On the basis of the BAM scheme, successful modeling depends on the answers to three questions:

1) What am I trying to model? This is a simple question with complex answers. Most often, people wish to model the intersection of the A, B and M circles (the actual distribution), which would require knowledge of the niche requirements, the abiotic environment and the dispersal capacities. (which begs the question of from where, and how long ago). Conventional "niche modeling" allows estimation of the potential area of distribution (the geographic expression of the circle A)

2) What kind of data I have? When one has true-absences data (i.e., reliable, well obtained sampling suggesting a species is absent from a place), it is possible to do regressions of the presence-absence data hoping that the combination captures information about the three circles above. True absence data is very scarce. What is truly abundant is presence-only data, from sources like GBIF, iDigBio and other similar organizations (CONABIO in Mexico, SANBI in South Africa, ALAS in Australia and several others). Modeling with presence-only data comprises the vast majority of niche-based distribution modelling. Some algorithms (like Maxent) require BACKGROUND data (which should not be confused with "pseudo-absences")

3) The ecological structure of the problem. Are we modelling a very mobile species (i.e., with a large M circle)? A species very dependent on the presence of competitors, predators, food plants? This would mean the B circle, interactions, cannot be dismissed.