Google Maps : PoC Requirements
Summary
The address validation service must parse, validate, and enrich user-provided address information. It should process input that includes fields such as house number, street name, apartment/building, subarea/district, city, region, country, postal code, and P.O. box comprehensively.
The service must perform address suggestion, validation, and conformance. It needs to suggest corrections for ambiguous or erroneous inputs, validate the address against authoritative databases, and ensure the output conforms to local postal standards. By enriching incomplete data, verifying accuracy, and standardizing format, the service should be able to significantly improve (existing & new ) address data quality leading to operational efficiency.
Functional Requirements
Address validation services should encompass the following functionalities.
1. Address Suggestion
The service should provide suggestions based on the user entered address.
- Address suggestions should be closely aligned with the user-entered address.
- The output must indicate the level of match (e.g., house number match, street match, or geographical match) and provide a match confidence score (match score).
- Address suggestions should align with the country specific formats
- Nice to have Time zone information as part of results.
Address Autocomplete
The service should provide an autocomplete capability as the address is entered.
- Partially enter addresses, prompt with list of values aligned to the partially entered address
Address Validation & Conformance:
The service should validate the parsed address using an address validation service :
- Confirm the existence of the address
- Verify the accuracy of each component
- Correct any minor errors or inconsistencies
- Standardize addresses to meet local in-country formatting rules
- Verify Telephone number format by country
Multilingual Support
The service should be able to translate to & from the country-specific official languages. It would help to have a visibility on accuracy of results & some examples where the translation may not be optimal.
Address Parsing
The service should parse the given address into its component parts, identifying each of the input fields (house number, street name, apartment/building, subarea/district, city, region, country, postal code, and P.O. box).
Geolocation
The service should provide accurate latitude and longitude coordinates for validated addresses.
Data Enrichment
The service should attempt to auto-enrich the data. It should utilize the validated address information to populate any missing fields.
UNLOC codes - For each city, Google Maps should automatically update the corresponding UNLOC code if it exists. If a city does not have an associated UNLOC code, the system should provide the details of the nearest city with a UNLOC code, along with its distance from the original city.
Vessel Flag data and External Systems - The service should connect with AIS (Automatic Identification System) or other systems to retrieve accurate vessel flag information (such as the Underlying flag or Flag state)
Country coverage
We should be able to assess the extent of coverage offered by the address validation service to establish clear expectations for the users. A confidence score per country for coverage would help in validating the results.
Comparative Analysis Scope & Success Criteria
Below problems are currently faced with current solutions and we’ll perform a comparison as part of this PoC to understand the success criteria.
| Challenge | Expectation / Success Criteria |
|---|---|
| Accuracy level of suggested addresses is low as we navigate towards Street Number / House Number, Company/Building Name. The problem is broadly across the globe.Scenario #1e.g. 1901, Corporate Center Drive East, Monroe, Tobyhanna, Pennsylvania, United States.The address suggestion selected changes the City of this address to Coolbaugh.Scenario #2The current 'address suggestions' do not include the suite or apartment number. Thus users will ALWAYS have to 'over-ride' the suggestions to have the suite number included in the request. e.g. 3521 Warp St, Suite 305, Charlotte, NCScenario #3The address accuracy confidence score is not correct sometimes in the output of matching suggested address list.e.g. Below address is of good quality & matching but has low match score : 100, Gary Avenue, Roselle, Dupage, Illinois, United States, 60172Same scenarios can be replicated with other addresses | The address search is able to accurately provide suggestions based on input up to roof-top level. The service is able to accurately locate location based on the company/building name.We’ll also review the cases reported by our Area Data Stewards where address data quality is poor and confirm whether Google Maps is providing higher accuracy.Accuracy of detailed address information is expected > 99% |
| Improvements needed for address parsing capability & country specific address format standards | Effective address parsing is performed for the given input format, and the parsed suggestions are valid addresses.Accuracy of address parsing should be > 99% |
| A few countries only support local language output | We’re able to get the results in English as a bare minimum. Availability in local language is an added advantage - multilingual requirement.Availability of English language support in output should be 100% |
| Postal Codes are sometimes unavailable in address suggestionse.g. 94 Shehab St, Mit Akaba, Agouza, Giza Governorate 3752641, Egypt | We’re able to fetch the postal codes for missing examples, and for other happy path as well.Availability of accurate Postal Codes should be 100% |
Below are additional requirements which we’ll target to analyze independently (or compare with Azure Maps as per feasibility) :
| Criteria | Comparison / Success criteria |
|---|---|
| Multilingual Support (future plan) | Support for all languagesExpected accuracy for multilingual support is > 99% |
Geo-component validation (hierarchy Country-Region-City-Subarea/District-Postal Code <wherever applicable>) | Expected accuracy for Geo-components is > 100% |
Technical Requirements
Documentation
OAS/Swagger definitions for the API offerings along with availability & latency metrics.
API requests
A minimum of 50-75K address validation service requests available to evaluate the functionality and quality of the output for different countries.
Client secret and auth token requirements of APIs to be shared.
Relevant Questions
- Is there availability of 'what3words'?
- Is there a bulk validation endpoint / thick client software available?
Evaluation:
Please find the Google maps address validation api POC results in the attached link Googe Maps POC.xlsx