Understanding consumer behavior during and after a Pandemic: Implications for customer lifetime value prediction models

•Structural breaks like COVID-19 significantly alter online grocery CLV predictions.•Behavioral shifts in pre- and during COVID-19 cohorts impact CLV predictions.•Cohort models excel in accuracy for individual CLV, outperforming overall models.•Pre-COVID models underestimate, post-COVID models overestimate cumulative CLV.•Among tested models, MBG/NBD with Gamma-Gamma ranks highest in CLV prediction. Our study uses a cohort analysis to investigate Customer Lifetime Value (CLV) for customer cohorts acquired before and during the COVID-19 pandemic. The research estimates CLV in a continuous-time setting of customer transactions within the online grocery sector. Stochastic models are combined with the Gamma-Gamma spending model to predict CLV at individual and aggregate levels. The findings reveal the satisfactory fit of the models at both individual and aggregate levels. Combined with the Gamma-Gamma model, the MBG/NBD model stands out as the top performer, accurately classifying over 60 % of the best-CLV customers (top 10 % and 20 %). Cohort-based analyses outperform overall sample models in terms of out-of-sample errors. Furthermore, CLV prediction models differ between the customer cohorts analyzed. The models for the pre-COVID-19 cohort underestimate the cumulative CLV, whereas models for the COVID-19 cohort overestimate it. These discrepancies can relate to the shifting behavior of the COVID-19 and pre-COVID-19 customer cohorts.