Usually a software product line (SPL) is developed by exploiting available resources of a set of software variants that deem similar. In order to reengineer such variants that are developed by ad-hoc reuse into software product line that are developed by systematic reuse, it is necessary to identify traceability links between features and source code in a collection of product variants. Information retrieval (IR) methods are used widely to achieve this goal. These methods handle product variants as singular entities. However when product variants are considered together, we can get additional information that improves IR results. This paper proposes an approach to improve IR results when they are applied to identify traceability links in a collection of product variants. The novelty of our approach is that we exploit commonality and variability across product variants at feature and implementation levels to apply IR methods in efficient way. The obtained results proved that our approach significantly outperforms direct applying IR technique in conventional way in term of precision and recall metrics.