You are on page 1of 107

Defect Management &

Metrics
Society for Software Quality
April 24, 2010
2010 Holl
Consulting
021209.01 SSQ Presentation
!o "s #!is $uy%
&
30 years in software
and hardware/software
Quality
&
19 years in Quality &
Process Management
&
6 years as Director of
Software Best Practices
and Corporate Software
Metrics Leader at Intuit
&
Past Q President
Q
U
A
LI
Y
T
Q S
S
2010 Holl
Consulting
021209.01 SSQ Presentation
'ri(ary )ele*ant +,perience
&
-esigne. "ntuit/s corporate .efect
(anage(ent process
&
Assiste. in rollout to 01100 engineers
&
$ui.e. its e*olution for si, years
&
As Corporate Software 2etrics 3ea.er,
fostere. 4uantitati*e pro5ect (anage(ent
an. process i(pro*e(ent
&
2a.e (ista6es an. learne. fro( t!e(
2010 Holl
Consulting
021209.01 SSQ Presentation
Agen.a
&
Su((ary of 're*ious #al6
&
-efect 2anage(ent 2etrics
&
2etrics for 'ro5ect 2anage(ent
&
2etrics for 'rocess "(pro*e(ent
&
A.*ance. 2etric 2o.els
2010 Holl
Consulting
021209.01 SSQ Presentation
3ast ti(e7
&
e re*iewe. cultural issues relate. to
.efect (anage(ent
&
e went o*er a .efect (anage(ent
process8

'olicies

or6flow9States

:iel.s
2010 Holl
Consulting
021209.01 SSQ Presentation
-ata Collection
&
At .ifferent states in t!e wor6flow, .ifferent
.ata are collecte.
&
2ost .ata are collecte. at su;(it ti(e
&
A..itional .ata are entere. ;y t!e
.e*eloper upon resolution
&
An. finally ;y t!e tester upon *erification
2010 Holl
Consulting
021209.01 SSQ Presentation
Agen.a

Su((ary of 're*ious #al6


&
-efect 2anage(ent 2etrics
&
2etrics for 'ro5ect 2anage(ent
&
2etrics for 'rocess "(pro*e(ent
&
A.*ance. 2etric 2o.els
2010 Holl
Consulting
021209.01 SSQ Presentation
!en -oes A -efect Count%
It de!ends on your goals
&
Defect Management8 #o trac6 .efects for resolution in
or.er to release pro.ucts

<ust log functional pro;le(s foun. in test= .e*iations fro( t!e


re4uire(ents t!at t!e custo(er will see= .efects t!at you .on/t
want to fall t!roug! t!e crac6s.
&
Defect "nalysis# #o i(pro*e t!e process or pre.ict

"f your goal is to get ;etter >learn? or pre.ict t!e future, you nee.
to capture (uc! (ore infor(ation.
$ithout early lifecycle data %re&uirements' etc() !redicti*e
models aren+t that relia,le
$orld-class com!anies don+t .ust manage defects / they analy0e information(
2010 Holl
Consulting
021209.01 SSQ Presentation
-ata Collection Costs
& 1stimate "ssum!tions8
11,000 .efects
@ t!e .efects are correcte. >9,000?
-ata collection ti(e8 10 (inutes >*ali.ate. at H'?
A*erage .efect cost8 1A !ours >.ata fro( 2ic!ael :agan?
-ata collection re.uces .efects BC
#i(e to recor. >116 D10(? E,000 !rs
#i(e to correct >96 D1A!? 144,000
!rs
-efect re.uction >96 D
BC?
4B0 .efects
#i(e sa*ings >4B0 D 1A!? F,200 !rs
23I 4506
Sa(pling is *ali., ;ut
sa(pling !as to ;e
(anage. to get goo.
ran.o( sa(ples an. to
ensure t!at a.e4uate
>statistically significant?
sa(ples are ;eing
collecte..
#!is .oesn/t 4uantify t!e *alue of ;eing a;le to pre.ict 4uality an. sc!e.ules
2010 Holl
Consulting
021209.01 SSQ Presentation
Si(ple 2easures
&
-efect cost G

