Addressing and the new postcode system by gyvwpsjkko


Addressing and the
new postcode system
by Pierre Rossouw, South African Post Office

The project for a new postcode system for South Africa has been approved and is under way. The system is
designed to take advantage of modern processing technology, advanced database functionality and integration
in a hierarchical structure, to delivery point.

    t is designed not only for postal        Thus, the coding of address data needs    entire address, provided that certain
    services (rural, urban, street and box   to be brought up to the same level        print fonts were used. Today, the
    delivery) but also for many other        to enable maximum performance.            best automatic reading technology,
purposes such as local and national          This project is part of the long-term     using different algorithms, can read
government service delivery, commercial      automation strategy to support the Post   up to 98% of machine-printed writing
and academic research. It is designed        Office’s vision and mission, which aims   (characters separated) and up to 85%
to be possible to integrate with other
                                             at being amongst the best in the world.   of handwriting (joined script), on first
location systems. This paper describes
                                                                                       pass reading. To enhance this, multiple
elements of the research, design,            Technology development                    cascaded methods are generally used.
development and implementation.
                                             Three generations of technology           The SA Post Office uses three cascaded
Historical background                        have come to pass since the early         levels of reading, giving an overall read
                                             technology. The early optical character   rate of over 99%.
The existing four-digit postcode no
longer satisfies the requirements of the     reading (OCR) technology of the 1980s     Automatic reading technology has
SA Post Office. The time has come to         was capable of reading the postcode       advanced beyond OCR. It also looks for
develop a new postcode system that           only. In the 1990s this technology was    patterns and shapes in the high and
is compatible with present-day and           capable of multi-line reading of the      low parts of script. Almost any known
future requirements, takes advantage
of modern technology and performs to
high levels of process efficiency and cost

The current postcode system was
launched on 8 October 1973. It contains
data indicating the postal region and
office identifier, but no data for finer
sorting than this. Finer downstream
sorting is done manually. Manual
processing is costly and results in a
certain amount of missorted mail that
must be reworked. The current system
was designed before the implementation
of the hub and spoke system and before
various technical developments such as
multi-line optical character recognition
and dynamic database integration. We
may find the first automatic reading
technology antiquated today, but at
the time this was the state of the art

In South Africa, many technical, social
and economic developments have come
to pass over the last three decades.
Keeping up with such changes has been
challenging. In recent years, the Post
Office has invested heavily in capital
equipment for processing in main mail
sorting centres, but the advanced
technical capabilities of our new sorting
machines are constrained by the
poor and inadequate addressing and
postcode systems.

PositionIT - Mar/Apr 2008                                                                                                          59
     Application           technical

