Customization of Function Point Analysis at Nokia

Petri Vesterinen and Kari Kansala (ed.) Nokia Research Center

P.O.Box 45, FIN-0021 1 Helsinki, FINLAND tel: +358 0 43761, fax: +358 0 4376 6855

email: petri.vesterinen @ research.nokia.com

9 NOKIA

Nokia Corporation

.Net sales 37 billion FIM (ca. 8 billion USD)

.Total personnel 33 800

.Sales in 120, production in 14 countries

.Main products: mobile phones ,

digital exchanges

-a I telecommunications networks -- -.*. --

satellite and cable receivers, multimedia monitors, car electronics

Introduction

A SPU (SW production unit) producing - telecommunications SW for Nokia's DX200 switches

importance of accurate estimation of effort and duration of SW projects formal estimation technique based on FPA approach wanted no earlier FPA experience within that SPU

NOKIA ?f

Problems with FPA

System boundary choosing the right abstraction level

Identification of basic function types difficult architectural impact strong, much variation, dependable on person who counts

IFPUG function types do not depict all aspects of switching systems separate new and modified functions

General system characteristics SPU-specific, unequal impact range

@ NOKlA COCOMO Forum, 9- 1 1th October 1996 / 4 / Peln Veslerrnen and Kan K&xBld (ed ) NOKlA

What to do?

All these problems could certainly have been resolved to some extent, but

An obvious. question was put: why should the SPU adapt to FPA instead of adapting FPA to current concepts of the SPU's SW specification and design processes, and SW architecture?

7 NOKIA Q NOKlA COCOMO Forum, 9- 1 1 th October 1996 / 5 / Petri Vestennen and Kan KansBla (ed.)

Pros and cons of FPA cuztomization

easy answers to our problems with FPA more concrete function types, i.e. more accurate estimates

9 not available early 9 benchmarking impossible

Selected function types

Lots of potential items were studied ->

Eight function types was selected master processes hand processes other programs messages

@ fields interface routines

@ MML commands MML command parameters

Customization decisions

System boundary at the (embedded SW) process level

Data from a large number of SW projects collected for weight calibration

originally 13 SW projects now dozens of projects

Definitions and glossary using existing terminology

no need for extensive training

9 NOKIA

Categorization of function

No complexity determination comple,xity already involved in many of the customized function types

Instead of that, classification based on the originality of corresponding SW blocks

new, modified, andlor deleted; each class with own weights

w i

Productivity factors

6 key productivity factors, have clear impact to the effort

architectural complexity communicational complexity

number of tasks number of persons number of working groups

a m . .

Specific to the SPU

Economy-of-scale vs. diseconom y-of-scale

Effort I

I

I I Diseconomy-of-scale I I because of increased

communication

Diseconomy-of-scale because of 'get-started' overhead

'fv NOKIA

Early results are very positive

Customized method is used extensively Method has been taken into use in all SW new projects at the SPU A couple of other SPUs is using the customized method There are hundreds of users now

Accuracy seems to be improving substantially

Early experiences suggest ca. 40% improvement in effort estimation accuracy No statistical data yet for full comparisons

?j NOKIA