Data quality improvement is becoming an increasingly important issue. In contexts where data are replicated among different sources, data quality improvement is possible through extensive data comparisons: whereas copies of same data are different because of data errors, comparisons help to reconcile such copies. Best quality copies can be selected or constructed in order to correct other copies.Record matching algorithms can support the task of linking different copies of the same data in order to engage reconciliation activities; for instance, a periodical running of record matching algorithms canbe performed in order to reconcile copies with different quality. Nevertheless, the extensive running of such algorithms is typically performed in fixed instants. This allows for periods in which the qualityof data can deteriorate, while no quality improvement action is performed on data. In this paper, we describe the DaQuinCIS platform for data quality improvement in contexts where data are replicatedamong heterogeneous and distributed sources. The quality improvement strategy underlying the proposed platform complements a periodical record matching activity with an “on-line” quality improvement,performed at query processing time. We experimentally show the feasibility and effectiveness of our approach by applying it to real databases; we also quantitatively evaluate the efficiency of oursystem.