     type of writing and language, from Latin     examples are national census, national       However this is not always the case.
     characters to Cyrillic, Arabic, Hebrew and   elections, infrastructure planning and
                                                                                               Address structures were thoroughly
     Chinese can be recognised. They may          standardised address structure. This also
                                                                                               researched. This topic is well covered
     use several cascaded methods of reading      implies a geographical structure which
                                                                                               in the draft SANS 1883 “South African
     and, using neural network methods            would permit other possible uses such as
     (“brain modelling”) and fuzzy logic, are     emergency services location as well as       Address Standard”, expected to be
     also self-teaching.                          for academic and commercial research.        published in 2008 [4].
                                                  Such applications require that the system    Addresses are classified into specific
     More recently, pattern recognition
                                                  conforms to open interface standards.        categories or types: they may be
     techniques also enable the reading of
     logos, stamps, franking indicia and the                                                   urban or rural, formal or informal,
     like, enabling integration with revenue                                                   street or box. People do not always
     control systems.                             In the first project, the postcode           write formal standard addresses, so
                                                  systems of all Universal Postal Union        other forms must be accommodated.
     However, world class service to
                                                  (UPU) (a United Nations body)                The most basic is a typical PO Box
     customers requires higher performance
                                                  member countries were researched.            address, which comprises four lines as
     and these automatic reading capabilities
                                                  International standards and compliance       follows:
     are often considered not good enough.
                                                  requirements were considered.
     Barcode reading offers even higher                                                          Addressee name
     performance. Readers of special              From the research, several important           PO Box number or Private Bag number
     barcodes, incorporating error-correction     principles were established and the            Post Office name
     algorithms, read correctly at rates          design of the new structure evolved.           Postcode
     in excess of 99,9%. The same error           One very important and influential           This is a very simple and standard
     correction algorithms are used for radio     finding of the review project was that:      structure. Although simple, this type
     signals such as those transmitted from
                                                  “The good function of a postcode             of address does not satisfy various
     deep-space probes.
                                                  system depends entirely on having            requirements for service delivery. Also,
                                                  a good addressing system and good            various legislation (such as the Financial
     Current requirements
                                                  base data. Unless the basic addressing       Intelligence Centre Act) specifies that
     The shareholder of the Post Office issued                                                 physical addresses are required. An
                                                  systems and structures were corrected
     a mandate which included, amongst                                                         example is the formal urban street
                                                  and formalised, a redesigned postcode
     other things, that “…a postal code                                                        address structure, which is usually fairly
                                                  on a flawed address structure would
     system facilitating more accurate market                                                  simple and standard:
                                                  not solve problems. Automating flawed
     segmentation and address verification”
                                                  base data will make address flaws             Addressee name
     [1] was required to be developed.
                                                  more difficult to find and resolve; non-      Street number and street name
     This gave rise to two specific projects:     conformances and failures would in fact       Geographical place name
     Firstly it was required that the existing    increase” [3].                                Postcode
     postcode system be reviewed and
     researched (completed and reported           Thus it was necessary to substantially       However this is often not adequate
     in March 2004) [2], and from this,           formalise and standardise the address        for unique delivery point identification
     secondly, that a new postcode and            structures first, before redesigning the     (DPID). The same place name can exist
     addressing system be designed,               postcode system.                             in different areas of the country, so a
     developed and implemented [3]. The                                                        town/city or region may be required in
     latter is the current project which has      Standardised address structures
                                                                                               the second last line. The address may
     now received full approval from the
                                                  People structure and write addresses in      refer to a multi-dwelling location such
     Department of Communications.
                                                  a hierarchy, from the bottom up. The         as an apartment block or residential
                                                  concept is that, within a specific level,    complex or multi-commercial building
     Objective of the project
                                                  the sub-level is uniquely identifiable.      such as a retail shopping complex or
     The primary objective is to design and
     develop a system that satisfies the
     service delivery requirements of the Post
     Office and also those of interested and
     concerned parties.

     The system is required to be compatible
     with the mechanised and manual
     sorting processes, the hub-and-spoke
     distribution system (based on
     geographical location and transport
     routes) (see Fig. 1.), relational database
     mapping technology and quality
     conformance. It must be reliable,
     stable, robust and universally applicable
     throughout various environments for
     many years into the future and thus
     must be flexible for maintenance and
     upkeep purposes.

     The system should also find application
     outside of the postal system. Such           Fig. 1: Schematic hub and spoke structure.

60                                                                                                               PositionIT - Mar/Apr 2008
     Application           technical

