Different lat and long for same address

[Stella-MarisA]

I Geocoded a list of address (Mostly intersections) for bus routing and the same address was geocoded by a different company. The lat and long values that I get is different from the one from the company. I calculated the difference in excel as shown below. Some seem too big. My question is how much is too big? (Calculation in metres). Looking at the table and map it seems not too far off but when I measure the distances in a straight line it seems much as shown in the table. I used projected coordinate system. I am trying to decide which of the coordinates(Lat/long) is the more precise and accurate so the bus will not miss a student. Is it okay to get varying lat and long for same address? Please can anyone help? Any help with geocoding intersections will be appreciated. I checked the match with no house number to Yes.

Thank you

[JoeBorgione]

You mention that you geocoded a list of addresses.  What data did you match against?  Is it your own?  Is it from a reputable source? 

If you add the company address points to the street data you'll use to create your network dataset, what sort of precision do they have compared to your points? That is, which points fall in the intersections as they should?

[Stella-MarisA]

I used the City's street network data to create my reference data before geocoding. I am not sure which data the company used. Could this be a factor in the difference. Is the difference too much to neglect?

[Stella-MarisA]

Also the points are close most of the time just not on top of each other until you zoom out. I am just curious why they don't match very closely.Thanks for replying me very fast. I appreciate

[DanPatterson_Retired]

What is the difference between expectation and observation?  A meter?, a couple? 10, 20, 50?.

big differences could be due to errors in coordinate systems and/or projections.  smaller differences could be attributed to precision and accuracy

[Stella-MarisA]

MeterDifference	ArcGisCoords_For_Google	GeoFrom_trapeze
552.7224544	53.4608055278, -113.37762709	53.460658, -113.385945
550.109075	53.4605877774, -113.385961346	53.460739, -113.377683
300.4362706	53.4285407575, -113.53653838	53.430043, -113.532783
299.5025307	53.626076525, -113.533326156	53.624516, -113.537014

Thanks Dan for your reply

I sorted the coordinates and above is the ones with the most difference. The difference is in metres.  I am trying to figure out if I should exclude/ re-geocode those with the big difference like 552 metres. Thank you

[DanPatterson_Retired]

if the rest are much smaller, then one or the other is in error... buildings don't move unless they are on wheels.  I would do the google thing and check the points to see the the long/lats look plausible.  Perhaps one of the two points is in a field or a lake or something obvious, indicating an error in geocoding or the collected data

[Stella-MarisA]

Thanks Dan. I think the difference maybe due to different reference data used. On checking through I realized the high numbers are not 100 % match hence the big difference. My question is does different reference data produce different lat and long for intersections and buildings, in the case where it goes with one side of the intersection versus the other side(L/R) OR a pin on the building roof versus parking lot or back door.

[JoeBorgione]

Varying results of geocoding against two different data sources in not a surprise to me.  In my original reply I mentioned 'reputable data source'.  That's a key element to geocoding.

[Stella-MarisA]

Much thanks for clarifying Joe!