	X	X	X	X
Debit order type (e.g., full balance, minimum, fixed amount, etc.) (clustered)			X
Customer gender			X
Indicator of overall marketing consent	X	X	X	X
Month-on-book since customer joined the bank	X	X	X	X
Product group in which customer falls (clustered)	X	X	X	X
Three-month average of the total limit at outside cards before CLI (scaled)	X	X	X
Utilisation month-end before CLI	X	X	X	X

Table 3. Result comparison of utilisation rate models

MODEL	ORDINARY LINEAR REGRESSION (Model 1)		WEIGTED LOGISTIC REGRESSION (Model 2)		NEURAL NETWORK (Model3)
	MSE	RMSE	MSE	RMSE	MSE	RMSE
TRAIN	0.07	0.27	0.07	0.27	0.06	0.24
VALIDATION	0.08	0.29	0.17	0.41	0.05	0.23

Table 4. Result comparison of best utilisation rate model and CCLU value model

MODEL	CCLU values derived from NEURAL NETWORK UTILISATION RATE model (Model 3)		CCLU values predicted directly by NEURAL NETWORK (Model 4)
	MSE	RMSE	MSE	RMSE
TRAIN	157.56	12.55	340.86	18.46
VALIDATION	174.93	13.23	358.71	18.94

Figure 3. Comparison of actual and predicted utilisation rates in panels (a), (c), (d) and CCLU values in panels (b), (e) and (f).

Table 5. Illustration of the effect of using CCLU values to inform limit-setting strategies

	Original strategy	Suggested strategy 1	Suggested strategy 2	Suggested strategy 3	Suggested strategy 4
CCLU ≤ 0	100%	10%	80%	85%	80%
0 < CCLU ≤ 50	100%	50%	90%	85%	90%
50 < CCLU ≤ 90	100%	90%	95%	95%	100%
90 < CCLU ≤ 100	100%	130%	100%	105%	105%
100 < CCLU < 150	100%	170%	95%	110%	110%
CCLU ≥- 150	100%	210%	90%	120%	115%
Capital: % increase		0,77%	0,19%	0,05%	0,09%
NII: % increase		0,90%	0,33%	0,07%	0,11%
IRORC: % Increase		0,030%	0,033%	0,018%	0,006%
CLI: % Increase		10,00%	−8,30%	0,00%	0,00%

Table A1. Summary of the factors (and examples of their underlying characterising) influencing credit card utilisation

Factor	Example of characteristics
Consumption needs and application characteristics	Age Marital status Number of dependents Household size Time at physical address Years at employer Occupation Highest level of education
Current and future resources	Income Liquid assets Investment assets Real assets Total amount of debt Number of credit cards Product holding
Consumer preferences and behavioural characteristics	User type (revolver or transactor)
Macro-economic factors	Interest rates GDP Unemployment rate

Table C1. List of the 22 variables available for model purposes

Number of outside cards

Average debit balance (scaled & binned)

Three-month average of fees on bank’s card due to transactions (binned)

Three-month average of the maximum utilisation of an outside card (binned)

Three-month average of the minimum utilisation of an outside card (binned)

Three-month average of the value of payment on the bank’s own cards before CLI (scaled & binned)

Three-month average of the value of the turnover on bank’s own cards before CLI (scaled & binned)

Three-month average of the volume of cash transactions on bank’s own cards before CLI (binned)

Average volume of payments (binned)

CLI amount (scaled & binned)

Current limit after CLI (scaled & binned)

Customer segment (clustered)

Debit order type (e.g., full balance, minimum, fixed amount, etc.) (clustered)

Customer gender

Indicator of overall marketing consent

Indicator that the customer does not have an outside card

Month-on-book since customer joined the bank

Product group in which customer falls (clustered)

Indicator of customers exhibiting revolving behaviour

Indicator of a secondary card

Three-month average of the total limit at outside cards before CLI (scaled)

Utilisation month-end before CLI