     large industrial facility or a farm. Thus    error-prone. Few people carry a GPS                The postal hierarchy is similar: There are
     more lines may be required above the         device or can use one satisfactorily.              six regions, 26 mail centres or hubs, 390
     street ID to indicate a subdivision.         Most importantly, people around the                postal routes, over 7000 postal offices,
                                                  world do not describe an address                   depots or agencies, about 45 000 walks
     Another Post Office initiative is the new
                                                  in terms of coordinates, but as a                  and over 10-million delivery points.
     rural addressing system, which was
                                                  structured hierarchy, such that it
     described previously [5]. This type of                                                          While the geographical structure is
                                                  forms a set of instructions to identify a
     address is structured as follows:                                                               largely a description of where a place
                                                  place. It is necessary to accommodate
                                                                                                     is and is a natural attribute, the postal
       Addressee name                             the structures and methods that
                                                                                                     structure is concerned with how to get
       Rural address code and place name          people use in practice. Also for many              there, which is synthetic. It thus lends
       Post Office name                           users of geographical information,                 itself to optimisation.
       Postcode                                   areas are required rather than point
                                                  locations (examples are for population             Mail is collected and brought to a mail
     This meets the requirements of a             segmentation, electoral wards, census              centre where it undergoes outward and
     physical address and is classified as        data). For such reasons, the hierarchy             inward processing. After cancellation,
     such.                                        system is used because it meets user               it goes through an outward sorting
                                                  requirements.                                      process where it is primary-sorted and
     It is intended that as far as possible
                                                                                                     transported to another mail centre
     all SA addresses conform to one of the       First, consider the geographical                   (except for local-to-local mail). There
     above three structures. One reason for       hierarchical structure. Within the                 it undergoes a secondary inward
     these specific structures is to suit the
                                                  country there are nine provinces, 53               sorting process to a delivery office or
     OCR or automatic reading technology
                                                  districts, 261 municipalities, 3125 main           depot, where it is sorted to a walk,
     used in processing machines.
                                                  places and 21 243 sub places. There                then sequentially sorted for final
     This follows a programmed logical            are about 800 000 streets and over                 delivery. As a letter goes through the
     methodology: After locating the address      12-million specific locations or delivery          mail stream, the number of possible
     block as the region of interest, the         points.                                            destinations increases. Sorting
     automatic reading system reads from
     the bottom of the address up, first              Origin (source)
     attempting to resolve the postcode. It
                                                      -   Mail originator
     verifies this against the place name or
     post office name, then on the next line          -   Outward (origin) mail centre
     up, reads the box number or street
     name and number. These are verified
                                                     Geographic location (where a place is):
     against an address directory.
                                                     -    Country
     The logic incorporates some decision
     making. If certain conflicts arise, it          -    Province
     uses a set of rules to make decisions.          -    District/municipality
     If it finds more than a certain number
                                                     -    Main place (town, city)
     of conflicts or incompatibilities or if
     the address is irresolvable, it sends an        -    Sub place (suburb, township, village)
     image of the address to a video coding          -    Street name (street, road, block, zone, section)
     operator who will key in the correct sort
     destination.                                    -    Street number

     In a mail centre, this image management
     system runs on one computer for three           Routing and delivery (how to get there)
     (or more) machines. This computer               -    Inward (destination) mail centre
     reads and processes in excess of 120
                                                     -    Transport route
     000 items per hour, more than 30
     per second. Despite this, there may
     still be failures or rejects for rework.        Street Delivery:          or       Box Delivery:       or         Rural Delivery:
     Most of these are due to irresolvable
                                                     -    Delivery Depot                -   Lobby Office               -    Post Office
     addresses, which are material faults. This
     is unfortunate because delivery quality         -    Walk                          -                              -    Village
     depends on address quality, which is            -    Walk Section                  -   100-box range              -    Village Section
     determined at the source, i.e., with the
                                                          -   Delivery point            -   Box number                 -    Delivery point
     mailer. To achieve the required levels of
     performance, this must be at the highest
     possible levels of conformance and           Table 1: Each set of data is structured as a hierarchy for coding.

     Address system design

     There is a frequent argument that
     coordinates will uniquely identify a
     location. The research finds that this
     is not very practical and there are
     some specific drawbacks. Capturing of
                                                  Fig. 2: Data structure in the code.
     coordinates is laborious, expensive and