&
:i, rate G
&
-uration G
3ast two inclu.e 4ueue ti(e, so t!ey .epen. on wor6loa.
"t/s !elpful to un.erstan. t!e stan.ar. .e*iation too
Honus8 Hy type, root cause, in5ection p!ase
D +,cel Ietor6-ays is a useful function
J >#otal )esolution +ffort?
Iu(;er of resol*e. .efects
Iu(;er of resol*e. .efects
'erio. of ti(e >wee6, (ont!, etc.?
J>>Close. -ate? >Su;(it -ate??D
Iu(;er of resol*e. .efects
2010 Holl
Consulting
021209.01 SSQ Presentation
)elati*e -efect Costs
&
"f you recor. in5ection an. .isco*ery
p!ases an. effort, you can co(pute
relati*e costs
&
+,a(ple8

Cost of a re4uire(ents .efect G 1

Cost of a .esign .efect G 10

Cost of a co.ing .efect G 100

Cost of a pro.uction .efect G 1000


2010 Holl
Consulting
021209.01 SSQ Presentation
Si(ilar to '!ase Contain(ent
&
Start wit! (atri, of in5ection an. .isco*ery
&
+nter t!e a*erage rewor6 >or .uration?
costs in eac! cell >resol*e. .efects only?
In.ection Phase
Disco*ery Phase )e4Ks -esign Co.e #est 'ro.uction
)e4uire(ents 2.4
-esign 1B.E 4.1
Co.e 22.F 9.B 1.2
#esting 4E.1 11.B 4.E 1.2
'ro.uction 1B.E E1.B 11.E 2.2 1
2010 Holl
Consulting
021209.01 SSQ Presentation
Si(ilar to '!ase Contain(ent
&
Ior(aliLe ;y .i*i.ing eac! cell ;y t!e
"n5ectionG-isco*ery *alue
In.ection Phase
Disco*ery Phase )e4Ks -esign Co.e #est 'ro.uction
)e4uire(ents 1.00
-esign A.E1 1.00
Co.e 9.4A 2.E2 1.00
#esting 11.2B 4.B1 E.B1 1.00
'ro.uction EB.B4 9.E9 1B.2B 1.1E 1.00
2010 Holl
Consulting
021209.01 SSQ Presentation
3ogging -efects
&
#!is is not
How many defects leak?
&
#!is is
What percent of known defects are recorded?
&
"f you only start logging .efects in test, you/re
(issing o*er !alf of your .ata
&
#ry to capture .ata fro( re*iews9inspections an.
.e*eloper unit tests
2010 Holl
Consulting
021209.01 SSQ Presentation
'rioritiLing -efects
&
#!is is not a;out !ow to set priority, ;ut7
&
'riority (eans t!e (ost i(portant .efects get
resources first
&
#o (easure t!is, we nee. to 6now t!e 4ueue
ti(e >fro( Msu;(itN to Mwor6 startsN?
&
#!erefore, we nee. an MAccepte.N state, or
so(et!ing to note w!en wor6 ;egan
&
!y can/t we use u,mit to 7erify .uration%
2010 Holl
Consulting
021209.01 SSQ Presentation
Aging -efects
&
#i(e in )esol*er 4ueue
&
#i(e in Oerifier 4ueue
&
Alternately, use
.urations (inus
effort as a pro,y
"*erage Queue 8ime
0
2
4
A
1
10
12
0P#est
Stopper
1 2 E 4 B
Priority
D
a
y
s

