There is no question that there are enormous benefits from having the
ability to predict in detail, and as early as possible, what a tobacco
crop will be, and when various final blended and processed products
will be available.But predictions by "rule of thumb"
traditional methods can be disappointing and unreliable. So, as he has
in the past, Adams International Ltd. Managing Director Wing F. Chung
turned to state-of-the-art, cutting-edge technology to create a
reliable new system.
Growing and selling quality oriental tobacco is
Adams International Ltd.'s hallmark. But the Thailand-based company is
also known for its tradition of introducing high tech computer
innovations into its tobacco operations.
This season's new crop
forecasting innovation is particularly note-worthy and the most
ambitious and revolutionary computer application Adams has ever
undertaken.
Chung's new, unprecedented system - a pilot project this
year-seems to offer not only on-target, reliable crop predictions, but
also countless other benefits, including dependable leaf processing
scheduling.
Chung and his colleagues have invested in a new concept
in software-Object Oriental Programming (OOP), which computer experts
predict will transform the computer industry.
Software is the major
bottleneck and expense all computer users face. OOP removes the
bottleneck and, after initial investment, the expense.
OOP is a way
of building software from tiny, independent modules called
"objects." Think of building a hardware computer from a
multitude of standard small parts. Or even a Lego set, where you have
many generalized blocks that fit together easily to make just about
anything.
Each "object" encapsulates a specific, tiny bit
of data and function. It communicates with the outside world of other
objects via a message-passing system. Each object is designed so that
the details of the contents do not matter to the outside world.
If you send the right message, an object returns the right response or
performs the correct action.
The implications of this simple idea
are enormous. It means that software is easily changed or modified-any
change to the behavior or the internals of any object is completely
local to that object; such changes do not affect any other parts of
the system.
Objects can be wholly replaced, or new objects can be
quickly attached to the system to provide new functionality without
needing to deal with possible side effects on the rest of the system.
With OOP, there are no side effects if the system is correctly
implemented.
Therefore, software maintenance-the adding or changing
of software function-is cheap and easy compared with traditional
methods. The product and its descendants can be easily adapted to
other similar requirements, again very cheaply and rapidly compared
with the old way of doing things. In a word, an OOP system is as
generic as a software system can possibly be.
In addition, with OOP
the development environment is graphically oriented. Adams added
powerful artificial intelligence that can be brought into play as
required.
The application of OOP technology to business is
far-reaching, extending beyond easy and flexible programming. Adele
Goldberg, a former Xerox Parc researcher and now president of
Park-Place Systems, a maker of object based software, told
International NewsWeek last September that business people "want
to describe information in more general, real-world terms and create a
full simulation of what they
think is going on."
OOP technology is allowing Adams International to do just that.
The
initial OOP pilot project at Adams provides crop forecasting,
schedules the actual buying of that crop, models the processing plant,
and predicts the availability of final shippable product.
The
whole process begins with planting data. Actual numbers of tobacco
plants that have been planted are entered at the beginning of the
season. This data is gathered by Adams' extensive network of field
personnel. The first rough forecasts are generated from this planting
data alone, using a standard growth model developed by Karl Kunz many
years ago, and refined by actual experience.
The growth model takes
into account rates of growth, field-curing and so on, to estimate the
time from planting to availability of purchasable crop.
Of course
this forecast is very rough, since it must assume "typical"
growing conditions. Certainly many factors during any actual growing
season affect the resulting crop both in quantity and quality,
especially rainfall. So the growing crop is monitored in detail and by
sampling to provide refined data about how it is actually developing.
This additional data is added to the growth model as it becomes
available, and in this way the forecast becomes increasingly accurate,
step-by-step.
By the time the growing season is well underway, and
the first crops are actually becoming available for buying, the
forecast model is probably as good as it can be. Through OOP, a
forecast is provided in great detail. Adams knows how much tobacco
will become available, of what kind and quality, when and where. Using
this, it produces a calendar, taking into account holidays, minimum
"buy points" for locations and many other factors, that will
direct the buying teams to the appropriate buying stations on a
day-to-day basis.