62                                                                                                                         PositionIT - Mar/Apr 2008
                                                                                              Application           technical

decisions are made at each level. This        of an official, usually the postal branch   international standard. A first-pass read
progressively increasing complexity of        manager. Such items will include Poste      error rate of less than 1% is expected,
the process implies increasing processing     Restante mail and internal mail items       which is a ten-fold improvement
requirements. Thus, including data down       specifically for that branch.               compared to current automatic first-pass
to final delivery point is essential for                                                  read rates. This is made possible
                                              Alpha characters are used only for
automated sorting systems.                                                                by incorporating error-correction
                                              country ID and for mail centres. The rest
It is true that a random unique               of the data are represented by numeric
identifier can be allocated for automated     characters. In compliance with UPU          The layout and format
processing. However some processing           specifications and other international
is done by hand (for example, parcels,        conventions, the order of the data sets     The components of the data groups are
small packets, newspapers, low volume         is modified: the country ID is put first,   combined to form a character string
non-standard items, and not all               followed by postal routing data.            and printed at the top of the address
centres are mechanised), thus it is also                                                  block in Latin alphanumeric characters
                                              Alpha characters are usually avoided        and in barcode. The character format is
necessary to have a comprehensible
                                              because of confusion with numeric           purely to facilitate human interpretation
                                              characters and lower read rates.            without using a decoding device. The
The code design                               However, in barcoded format this is not     order is changed for some data: The
                                              an issue. This is discussed below.          UPU requires that the country identifier
Coding removes ambiguity and name                                                         is the first element; also, some data
confusion. Once a specific location is        The comprehensive address code is
                                                                                          is relevant only for mailing (origin,
coded, it is uniquely identifiable. The       represented by a specific barcode.
                                                                                          tariffs, and customer references) and
comprehensive address code describes                                                      thus is not shown in the alphanumeric
                                              Barcode representation
the journey of a mail item from                                                           representation.
origin to destination. It also contains       The barcode representation of the
internationally-compliant data and some                                                   Initially this system will be used by
                                              address code is for accurate high-speed
other useful information.                                                                 commercial bulk mailers (for example,
                                              machine sorting. Being linear, it is
                                                                                          banks, retail marketers and service
The code comprises four basic types or        suitable for low-cost printing as it can
                                                                                          providers like SARS, telephone
sets of data:                                 be electronically represented as a
                                                                                          companies and municipalities), which
                                              simple character font (as opposed to a
•   Origin: identification of originator.                                                 supply the major portion of mail
                                              graphic) allowing it to be printed by the   volumes. It is intended that the postal
•   Postal routing and delivery data,         mailer as part of the address.              routing code will replace the postcode in
    from postal region to delivery point                                                  the last line when the system is rolled
    identification.                           Being a base 4 construction it has a
                                              high data density for a linear code.        out to the general public.
•   Geographical data, from country to
                                              Two bars are sufficient to represent up     The structures for urban street
    specific delivery point location.
                                              to 16 characters (all numerics); three      addresses, PO Box addresses and
•   Other information such as postal          bars represent up to 64 characters (all     rural addresses are shown in Figs. 3,
    tariff data and customer data.            alphanumerics, upper and lower case).       4 and 5. These show some address
Each set of data is structured as a                                                       components represented as data
                                              Because the 4-state configuration
hierarchy for coding,                                                                     elements.
                                              allows base 16 and base 64, bars
(see Table 1). These sets of data are                                                     The formal urban street delivery
                                              can represent characters that do not
separate databases and combined for                                                       address (see Fig. 3) contains descriptive
                                              exist in base 10 or alphabets, such as
specific purposes.                                                                        geographic information. The second last
                                              punctuation. Hyphens and exclamation
These data are coded and structured as        marks are examples of characters used       line is a place name. Without the code,
shown in Fig. 2.                              in place names.                             postal sorting, routing and delivery
                                                                                          information is not explicit: it is derived
The first two sets represent the journey      The 4-state barcode configuration is        from the address using relational
of a mail item from start to end. Each        rapidly becoming widely used as an          database references.
delivery point needs to be coded once
only. Once this is coded, the system
is not affected by name changes
and spelling. Boundary changes and
cross-boundary issues are simple
to accommodate by changing one

There are certain rules used for the data
constructs. Zeros are avoided wherever
possible. A zero on this structure is
interpreted as a null value and indicates
no knowledge or confidence at a
particular level. If the exact location is
not known, the last few characters may
be zeros; however the delivery point
can still be located, usually through local
knowledge. Zeros also indicate that data
is required to be checked and captured.
Zero may also indicate special attention      Fig. 3: Urban street address format.