i
n

Q
u
e
u
e
2010 Holl
Consulting
021209.01 SSQ Presentation
-efect Hac6log
&
2any groups .on/t count .efects left o*er fro(
prior releases= t!ey only focus on .efects foun.
an. not fi,e. in t!e current pro5ect
&
#!is is .angerous ;ecause a pro.uct .efect
;ac6log can grow un(anage.
&
ALL of t!e open .efects are release., not 5ust
t!e new ones
&
!en pro.ucing t!e report of open .efects,
4uery for all open .efects in t!e product
2010 Holl
Consulting
021209.01 SSQ Presentation
#!e Calculation
$ee9
1
2
E
4
8otal num,er of defects
/ :um,er closed ,efore the wee9 started
/ :um,er that were o!ened after the wee9 ended
; :um,er o!en in the wee9
2010 Holl
Consulting
021209.01 SSQ Presentation
Calculating Hac6log in +,cel
&
Query Su;(it -ate an. Close. -ate
2010 Holl
Consulting
021209.01 SSQ Presentation
Calculating Hac6log in +,cel
8his uses named ranges' where
-.Iu(.C)s G #otal nu(;er of C)s opene.
-.#.Close. G )ange of Close. -ates
-.Su;(it.-ate G )ange of Su;(it -ates
Array for(ula. QS!iftPCtrlP)eturnR
2010 Holl
Consulting
021209.01 SSQ Presentation
#!e )esult
Product Defect <ac9log
0
200
400
A00
100
1000
1200
1
9
F
9
2
0
0
A
2
9
F
9
2
0
0
A
E
9
F
9
2
0
0
A
4
9
F
9
2
0
0
A
B
9
F
9
2
0
0
A
A
9
F
9
2
0
0
A
F
9
F
9
2
0
0
A
1
9
F
9
2
0
0
A
9
9
F
9
2
0
0
A
1
0
9
F
9
2
0
0
A
1
1
9
F
9
2
0
0
A
1
2
9
F
9
2
0
0
A
1
9
F
9
2
0
0
F
2
9
F
9
2
0
0
F
E
9
F
9
2
0
0
F
4
9
F
9
2
0
0
F
B
9
F
9
2
0
0
F
A
9
F
9
2
0
0
F
F
9
F
9
2
0
0
F
1
9
F
9
2
0
0
F
9
9
F
9
2
0
0
F
1
0
9
F
9
2
0
0
F
1
1
9
F
9
2
0
0
F
1
2
9
F
9
2
0
0
F
1
9
F
9
2
0
0
1
2
9
F
9
2
0
0
1
E
9
F
9
2
0
0
1
4
9
F
9
2
0
0
1
=

3
!
e
n

D
e
f
e
c
t
s
2010 Holl
Consulting
021209.01 SSQ Presentation
Agen.a

Su((ary of 're*ious #al6

-efect 2anage(ent 2etrics


&
2etrics for 'ro5ect 2anage(ent
&
2etrics for 'rocess "(pro*e(ent
&
A.*ance. 2etric 2o.els
2010 Holl
Consulting
021209.01 SSQ Presentation
So(e 'ro5ect 2anage(ent -efect 2etrics
&
)e*iews S "nspections
&
-efect -isco*ery S Closure
&
-efect -ensity
2010 Holl
Consulting
021209.01 SSQ Presentation
A 'ro5ect A,io(
&
#!e 4uality of pro.uct in one
p!ase is an in.icator of t!e
4uality in later p!ases
&
!y%

#!ere are two answers


2010 Holl
Consulting
021209.01 SSQ Presentation
)e*iew 2etrics
&
)e*iews S inspections !a*e
pro*en to ;e t!e (ost costP
effecti*e way to .etect an.
re(o*e .efects
&
"nspect your 6ey pro5ect artifacts
>re4uire(ents, .esigns, co.e, tests, etc.?
2010 Holl
Consulting
021209.01 SSQ Presentation
)e*iew9"nspection -ata
&
#!ere are (any .ata collecte. .uring a
re*iew or inspection
&
#!e ones t!at tell us a;out 4uality are8

Iu(;er of 3!erational .efects .isco*ere.

Iu(;er of Minor .efects foun.


per c!ange. line of inspecte. .oc9co.e
&
Iu(;er of re*iews per .oc9(o.ule
2010 Holl
Consulting
021209.01 SSQ Presentation
'ro5ect -efect -ensity
&
Consistency in categoriLing inspection
.efects is i(portant

Ha*e clear gui.elines for Tperational an.


