From insights to impact: leveraging data analytics for data-driven decision-making and productivity in banking sector

The paper addresses how data-driven decision-making (DDDM) and data analytics (DA) influence bank productivity amid accelerating digital transformation, particularly in Pakistan’s financial sector. With rising data volumes from digital services and fintech-led disruption, banks aim to replace intuition-based decisions with data-driven practices. Pakistan’s banking sector, despite significant digitalization and large data generation from transactions, exhibits gaps in adopting DDDM at the organizational decision level. The study seeks to quantify the effect of DDDM on productivity and assess whether DA investments strengthen this relationship. The authors posit two hypotheses: H1: DDDM significantly affects the productivity of Pakistan’s banking sector. H2: DA significantly impacts (strengthens) the relationship between DDDM and productivity. The study’s importance lies in offering empirical evidence on DDDM’s productivity effects, addressing endogeneity through IV-2SLS, and informing policy in a sector pivotal to economic development.

The review highlights DDDM as a key managerial process that transforms data into insights for faster, more accurate, and efficient decisions, contingent on reliable, accessible data and an organization-wide data-driven culture. Advances in analytics (e.g., decision trees, neural networks) enable automated, evidence-based decisions. Drawing on diffusion of innovation (DOI) theory, adoption of ICT and analytics fosters innovation, information availability, reduced transaction costs, and data-driven decisions that can enhance productivity. Prior evidence shows DDDM adoption rising alongside IT investment, with firms using DDDM being 5–6% more productive, and links between DDDM and improved profitability and plant productivity. The review motivates hypotheses that DDDM improves bank productivity (H1) and that DA investment positively moderates the DDDM–productivity relationship (H2).

Design and sample: The study covers all commercial and microfinance banks registered with the State Bank of Pakistan, excluding foreign and specialist institutions, yielding 26 commercial and 10 microfinance banks (36 total). Panel period: 2016–2020, with up to 180 firm-year observations; final IV regressions use 142 observations. Primary data: A structured survey (adapted from Brynjolfsson and McElheran, 2019) administered to CIOs, data analysts, IT heads, and/or senior managers (via Google Forms and in-person) across all 36 banks. Constructs measured on Likert scales include DDDM, Adjustment Cost to Change (AdjC), Exploration (EXPL), and Human Capital (HC). Reliability: Cronbach’s alpha—DDDM: 0.692; AdjC: 0.893; EXPL: 0.900; HC: 0.755. Index construction: Principal Component Analysis (PCA) used to create composite indices for DDDM, EXPL, AdjC, HC; diagnostics show KMO > 0.6 and Bartlett’s test significant, retaining components with eigenvalues >1. Secondary data: Productivity measured using an extended Cobb–Douglas production function under the intermediation approach: Output (Y) = log(loans + deposits/investments). Inputs: capital (K = net fixed assets) and labor (L = full-time employees). Controls: bank stability (Z-Score), non-performing loans (NPL ratio), bank type (commercial vs. microfinance), listing status, and DA investment variables (e.g., DA dummy; DA-age where applicable). Variables: DA is a dummy (1 if bank invested in DA; 0 otherwise). DDDM is a PCA-based index (range approx. 0–1). EXPL and AdjC are PCA-based indices. Models: (1) log(Y_it) = β0 + β1 DDDM_it + β2 log(K_it) + β3 log(L_it) + β4 log(ITF_it) + Σγ X_it + u_it. (2) log(Y_it) = β0 + β1 DDDM_it + β2 log(K_it) + β3 log(L_it) + β4 DA_it + β5 (DA_it×DDDM_it) + Σγ X_it + u_it. Estimation: To address endogeneity and reverse causality (common in IT–performance studies), the study employs 2SLS-IV. Instruments: For DDDM, lagged output and adjustment cost to change (AdjC) are used; in the moderation model including DA, exploration (EXPL) and lagged output are added, satisfying instrument relevance and exogeneity conditions. Diagnostic tests include Sargan overidentification tests, weak instrument tests (Wald χ²), and endogeneity tests.

- DDDM effect on productivity: DDDM increases banks’ output by approximately 10.5% (2SLS/IV coefficient ~0.105), statistically significant at 1%. - DA direct effect: DA investment is positively associated with output, with a coefficient of about 0.0676 (~6.7% increase), significant. - Moderation (DA×DDDM): The interaction term is positive (~0.0443), indicating that DA strengthens the DDDM–output relationship by about 4.43%. - Controls: Bank stability (Z-Score) and bank type (commercial vs. microfinance) are positive and significant; commercial banks benefit more from DDDM adoption. NPL ratio, listing status, and HC are generally insignificant (HC marginally negative in one model). - Diagnostics: Sargan overidentification tests yield p-values ~0.13–0.16 (fail to reject instrument validity). Weak instrument tests are passed (Wald χ² ≈ 12.35 and 14.583; p=0.000). Endogeneity tests are significant, supporting IV use. Adjusted R² ≈ 0.54–0.64; observations = 142.

Findings support the hypotheses and diffusion of innovation theory: adopting DDDM significantly enhances bank productivity, and investments in DA further amplify these gains, reflecting complementarities between analytics capabilities and data-driven practices. The positive role of Z-Score suggests that DDDM adoption aligns with greater stability and operational robustness. Commercial banks, typically larger and more digitally mature, appear better positioned to translate DDDM into productivity improvements than microfinance banks. Collectively, results indicate that to realize performance benefits, banks require both a data-driven culture and tangible analytics investments that facilitate timely, reliable, organization-wide access to data for decision-making.

The study provides empirical evidence from Pakistan’s banking sector that DDDM improves productivity by about 10% and that DA investment both directly raises output and strengthens DDDM’s impact. By employing IV-2SLS, the analysis addresses endogeneity, bolstering causal interpretation. Practically, banks should invest in analytics, ensure data quality and accessibility, and cultivate a data-driven culture to support real-time decision-making, risk management, and cost efficiency. Theoretically, applying IV-2SLS in this context contributes to literature on digital transformation and performance by clarifying the DDDM–productivity linkage and the moderating role of DA. Future research should extend to other industries and performance outcomes (e.g., profitability, credit risk), and explore integration with emerging technologies (e.g., AI) to further enhance decision processes.

Generalizability is limited as the study focuses on Pakistan’s banking sector and uses productivity as the primary performance metric. Other relevant outcomes such as profitability, risk exposure, and financial performance were not examined. Organizational challenges in adopting DDDM were not deeply analyzed. Future work should consider other sectors (e.g., healthcare, telecommunications), broader performance measures, barriers to DDDM adoption, and complementarities with emerging technologies like AI.