PositionIT - Mar/Apr 2008                                                                                                              63
     Application           technical

                                                                                              It will become possible to do all primary,
                                                                                              secondary and delivery sorting by
                                                                                              machine system; currently much of this
                                                                                              is done manually. Mail can be taken
                                                                                              directly from the processing machine and
                                                                                              delivered. This will have a major impact
                                                                                              on the secondary processing work, and
                                                                                              consequently on time and cost.

                                                                                              The impact on quality will be very
                                                                                              significant. Compared to a first-pass
                                                                                              acceptance rate of less than 90%, the
                                                                                              4-state barcode has been found to
                                                                                              achieve about 99%. This represents
                                                                                              a reduction in first-pass read failures
                                                                                              from over 10% to about 1%, a ten-fold
                                                                                              reduction in rework.

                                                                                              The system can be standardised across
     Fig. 4: Post Office Box address format.
                                                                                              all mail types: franked mail, hybrid mail,
                                                                                              parcels, courier, track & trace. Analysis
                                                                                              is possible by bulk mailers, hubs,
                                                                                              routes, transport, walks, addresses,
                                                                                              revenue, geography, demographics
                                                                                              or by any other data contained in the
                                                                                              code. The system is to be internationally
                                                                                              compatible according to standards and
                                                                                              provides capability for foreign sorting. It
                                                                                              will be possible to sort for international
                                                                                              destinations, subject to having the
                                                                                              relevant format controls and the foreign
                                                                                              sorting plans.

                                                                                              Return-to-sender (RTS) mail processing
                                                                                              is a very slow and expensive manual
                                                                                              activity for which no revenue is
                                                                                              received. Currently, RTS mail can be
                                                                                              machine-sorted, yielding a first-pass
                                                                                              acceptance rate of about 40%. With
     Fig. 5: Rural address format.                                                            the new barcoded system, because the
                                                                                              origin information will be included in
     The PO Box or Private Bag address             hybrid mail systems, and back to bulk      the new barcode, RTS can be machined
     (see Fig. 4) contains simpler postal          mailers, who will all be able to use the   to the same accuracy as normal mail
     information for sorting, routing and          same sets of data, thereby minimising      (about 99%). Further, mailers frequently
                                                                                              do not want/need the actual RTS letter
     delivery. The second last line is a Post      address errors.
                                                                                              returned, just the information to correct
     Office name. This address requires
                                                   The results of the mail process are        an address record. It will be possible
     fewer steps to sort and the database
                                                   passed through a data fault analysis       to run a program to scan and capture
     referencing is only for range checking
                                                   and learning system to analyse failures,   letter images and send just the address
     and address validation.
                                                   rejects and other non-conformances.        images back to the bulk mailer.
     The rural address (see Fig. 5) contains       The corrections are fed back to the
                                                                                              There are other significant potential
     descriptive geographic information as         original sources of the data. It is most
                                                                                              impacts for mailers. Lodging patterns
     well as postal routing information. The       important that the original sources
                                                                                              analysis for operational planning
     second last line is a Post Office name.       are part of the entire data flow,
                                                                                              can be linked to machines. Accurate
     The dwelling identifier is associated with    otherwise the same errors arise again.
                                                                                              measurement of quality and volumes will
     the Post Office and the local agent. Postal   The objective is that any particular
                                                                                              improve, resulting in improved costs and
     sorting information can be readily derived.   error, once corrected, should never be
                                                                                              rebates. Address verification has already
                                                   encountered again.
     Integrated data flow and                                                                 been indicated as very important. It will
     feedback                                      Features, impacts and benefits             be possible to check addresses on-line
                                                                                              and to verify geographical location
     Operationally, a master data coding           The system offers a number of possible     through GIS integration. This needs to
     engine collects mailer data, postal data,     benefits for the Post Office. The          be done once only for each address.
     geographic data and other data. It            centralised address data and sort plan     Receivers will be able to verify that a
     verifies the data and combines the data       management promotes significantly          letter was sent from a mailer; it will be
     to generate an address code string for        increased accuracy and consistency.        possible to follow a letter through the
     each item. It updates the automation          There will be a substantial reduction      mail stream, (although not quite the
     address database and machine sort             in secondary manual processing and         same as Track & Trace). Mail forwarding
     plans. These coded data are fed to            delivery processing. Rebate structures     for change of address purposes will be
     sorting machine systems, electronic and       will be improved.                          vastly improved.

