i 4 vzesi xa 5 s
h MA aje aš PA F l B
Šadl ) | J i
Uri! s zadi. j
r a y j
j bu il !
] - - -.

Pace 54

APPLIED INTELLIGENCE

PC WEEKNAPPLIGATION DEVELOPMENT.

MARCH 12, 1990

Mlaj the Timebox, Development Deadlines Really Work

but the function-
altty may slip.
People in many
walks of life
have to meet,
deadlines. Writers, television producers
and seminar developers, for example,
must deliver material by a certain date,
which means the content of the materi-
al may slip.

Methodologies such as RAD apply a
similar constraint to the building of IS
applications: There is a deadline that; is
immovable, but the functionality of the
system may slip. The system must work,
carrying out its basic functions, but the
refinements may have to be postponed
in order to meet, the deadline.

Seventy-five percent of a system's
functionality can be created relatively
guickly with an automated code genera-
tor, especially if reusable structures are
employed.

The next 15 percent may take as long
to create as the first 75 percent, and the
system may have to be successively re-
fined before the last 10 percent is com-

MARTIN

er on

Many of the features in the last 10
percent could be dropped or postponed
for a subseguent, release.

If the refinements are added after the
first version of the product has been in
use, they will probably be subject to
change.

Adding Refinements

it is often better to add refinements
after the first version has been in use
because users change their minds about
the reguired functionality of a system
moi sle with it for a while.

e dangers of prototypin
methodologies or iterative  oninent
is that the functions of a system can
grow in an uncontrolled fashion. Users
or developers pušo functionality so
design does not converge guick-
ly into a usable system. This is releje]
to as creeping functionality."

Ne | se | eš z a
| , t
ni sa »d ii

; built | | K and the : efi |

9 The first version must be built
his. guickly,. |
event. The application must be built so that

- ., j LA
Me žek di Uber
ari Bi " ML - | sei boma A m a " ij
nk v ai a i a NM | a
Vač K, ' ii ri sa 5 . k, h La zi 4d: ka im : LA j k. Z ki j k
išče s J a ai db K | id - ; gi , a je k ini
Aamha — o ZUCLNOGOLOSVY Š
Be Metnodology works. for the compa-
E o | pet: ke a. se, ci ni se - ! in - |
PE AE MES lit Va |
-. ka. bišti: Kak beni h s A a i ž ri

ting a deadline practical. The Timebox

e cycle allows refinements to be made
until the deadline nears, at which time a
working system must be delivered.

Du Pont's Timebox was first used in
the company's fibers division, which

was moving to a highly automated man- |

ufacturing environment. It was neces-
sary to create complex application soft-
ware guickly, and Du Pont recognized
that it is better to get a basic version of
the system working, learn from the ex-
perience of operating with it and then '
design an enhanced version than it is to
wait for a comprehensive system at a
later date.

Du Ponts experience implies that:

Applications
Development
Within a

išči (or SWAT) team is given a
e frame within which a syste
be constructed. zene

Prior to the assigned time frame, the
functions and design framework of the
system are defined. After the time
frame, the system is evaluated and a de-
cision is made as to whether to put it
into production.

The time frame cannot: be extended;
however, the functionality of the system
may be trimmed to complete the system
within the time allotted.

Within the time frame, continuous it-
erative development is done with end
users and IS developers working closely
together. The team is under pressure to

Reguirements Planning X

N
4
ŽA
n N
Ma .
H 5,
aa
1 ] 4
V X

| User Review XX

— Reguesi for Change A

j ; ——————-—- J.J JJ JJ JJ JJ

-Timebox

"ad zd, A p 5. |
AE ai ok ae S
tie ni deia
krešuiij

O RRA A A  Aai R EI
» sla Ai RA Ma: le di
dal, HE tla nejettik > sh, krti

gi k iš MP ii ka LA A uli P
m em Era ha ip ure,
f He: Re ea ja Ma ME
a %.: Seiko " mi dodaja ji š
vi

| Evaluate System V

X
|
|

mi

John Avakjan

k 5 - - zij ka i: uč B ue j k
k, j ri] ni 4 A, un % d a k JE

Within the time frame, continuous iterative development is

done with users and IS developers working closely together.

The team must produce a working system by deadline.

UE DOFrALJONS

č Erepok ra nm

produce a working system by deadline.

A Timebox approach, or team devel-

opment in general, does not. work well if

—there'is pressure to meet an impossible

deadline.

The setting of the deadline, or the se-

lection of functions to be accomplished
within an allotted time, should be the
responsibility of the team or systems an-

miliar with the tools and technigu
and confident in what. technigues,

Before deciding ona

pa t working on the project. Meeting

deadline is easier if the team is fa-  .

they can accom-

alistic deadline,

— the SWAT team should make its own es:

timation as to how long the project will
take. This estimate should be made at or
before the end of the final joint applica-
tions design (JAD) workshop, within the
RAD user design phase.

By the end of the final JAD work-
shop, the following should be estimated:
the number of function points, inputs,
outputs, logical internal files, external
files and gueries, and their complexity
Chigh, medium or low).

An experienced SWAT team knows
what it can achieve. It examines the
JAD and uses its own guidelines for de-
termining how long the project will take
and whether additional assistance is
needed.

It may allocate a certain period of
time for startup, a number of person-
days for each transaction type, a period
for integration testing and a period for
handing over the project.

SWAT teams should work on systems
or subsystems that, can be completed in
a short time. If the overall project is
large and complex, it should be divided
into subsystems, which can be built at
the same time by separate teams.

The SWAT team effectively signs a
contract with management, saying, "We
will build a system of that scope by a
certain date."

Changing the Scope

If the scope changes during the con-
struction phase, more development time
may be needed, or reguired functionali-
ty must be dropped. Any change in
scope should be agreed upon with man-
agement-during the construction phase.

Limiting functionality to meet the
deadline does not, mean guality must, be
compromised. It is essential that SWAT
teams have pride in their work; no self-
respecting team would put out. work
that, was less than the highest guality.

The delivered system must be as bug-
free as possible and must; provide a set
of functions that. meets the users needs
when the system is put into production.
Integrated CASE tools make it much
easier to achieve technical guality.

Functionality can vary substantially
without reducing the usability of the
system.

Many software and electronic prod-
ucts today have excessive funetionality
that often bewilders users. The return
on development, investment, goes down
when unwanted or marginal functions
are added.

The Timebox development concept
places an inflexible limit on the func-
tionality to be supported and the time
taken to produce a working system.

Next, week I will discuss how innova-
tive management technigues can lead to
higher productivity. B

The concepts embodied in RAD are de-
soribed ix a neto volno ga Nana
Martin Report Series. For more tnjor-
mation, on this volume, call (800) 242-
1240. For information on seminar, con:
tact (in the United States and Canada)

Technology Transfer Institute, 741 10th

4. Santa Monica, Calif. 90402 (213)
994.8305. In Europe, contact Savani, 2
New St., Carnforth, Lancs., LA5 9BX
United Kingdom (0524) 734 505.

nsaolo MM | K«