A — PAGE 76 Gain Management SUPDD This is the third of a sertes of articles on rapid applica- tion develop- ment (RAD , a development methodology de- signed to be much faster than the tradt- tional method. SE delujete the information: K: (s NefIN systems (IS) or- oi ganizations strive | to develop sys- tems that are faster, higher in guality and lower in cost. Uhis can be achieved by using new de- velopment technigues such as RAD. The RAD life cycle is based on the use of in- tegrated computer-aided software engi- neering (CASE) tools, small teams of highly trained and motivated personnel, interactive joint application design (JAD) technigues, and management pro- cesses focused on breaking down obsta- cles to high-speed development. RAD is an effective management, tech- nigue for application development. It fo- cuses on the management technigues needed to facilitate communication be- Wween end users and analysts, motivate the development team, build complete applications within a strictly defined window of time, and ensure that com- pleted applications meet end users' needs. Approximately 80 percent, of RAD's elfectiveness is attributable to better management processes; only about, 20 percent is due to the use of computer technology. - One of RAD's management technigues is the measurement, and reporting of de- velopment productivity. Appropriate measurements enable managers to moti- vate and set goals for developers. They nigues, and enable managers to detect, examine and, possibly, replicate exam- ples of excellence. Periodic measurements of produtctivi- - sjon lines, but should exclude comments. nization. permit the comparison of tools and tech: | IR ai zirE | | Ki ih | | | ti gi | s j u h "PG WEEKNAPPLIGATION DEVELOPMENT APPLIED INTELLIGENCE EI Dei - mra cutable lines of code and the Data Divi- vendors' claims. Some individuals re- It's desirable to measure project com- plexity independent of any particular computer language. dkamijj? The complexity metric that's most commoniy used is the function-point method. This was developed at IBM and improved in GUIDE, the IBM users' orga: per person-month. s PLE For systems of average complexity The function-point method lists and without elaborate user interfaces, one counts the elements of a system, in- cluding; about, five lines of COBOL per person: 8 inputs (screens, messages and batch day. | | | transactions); | For systems with human interfaces % cutputs (screens, reports, messages such as a mouse, icong and pull-down and bateh transactions); menus, the ratio becomes higher. Differences in Development — Productivity RA MA NI ae s nil a na ti (raza Productivity Function points per person-month I irtr ra 4 | ji 1 k! Ee MLAY | 4990 49832. | lk | Mi MI 1 8 1K trat TE Ii 'y enable managers to establish guali | kih Z EP veji PE JaN ie rds and reward errmeliknjit! Noe h OMEJI | bp eee: le tiči č | ume jat achleve high levels of pro- a S | ana ai avality. Measurement Periodic measurements of productivity enable managers to sregoidenih a aii ME po establish guality standards, identify areas in which improve- to ao dogolnij prosrca ments are and establish a program for improvement 4 com system data stores (logical files): | | x nei (data stores shared with other systems ti 118 US. average development produc- pbezlev, O terface files); and morah Ve function points per person: ema pa ooo re podi to about 26 ine Zase when Nag: Mar ly organizations develoni ma bul wi tir ; tems with COBOL database Mojo and ie pomiki ig the i h o njoki o Cycle average lines of code, a sabout eight NE per person-day for | bek month) PRA points per person: tion | vad, aids with BOL projects t sne elopment pesi s; SA. O COBOL must be counted with ee mane |EINA | Saution, The count should include exe. sadi s dliTec ko have a basis for com. id 11 "'rent projects, estimates or' peče today's power tools | | | za " io achleve substantial- member measurements such as eguiva- lent lines of COBOL per person-day, re- lating this perhaps to their own experi- ences in programming. Other individuals 'remember function points per person- month. I recommend that IS executives measure and remember function points function point per person-month eguals (0524) lede td rila varov devete | end eet or tete rooza; Mao | si sa Mili bei bbika JANUARY 8, 1990 ort by Measuring Productivity Iy higher productivity numbers. The figure shows measurements of IS productivity at Marine Midland Bank in Buffalo, N.Y. COBOL development had reached a plateau of about 12 function points per person-day across all mea- sured projects. The first introduction of 4GL technol: ogy using the Telon code generator from Pansophic Systems Inc. resulted in slightly lower productivity, but the staff rapidly climbed a learning curve, build- ing to about 40 function points per per- son-day. This is a fairly typical number for well-managed use of a 4GL/CASE tool with a code generator. Using an inte- grated CASE environment and a tighter methodology, the productivity figures are likely to increase. Highly trained teams using integrated CASE tools and a RAD methodology can achieve much higher productivity [ig- ures. Some teams regularly achieve 1,000 eguivalent, lines of COBOL per person-day (or 200 function points per person-month) on projects of moderate complexity. Measurements in the Du Pont Fibers division show skilled developers averag- ing 13.8 hours per function point with a traditional Yourdon structured method- ,Ology, while other Du Pont analysts were averaging two hours per function point using its "time box" methodology, which is a version of RAD. Some large and bureaucratic applica- tion-development projects average only two or three function points per person- month, at a cost of more than $1,000 per function point. Some software-devel- opment companies charge more than $1,000 per function point. IS organiza- tions typically average about $500 per function point. A reasonable target for a RAD project is about $100 per function point (corresponding to 100 function points per person-month).. CEO Productivity Report A corporate president should monitor IS productivity because of its vital ef: fect. on corporate capabilities. The CEO report should include five types of measurements: 8 speed of development; $ development costs; ? development productivity; ? guality of delivered systems; and ' maintenance. When this type of report is reguested on a guarterly basis by the CEO, the IS manager is motivated to strive vigorous- Iy for the most, efficient use of powerful development technigues, Next. week I will discuss the automat- ed tools that are appropriate for use in a RAD life cycle. 8 The concepts embodied in RAD are de- seribed in a new volume in the James Martin Report Series. For more infor- mation on this volume, call (800) 242. 1240. For information on seminars, ado) Tobak vaj jE ao A »ehnotogy Transfer Institu 741 10th St, Santa Montca, Calif. | 90402 (213) 3948305. In Europe, con- tact Savant, 2 New St., Carnforth, LA5 9BX United Kingdom 734 505.