The management of invasive plant species (IPS) requires knowledge of areas susceptible to invasion and the origin of the invasive biotypes. Ecological niche models (ENMs) are useful for these purposes, but modeling results depend on the data sources. We propose a synthetic approach to determine the selection of data source areas considering the invasion status of an IPS and management objectives to deal with the IPS. We assessed the importance of data source for ENMs and their projections to invasive areas using Chromolaena odorata, a Neotropical weed, in South Africa where this IPS is invading. We used MaxEnt to perform ENMs using different datasets from C. odorata's native range and from South Africa. We employed reciprocal ENM projections to find the probable native region of the plants invading South Africa. ENMs varied depending on the native area selected as the hypothetical invasion source. The modeling approach using worldwide data was most appropriate for prevention purposes, whereas the modelling approach using data from the Americas was most suitable for estimating invasion-susceptible areas in South Africa. The South African ENM was useful for reciprocal modelling but not for prediction of areas susceptible to invasion. ENM projections from the Americas to South Africa and vice-versa identified two native areas as possible invasion sources (northern Mexico and southern tropical South America). Their concordance with the South African ENM can be useful to search for natural enemies of C. odorata's and to reinforce the identification of invasion-susceptible areas in South Africa. We suggest that the various ENM obtained with the synthetic approach in modeling with different data sources for C. odorata cover the scenarios that depend on management purpose and invasion status for this weed.