2inor .efects
& -efect -ensity
'ro5ect
G
Tperational -efects foun. ;y inspection
-elta S3TC >or 3T#?
2010 Holl
Consulting
021209.01 SSQ Presentation
'ro5ecting -efect 3ea6age
&
"nspecte. total of BE40 S3TC
&
:oun. 4F operational .efects
&
-ensity G 4F9BE40 G 1.1 .efects9US3TC
&
'ro5ect c!ange. 11,4AE S3TC
&
+sti(ate. .efects G 11.4AE D 1.1 G 1A2
&
)e(aining G 1A2 P 4F G 11B
!at s!oul. you .o% "nspect (ore% Tr #est for t!e(%
2010 Holl
Consulting
021209.01 SSQ Presentation
Oerification
&
Vse -efect -ensity to esti(ate e,pecte.
yiel.

Si(ilar to esti(ate .escri;e. wit! re*iews


&
-efect .isco*ery S closure
&
Vse e,trapolation to pre.ict w!en .efect
e,it criteria will ;e (et
2010 Holl
Consulting
021209.01 SSQ Presentation
-efect -isco*ery
#i(e
C
u
(
u
l
a
t
i
*
e

-
e
f
e
c
t
s

:
o
u
n
.
8otal Defects 3!ened
"
.
e
a
l

C
u
r
*
e
W
$ee9ends
2010 Holl
Consulting
021209.01 SSQ Presentation
-efect -isco*ery Cur*e Assu(ptions
&
#i(e is a pro,y for Me,posureN

)e(o*e wee6en.s to s(oot! t!e cur*e


&
Iu(;er of test cases run woul. ;e ;etter,
assu(ing

#est co*erage is goo.

Iot continuing to a.. new test case


2010 Holl
Consulting
021209.01 SSQ Presentation
ellPHe!a*e. Cur*e
#i(e
C
u
(
u
l
a
t
i
*
e

-
e
f
e
c
t
s

:
o
u
n
.
8otal Defects 3!ened
)elease #arget -ate
2010 Holl
Consulting
021209.01 SSQ Presentation
A..8 -efect Closure
#i(e
C
u
(
u
l
a
t
i
*
e

-
e
f
e
c
t
s
8otal Defects 3!ened & >losed
Tpene.
Close.
)e(aining
Tpen
2010 Holl
Consulting
021209.01 SSQ Presentation
A..8 #o QA
#i(e
C
u
(
u
l
a
t
i
*
e

-
e
f
e
c
t
s
8otal Defects 3!ened' >losed' & 8o Q"
Tpene.
#o QA
Close.
Tpene., waiting for
.e*elop(ent fi,
#urne. o*er fro(
.e*elop(ent to QA,
awaiting *erification
2010 Holl
Consulting
021209.01 SSQ Presentation
:in.ing #!e Halance
#i(e
C
u
(
u
l
a
t
i
*
e

-
e
f
e
c
t
s
8otal Defects 3!ened' >losed' & 8o Q"
Tpene.
#o QA
Close.
$hat+s
ha!!ening
here?
QA is a
;ottlenec6
-e* is a
;ottlenec6
2010 Holl
Consulting
021209.01 SSQ Presentation
A..8 3egacy S #arget )elease
#i(e
C
u
(
u
l
a
t
i
*
e

-
e
f
e
c
t
s
8otal Defects 3!ened & >losed
#otal
#argete.
#o QA
Close.
"n!erite. fro(
pre*ious !istory,
targete. for or foun.
in t!e pro5ect9release
2010 Holl
Consulting
021209.01 SSQ Presentation
+,trapolation to )elease -ate
#i(e
C
u
(
u
l
a
t
i
*
e

-
e
f
e
c
t
s
Pro.ect Defects
#otal
#argete.
#o QA
Close.
)elease #arget -ate
S!oul. " ;e
concerne.%
2010 Holl
Consulting
021209.01 SSQ Presentation
+,trapolation to )elease -ate
#i(e
C
u
(
u
l
a
t
i
*
e