64                                                                                                              PositionIT - Mar/Apr 2008
                                                                                             Application         technical

There are also potential benefits for        Another example: When using               This project is expected to run for
other organisations. State/government        internet access, such as through          several years. The Post Office will
planning and analysis will be facilitated.   a Public Information Terminal at a        endeavour to involve organisations,
Research and analysis for commercial,        remote Post Office, and by inputting      private, public, municipal and national,
marketing, academic and other                your code, your geographical              to participate in this project in order to
knowledge-based research and analysis        location can be verified via links to     gain maximum benefit for all, according
will become simpler.                         the Post Office system or another         to the mandate and project objectives.

There are opportunities from data            organisation’s GIS. You will be able to
sharing agreements with and services         see a map or aerial photo or satellite
                                                                                       [1]   Government White Paper on Postal
to external organisations regarding          image of your location. You will be
                                                                                             Policy (14 March 1998) section 2.2
geographic and demographic information.      able to access another GIS system
                                                                                             and other sections.
Government departments, e.g. land            in another organisation such as your
                                                                                       [2]   Pierre Rossouw: Postcode System
affairs, agriculture, SA Police Services,    local municipality or StatsSA to see            Review Project – Preliminary
home affairs have indicated requirements     maps and aerial or satellite images of          Investigation and Proposed Way
for our system. Some government              your location.                                  Forward, March 2004
related organisations, e.g. SABS, CSIR,                                                [3]   Pierre Rossouw: Postcode Renewal
StatsSA, Eskom, Telkom and IEC have          Or, you will also be able to verify
                                                                                             Project – Proposed New Postcode
shown interest. Some municipalities are      your location using a cellphone. You            System, 20 January 2005
keen to share street coding data and         will be able to access your location      [4]   Serena Coetzee, Antony Cooper and
the new postcode system data to enable       information by text message on your             Paul Strydom: “Spatial standards
advance notification (before deed) of        mobile cell phone, which indicates              make your life easier” PositionIT,
suburbs/townships, streets, addresses.       your GSM cell, geographic location              Jan/Feb 2007
                                             and postcode. Further, this will          [5]   Pierre Rossouw and Keletso Kgope:
This cooperation makes it possible to                                                        “Rural Addressing in South Africa
                                             not be limited to South Africa; it is
register and verify an address before                                                        PositionIT, Sep-Oct 2007
                                             intended to be international. These are
a residence is bought or occupied.
                                             just some of the possibilities of the     Contact Pierre Rossouw, SA Post Office,
Sharing of information of infrastructure
                                             new postal addressing and postcode        Tel 012 401-7465,
development, from source, on deed,
well in advance of ownership or
occupation, for our strategic planning
will be mutually beneficial. Private
sector businesses are also interested
in data sharing; pricing concepts
according to raw data and added value
have been discussed and principles
agreed. There is also much interest
from the GIS, real estate and market
research industries.


Imagine that you are at a point where
you fill in your address on a form and
it is captured in a system. This may
be to open a retail account or bank
account, to apply for municipal or
government services or some other
type of registration.

The service person on the other side
of the counter captures your address
details and checks this through the
secure Post Office website. Your
address is verified and your address
delivery point ID code is sent back
to the retailer, on-line. The service
provider’s mailing address database
is updated accordingly with your
correct, valid address (this needs to
be done once only for each address).
The service provider can give you
your delivery point ID code of your
address, together with your address
barcode, for your own general use.
The server, when invoicing you, prints
the barcode as part of the address.

PositionIT - Mar/Apr 2008                                                                                                           65

To top