Animals continuously evaluate past and current sensory and contextual information to make appropriate decisions during behaviour. How animals employ specific behavioural strategies while learning a task and how the orbitofrontal cortex (OFC) contributes to such strategy-based learning remains understudied. To investigate this, we trained mice on a tactile reversal learning task and implemented a trial-resolution Bayesian evidence accumulation model to measure the probability of the strategies used. We analysed multiple exploratory strategies across key task-learning phases and subsequently after rule reversal. During initial task learning, mice gradually switched from choice-driven strategies (e.g., ‘hit-stay-choice’) to cue-driven strategies (e.g., ‘hit-stay-cue’). Following the rule switch, mice reused a cue-driven strategy for reward-guided exploration. Mice did not rely on one exclusive strategy but adjusted the use of relevant strategies dependent on task phases. Silencing lateral OFC (lOFC) resulted in delayed choice-to-cue transition and impairments in behavioural flexibility. To study the contributions of orbitofrontal signalling in updating behavioural strategy representations to relevant sensory areas, we additionally measured functional responses from excitatory layer 2/3 neurons in the primary somatosensory cortex (S1) using two-photon Ca2+ imaging. Using tensor component analysis and a novel method of temporal decoding, we revealed that cue-driven strategies and relevant history-dependent variables were decoded during learning across mice. lOFC silencing resulted in a reduction of encoding, highlighting its critical contribution to guiding behaviour. Our study revealed how animals employ distinct exploratory strategies during flexible behaviour and highlighted the role of lOFC in leveraging prior knowledge supporting reward and error-guided learning.