Literally hundreds of pages of relevant printouts
can be generated by the system. For example, Adams generates detail
predictions for each tiny area under Adams' control. Through
individual area predictions, local agents know what is expected. Also,
Adams returns to each individual agent (there are hundreds) a detail
sheet showing the data that the agent has provided to the company's
system. In this way, the field agent has a chance to verify that Adams
has the correct figures.
Detailed predictions include breakdowns of
the tobacco by variety and quality, as well. In all, there will be
some 300,000 relevant "data objects" processed by Adams'
system this year. There is no possibility that any manual system could
hope to deal with this much data in methodical way or in a timely
fashion.
The use of graphics output which allows one to see at a
glance what is developing in the fields, is an extremely important
product of the system, These printouts are much easier to use than
lists of numbers, which have been produced by earlier systems. Such
lists can be produced as required, and in some cases they are
necessary. But for most purposes, the graphical summaries are more
than sufficient-in fact, they are a much better and more digestible
way to see what is happening or what will happen.
Given the system's
prediction of the actual buying of tobacco, Adams can obviously
predict how fast the inventory of green tobacco will grow. This
information is input to the factory - the Adams processing plant where
tobacco is cleaned, blended and otherwise processed into the final
baled and aged product, ready for the customers.
Factory rates of
production and other factors are modeled in the OOP computerized
system, too. Running this "model factory" against the
accumulating green inventory allows a prediction of final product
availability. In effect, Adams emulates the factory in the computer,
and runs it against the predicted available tobacco.
In this stage,
there are of course many options. Just how the green tobacco inventory
(variety, quality) builds allows various production options to be
brought into play. Depending upon the requirements of the customer,
various production options can be "tried out" on the
computer to optimize the resulting finished product. One can,
effectively, run "the factory" on the computer in "what
if" conditions to see what happens. If at any time some factory
conditions or customer conditions change, the model can be quickly
rerun to take new factors into account.
The end result is a very
clear prediction of Adams' entire production for the year- early
enough to help immensely with planning and marketing. No technology
offering anywhere near these capabilities has been available before.
Now Adams' experts have a methodical and precise tool at their
disposal, rather than having to rely on guesswork, which inevitable
invites mistakes. OOP, experts report, is very welcome indeed.
Adams
knows that as employees become accustomed to the system, and the
system gets used to the employees, it will be improved.
As well,
since this is its first year in action, the company knows it has not
provided data in every possible, useful way-but then, no one has had
such a system before. Adams International fully expects that many good
suggestions for improvements will be forthcoming from system users as
a result of this first year's trials.
But the beauty of OOP
technology is that Adams will be able to rapidly and easily
accommodate enhancements.
In the project's beginning, a special,
small internal group was formed at Adams to explore new OOP computer
technology and develop a coherent plan for using it. The group spent
more than a year on the architecture of the OOP-based generic business
concept.
The project was developed, and presently runs, on Sun
workstations. Very little about the project confines it to these
platforms; they were chosen because they were the most expedient at
the time. The very generic nature of the OOP design makes it possible
to move the project to other platforms.
Operating an OOP system is
exceedingly easy and requires no special skills. Probably less skill
is required to operate an OOP system than more conventional systems.
But, because the technology is so new, there are few people in the
world with OOP programming experience. As OOP moves into the
mainstream of computing, there will be a serious learning curve.
Often the payoffs of computer systems are hard to predict in advance.
Some payoffs are not easily quantifiable or are unexpected. But after
the fact, Adams discovered that its field buying system paid for
itself in its first season of operation. This has been the consistent
pattern for all of Adams' computerization projects-quick return on
investment.
How soon the new forecasting and planning system pays
off is hard to say, because the benefits of accurate, early forecast
data are particularly hard to quantify. No one doubts that benefits
will be substantial. Adams will have a solid handle on these benefits
after a post-mortem of its actual use for the first season.
TR's
prediction: look to Bangkok-based Adams International Ltd. to market
its tobacco OOP-based system to the industry once all the bugs have
been worked out of the system.
