Big Data Analytics and Sales Performance

The study used SPSS v25 to ensure the validity and reliability of the exploratory factor analysis (EFA). A structural equation modelling (SEM) approach was adopted to test the proposed research model, and AMOS v25 was used to conduct SEM and confirmatory factor analysis (CFA). AMOS is an appropriate tool for CFA and SEM and a powerful tool for estimating specific indirect effects. The demographical information presented through SPSS is shown in Table 4. The results indicated that the respondents were equally diverse in gender, with 53.6% males and 46.4% females. Furthermore, 96.4% of the respondents had bachelor's or master's degrees, and 94% of the respondents were aged below 46 years. Therefore, the respondents of the study were young, educated, and equally diverse in gender.

Table 4 Demographical information.

	Category	Frequency	Percentage
Gender	Male	223	53.6
	Female	193	46.4
	Total	416	100
Education	High school/diploma	7	1.7
	Bachelor	207	49.8
	Master	194	46.6
	Doctor	8	1.9
	Total	416	100
Age	18–25	57	13.7
	26–35	163	39.2
	36–45	171	41.1
	46 and above	25	6.0
	Total	416	100

Measurement Model

The study first confirmed the samples' adequacy using the Kaiser–Meyer–Olkin (KMO) test, and the KMO score was 0.914, which exceeded the cutoff value of 0.8. After ensuring the samples' adequacy, the central concern of common method bias (CMB) was addressed by adopting Harman's single factor test. The first factor explained the total variance after categorizing all items into eight subgroups below the threshold value of 50%.

The validity and reliability were ensured through Cronbach's alpha, average variance extracted (AVE), and composite reliability (CR). The values of Cronbach's alpha, CR, and AVE ranged from 0.967 to 0.882, 0.971 to 0.883, and 0.892 to 0.654, respectively. The values of Cronbach's alpha, CR, and AVE exceeded the threshold values of 0.7, 0.7, and 0.5, respectively, thus indicating the absence of reliability and validity issues. Subsequently, EFA was performed to ensure that the measures were according to the respective variables; the results of factor loading ranged from 0.717 to 0.866 and divided the total items into eight factors. The factor loading values exceeded the threshold value of 0.7, which guaranteed the absence of any factor loading issue. Table 5 presents the results of factor loadings, Cronbach's alpha, CR, and AVE.

Table 5 Results of factor loadings, Cronbach's alpha, composite reliability (CR), and AVE.

Sr. no	Constructs	Items	Loadings	Cronbach's alpha	CR	AVE
1	Better customer services	BCS1	0.736	0.921	0.919	0.701
		BCS2	0.764
		BCS3	0.845
		BCS4	0.852
		BCS5	0.849
2	Personalization	PR1	0.801	0.927	0.935	0.829
		PR2	0.779
		PR3	0.825
3	Advanced analytics	AA1	0.807	0.915	0.917	0.785
		AA2	0.754
		AA3	0.733
4	Improved relational knowledge	IRK1	0.830	0.967	0.971	0.892
		IRK2	0.812
		IRK3	0.842
		IRK4	0.866
5	Customer interaction management capability	CIMC1	0.765	0.917	0.913	0.679
		CIMC2	0.776
		CIMC3	0.809
		CIMC4	0.771
		CIMC5	0.761
6	Customer relationship upgrading capability	CRUC1	0.762	0.882	0.883	0.654
		CRUC2	0.735
		CRUC3	0.741
		CRUC4	0.815
7	Customer win-back capability	CWBC1	0.717	0.902	0.905	0.712
		CWBC2	0.884
		CWBC3	0.718
		CWBC4	0.875
8	Perceived sales performance	PSP1	0.768	0.884	0.885	0.720
		PSP2	0.792
		PSP3	0.804

The study elaborated on the square root of AVE to ensure the discriminant validity, as suggested by a prior study. As shown in Table 6, the values of the square root of each construct exceeded those of all interconstructs linked with the variable; hence, no discriminant validity issue existed.

The study subsequently performed path analysis. The value of CMIN/DF was 1.099. The values of CFI, NFI, RFI, IFI, and TLI were 0.999, 0.987, 0.965, 0.999, and 0.997, respectively. The values of RMSEA, PClose, and SRMR were 0.015, 0.960, and 0.036, respectively. The values were within the acceptable range and ensured the good fitness of the model. Figure 2 presents the path coefficient and significance level of the proposed research model. According to the results, BCS was positively associated with CIMC (β = 0.198, p<0.001)), CRUC (β = 0.095, p<0.050), CWBC (β = 0.282, p<0.001), and PSP (β = 0.104,  p < 0.050). Personalization was positively associated with CIMC (β = 0.125, p<0.050), CRUC (β = 0.214, p < 0.001), CWBC (β = 0.156, p < 0.010 ), and PSP (β = 0.129, p < 0.050). AA was positively associated with CIMC (β = 0.189, p < 0.001), CRUC (β = 0.258, p < 0.001), CWBC (β = 0.137, p < 0.050), and PSP (β = 0.117, P < 0.050). IRK showed significant relationships with CIMC (β = 0.292, p < 0.001 p < 0.001), CRUC (β = 0.257, p < 0.001), CWBC (β = 0.137, p < 0.010), and PSP (β = 0.106,p < 0.050). Furthermore, CIMC was positively associated with PSP (β = 0.105, p < 0.050), CRUC had a significant impact on PSP (β = 0.106, p < 0.050), and CWBC had a significant impact on PSP (β = 0.198, p < 0.001). Moreover, the adjusted R-square (R²) of CIMC was 0.41, indicating that the predictor variables caused 41% of the variance of CIMC. Furthermore, the R² values of CRUC, CWBC, and PSP were 0.45, 0.32, and 0.42, respectively.