-
e
f
e
c
t
s
Pro.ect Defects
#otal
#argete.
#o QA
Close.
)elease #arget -ate
!en will t!e
pro5ect ;e .one%
2010 Holl
Consulting
021209.01 SSQ Presentation
Co(puting in +,cel
Query .efects into +,cel
Con*enient8 Vse na(e. ranges for .ata
2010 Holl
Consulting
021209.01 SSQ Presentation
Co(puting in +,cel
&
"n a secon. s!eet, create a .ate colu(n an. count
t!e .efects su;(itte. an. close. for eac! .ate.
Array for(ula. QS!iftPCtrlP)eturnR
2010 Holl
Consulting
021209.01 SSQ Presentation
)eal 3ife +,a(ple
"ard*ar9 5(@ Defect Disco*ery & >losure
>it! e,trapolations?
0
B0
100
1B0
200
2B0
E00
EB0
400
4B0
B00
A
9
1
F
9
0
1
A
9
2
4
9
0
1
F
9
1
9
0
1
F
9
1
9
0
1
F
9
1
B
9
0
1
F
9
2
2
9
0
1
F
9
2
9
9
0
1
1
9
B
9
0
1
1
9
1
2
9
0
1
1
9
1
9
9
0
1
1
9
2
A
9
0
1
8
o
t
a
l

>
2
s
0
10
20
E0
40
B0
A0
F0
10
90
100
A
o
u
n
d

i
n

5
(
@
J Tpen
J Close.
J :oun. in 4.F
A91A G Start of testing
#arget
)elease
-ate
2
n.
or.er polyno(ial fit
2010 Holl
Consulting
021209.01 SSQ Presentation
)elease Criteria8 -efects
&
$oals8

'ro*i.e goo.Penoug! 4uality to custo(ers


&
-on/t release any M;a.N .efects
&
-on/t release Mtoo (anyN .efects

2ini(iLe support costs


&
Custo(ers will nee. support proportional to
pro.uct 4uality
2010 Holl
Consulting
021209.01 SSQ Presentation
-efect -isco*ery
#i(e
C
u
(
u
l
a
t
i
*
e

-
e
f
e
c
t
s

:
o
u
n
.
8otal Defects 3!ened
)elease #arget -ate
S!oul. " release%
2010 Holl
Consulting
021209.01 SSQ Presentation
Ha. -efects
&
Ha. is a function of

e*erity G t!e conse4uences of t!e failure

Pro,a,ility t!at it will occur

#!e num,er of customers li6ely to e,perience it

2ay also inclu.e Mi(portantN custo(ers or functions


&
#!ese are often co(ple, to calculate, so (ost
co(panies 5ust use Se*erity, te(pere. ;y ot!er
factors
2010 Holl
Consulting
021209.01 SSQ Presentation
Quality #!res!ol.
&
Io (ore t!an t!e following nu(;er of open
.efects8

!ere Iew G .efects .isco*ere. in t!e current pro5ect, an.

Tl. G .efects .isco*ere. in pre*ious releases


e*erityB :ew 3ld
Critical 0 0
"(portant B 10
2o.erate E0 100
#ri*ial B0 200
Release team (stakeholders) need to
determine values based on history. The
goals are:
& Not to release any Critical defects
& Not to introduce too many ne
defects
& !anage the backlog
2010 Holl
Consulting
021209.01 SSQ Presentation
!at/s $oo. +noug!%
The two market leaders Quicken
TurboTax, produced by Intuit, and
TaxCut, produced by H&R Block
It also oers better end!
o!return re"iews #
TurboTax oers better "ideos than
TaxCut #
2010 Holl
Consulting
021209.01 SSQ Presentation
!at/s $oo. +noug!%
TaxCut oers easier ta$$in$ o
items you%d like to deal with later&
It also appears to be more
scrupulously prooread than
TurboTax' Twice in TurboTax%s
screens in my tests, incorrect
inormation was displayed, thou$h
the tax returns themsel"es were
unaected&
2010 Holl
Consulting
021209.01 SSQ Presentation
!at/s $oo. +noug!%
# i you%re choosin$ which pro$ram
to $o with or the irst time this year,
I%d $o with TaxCut solely because
o the text errors in TurboTax&
I ha"e a hi$her standard or accuracy in tax
pro$rams than I do or any other type o
sotware, and while the errors we ound didn%t
aect the bottom line, they did aect my
conidence in the pro$ram%s makers&
It%s unortunate, because I elt
TurboTax oered better eatures&
2010 Holl
Consulting
021209.01 SSQ Presentation
-etroit :ree 'ress
R 1.F2 )ea.ers
X we; site
X wire ser*ice
#!e article ca(e fro( t!e
Monterey Peninsula Herald
2010 Holl
Consulting
021209.01 SSQ Presentation
're.icting -efects
&
#!ere are se*eral (o.els for pre.icting
.efects
&
#!e pre.ictions of interest are8

