Geocoding Scenarios

You can use Enterprise Manager to create dataflows that are appropriate for your business requirements and for the nature and quality of your data.

Multiple Country Stage with Geocode Address World as Last Geocoding Pass

You may be able to optimize your results by geocoding your input in several passes. In general, you can use more strict matching criteria in the first pass. In subsequent geocoding passes, you can apply less restrictive matching criteria to any addresses that previously failed to return a close match candidate. This strategy can produce accurate matches for your high-quality addresses and still give you the best possible matches for less accurate addresses, or for addresses in countries that do not have a comprehensive level of coverage.

Let's assume the following scenario:
  • Your input file includes addresses for six countries: Argentina (ARG), Brazil (BRA), Mexico (MEX), Chile (CHL), Venezuela (VEN), and Panama (PAN).
  • You have geocoders for three of these countries (ARG, BRA, and MEX) are deployed in a multiple country stage.
  • Geocode Address World is deployed in a separate stage to geocode addresses that could not be identified by the country-specific geocoders.
  • Your stage uses conditional routers (and optionally stream combiner) to manage the geocoding flow.
  1. Read input into the multiple-country stage. Geocoded addresses can be written out to a file or optionally sent to the stream combiner.
  2. Some Addresses that could not be geocoded in step 1. This may be because they were addresses from CHL, VEN, or PAN, and you do not have geocoders for these countries in the first stage. Or they may have failed to return a close match candidate in the first stage because of input errors or ambiguities in the addresses. These ungeocoded addresses are sent to the Geocode Address World stage.
  3. Addresses can be geocoded to postal or geographic accuracy by Geocode Address World. Successfully geocoded addresses can be written out to a file or optionally sent to the stream combiner.

    Postal geocoded candidates will have a Z1 result code. Postal geocoded results may be very accurate in countries with robust postcode systems. See Postal Geocoding. Geographic candidates will have a G result code (for example G3 for a town/city match). See Geographic Geocoding.

  4. The stream combiner (if used in your dataflow) can combine all geocoded addresses and write them to a file or direct them for further processing.
This is one scenario. You could use Enterprise Manager to design more complex dataflows that are suitable for your needs.

Using Geocode Address World as First Geocoding Pass

You could also use a strategy with Geocode Address World as the first geocoding pass.

Assume the following:

  • Your addresses typically do not specify a country (although some may).
  • Some addresses contain only street and city address information.
  • You have country-specific geocoders for some countries, but not all.
  • You use a main dataflow with subflows to manage the geocoding process.

Use a dataflow (possibly with subflows) that perform the following actions. Note that these steps illustrate a simplified view of a sample dataflow.

  1. Read input into the mulitple-country stage that also includes Geocode Address World. Based on city name (and possibly state name for USA addresses), each address can produce one or more potential close match candidates for several different countries. Each candidate will now be associated with a country, even though the input address may not have included a country.
  2. If a country-specific geocoder is available, the candidate is sent to that geocoder. This processing involves conditional routing, stream combiners, and other Spectrumâ„¢ Technology Platform control stages. Depending on the completeness of the input address and capabilities of the country-specific geocoder, candidates may be geocoded to a street (S result code), geographic (G result code), or postal (Z result code) level.
  3. If no country-specific geocoder is available, the candidate is routed to Geocode Address World, where candidates can be geocoded to a geographic or postal level.
  4. Candidates from all subflows are combined and ranked using a number of criteria. Ranking could be based on population of the city (city rank), accuracy of the match (street, geographic, postal), proximity to a user's locality, or other criteria.