India was one of the most vulnerable countries to the COVID-19 pandemic considering the high transmissibility of the virus, exploding population, and fragile healthcare infrastructure. As an early counter, India implemented a country-wide lockdown and we aimed to study the impact of 4 lockdowns and 2 unlock phases on 6 outcomes: case growth, death count, effective reproduction number, mobility, hospitalization, and infection growth by two methods: interrupted time series (ITR) analysis and Bayesian causal impact analysis (BCIA) for nationals and sub-national levels. We observed that the effects are heterogeneous across outcomes and phases. For example, ITR revealed the effect to be significant for all the outcomes across all phases except for case growth in phase 1. BCIA revealed that the causal effect of all four lockdown phases was positive for deaths. At the state level, Maharashtra benefited from the lockdown in comparison to Tripura. Effects of lockdown phases 3 and 4 on death count were correlated (R=0.70, p<0.05) depicting the 'extended impact' of phase-wise interventions. We observed the highest impact on mobility followed by hospitalization, infection growth, effective reproduction number, case growth, and death count. For optimal impact, lockdown needs to be implemented at the sub-national level considering various demographic variations between states.