How (any .efects are going to ;e create.%

!en will " fin. t!e(%

How long will it ta6e to fi, t!e(%

!en will " ;e .one%

How goo. will t!e result ;e%


2010 Holl
Consulting
021209.01 SSQ Presentation
Vsing -efect -ensity
&
'!ase -efect -ensity
--
'!ase
G
&
At first t!is (ay not see( rig!t, ;ut it/s a reasona;le
pre.ictor
&
!ic! is (ore relia;le%

AF specification .efects per 200 pages of .ocu(entation


>0.E4 .efects9page?
AF specification .efects per 4E00 S3TC >1B.A .efects9US3TC?
-efects "n5ecte.
'!ase
#otal 'ro5ect Affecte. S3TC
2010 Holl
Consulting
021209.01 SSQ Presentation
Vsing -efect -ensity
&
'ro5ect9'rocess -efect -ensity

Iu(;er of .efects per


affecte. S3TC

Affecte. S3TC can ;e


.ifficult to get wit!out a
goo. tool.

#!e ;est "/*e foun. is


Co.e)eports
>a6a Source)T"?

!ttp899www.polariguana.co(9sourceroi.p!p
2elease " 2elease <
Defects
4B2 BAE
8otal C3>
1,A00 4E,B00
Defect
Density
B2.A 12.9
Pro.ect
>hanged
C3>
B,400 4,EB0
Pro.ect
Defect
Density
1E.F 129.4
2010 Holl
Consulting
021209.01 SSQ Presentation
Calculating -- per '!ase
Y -efects "n5ecte. 'er '!ase
'ro5ect US3TC
Iu(;er of -efects In.ection Phase
Disco*ery Phase )e4Ks -esign Co.e #est 'ro.uction
)e4uire(ents 4A 0 0 0 0
-esign 1 2A 0 0 0
Co.e 1E 19 1E 0 0
#esting EE 22 EF 2 0
'ro.uction E 2 1A 0 1
8otal In.ected 1FE A9 1EA 2 1
Defect Density 5(95 1(9@ 3(D9 0(06 0(03
'ro5ect
S3TC G
EB,000
Iote8 if t!is is
only test .efect
.ata, you/re
(issing an
opportunity to get
;etter
infor(ation.
--
'!ase
G
2010 Holl
Consulting
021209.01 SSQ Presentation
Ie,t 'ro5ect
&
+sti(ate. to ;e 20,000 S3TC
Eistorical
Defect
Density
Ci9ely
to
In.ect
Phase
>ontainment
Ci9ely
to Cea9
1ffort to
2esol*eB
8otal
2ewor9
2e&uirements 4.94 99 2A.AC 2A 1B.E 2242.2
Design 1.9F E9 1F.EC F E1.B 2AE.E
>ode E.19 F1 A2.1C 41 11.E 112.9
8esting 0.0A 1 11.FC 1 2.2 2.1
Production 0.0E 1 100.0C 1 1 0.A
#otal 211 Frand 8otal 3391(0
D orst case cost >.efects .isco*ere. in test?
2010 Holl
Consulting
021209.01 SSQ Presentation
-efect -isco*ery S Closure
&
're.ictions can ;e
(a.e in +,cel ;y
e,trapolating an.
seeing w!ere t!e
.isco*ery an.
