MI Address Geocoding Difficulties

[Valdis J Kalnins]

Dear List,   We are conducting some address geocoding tests and are getting puzzling results.  In essence, houses are not being proportionately located along the street segments.  Furthermore, house placement appears to change as "shaping" nodes are added or removed from continuous road segments.   In the simplest test case, we have a line with a start node and end node.  Of four houses that were "exactly matched" according to the result codes, two of them were placed 500 to 1100ft south of the starting node of a road segment bearing north.   After adding 3 intermediate nodes (but not splitting the segment) equally spaced along the segment, all four houses geocoded to either side of the segment as expected.  Visually the result looked correct.  However, the houses were not located proportional distances along the segment in relation to the house number.   It seems logical that the geocoding algorithm would divide the house number by the range of the TO and FROM address and multiply this % by the total length of the segment to get a proportional location.  Working from the total length of the segment, intermediate nodes shouldn't play a role in determining location so long as they don't split the segment.   In the worst case, we had 9 houses placed 15 miles from the nearest road segment with a result code of "exactly matched".  It turned out that the road segments they were supposed to be attached to were "combined" segments consisting of 2 or more lines.  While this error is explainable, it is unclear why the result code would not indicate the problem.   Our road layer is projected in NAD83 Meters, State Plane, Michigan South Our addresses run from south to north and east to west with a range of 200 possible numbers per mile We are using MapInfo v5.5.0   Any insight or explanation would be much appreciated.   Regards, Valdis Kalnins