Product recall is a challenge which may have a significant financial impact. Incidents should be anticipated to improve responsiveness and reduce potential harm. In this paper, we propose a product recall approach following the detection of a critical fault. It is applicable to foodstuffs industry characterized by complex processes with high variability, high-speed manufacture and very large lots sizes. In such a case, usual strategy which consists of recalling entire lots is expensive and does not foster continuous improvement. The proposed approach in this paper allow to identify root causes and other products likely to present the same noncompliance in order to make a targeted recall. The root causes are searched based on an analysis of traceability data using a Bayesian model. A data model suitable for product and process traceability is also proposed. The originality of our approach lies on the reconstitution of the conditions of manufacturing of each item through the coupling of product and process unitary traceability data.