close lines (eet
>umulati*e Defects u,mitted
0
B0000
100000
1B0000
200000
2B0000
E00000
1
0
9
1
4
9
2
0
0
A
1
0
9
2
1
9
2
0
0
A
1
1
9
1
1
9
2
0
0
A
1
1
9
2
B
9
2
0
0
A
1
2
9
9
9
2
0
0
A
1
2
9
2
E
9
2
0
0
A
1
9
A
9
2
0
0
F
1
9
2
0
9
2
0
0
F
2
9
E
9
2
0
0
F
2
9
1
F
9
2
0
0
F
E
9
E
9
2
0
0
F
E
9
1
F
9
2
0
0
F
E
9
E
1
9
2
0
0
F
4
9
1
4
9
2
0
0
F
4
9
2
1
9
2
0
0
F
B
9
1
2
9
2
0
0
F
B
9
2
A
9
2
0
0
F
A
9
9
9
2
0
0
F
A
9
2
E
9
2
0
0
F
F
9
F
9
2
0
0
F
F
9
2
1
9
2
0
0
F
1
9
4
9
2
0
0
F
2010 Holl
Consulting
021209.01 SSQ Presentation
+,trapolation
&
#!is (et!o. is
4uic6 an. in
so(e cases
surprisingly
goo.
&
Hut not in all
cases
2010 Holl
Consulting
021209.01 SSQ Presentation
Tt!er +,cel +,trapolations
&
A..ing a tren. line uses all t!e .ata
&
Zou (ay only want to use some of t!e
.ata
&
Create a parallel colu(n of t!e .ata you
want to e,trapolate an. e,trapolate only
t!at cur*e
2010 Holl
Consulting
021209.01 SSQ Presentation
+,a(ple
2010 Holl
Consulting
021209.01 SSQ Presentation
Si(ple 3inear +,trapolation
&
#o create a linear e,trapolation on t!e last
few .ata points, start ;y e,ten.ing t!e
.ate
colu(n8
2010 Holl
Consulting
021209.01 SSQ Presentation
Si(ple 3inear +,trapolation
&
Select t!e cells t!at
loo6 linear
&
Copy an. 'aste
Special P Oalues
2010 Holl
Consulting
021209.01 SSQ Presentation
Si(ple 3inear +,trapolation
&
#!en .rag t!e cells .own to t!e last
e,trapolate. .ate
&
+,cel will
.o a linear
e,trapolation
of t!e .ata
2010 Holl
Consulting
021209.01 SSQ Presentation
+,trapolation
&
"n or.er for e,trapolations to ;e relia;le,
t!e .isco*ery rate (ust !a*e pea6e.
$ee9ly Defects u,mitted
0
10
20
E0
40
B0
A0
1
9
1
9
2
0
0
1
1
9
1
B
9
2
0
0
1
1
9
2
9
9
2
0
0
1
2
9
1
2
9
2
0
0
1
2
9
2
A
9
2
0
0
1
E
9
1
1
9
2
0
0
1
E
9
2
B
9
2
0
0
1
4
9
1
9
2
0
0
1
4
9
2
2
9
2
0
0
1
B
9
A
9
2
0
0
1
B
9
2
0
9
2
0
0
1
A
9
E
9
2
0
0
1
A
9
1
F
9
2
0
0
1
F
9
1
9
2
0
0
1
F
9
1
B
9
2
0
0
1
F
9
2
9
9
2
0
0
1
1
9
1
2
9
2
0
0
1
1
9
2
A
9
2
0
0
1
9
9
9
9
2
0
0
1
9
9
2
E
9
2
0
0
1
1
0
9
F
9
2
0
0
1
2010 Holl
Consulting
021209.01 SSQ Presentation
-one%
&
!at .oes it (ean to ;e .one >fro( a
.efect perspecti*e?%

Io M;a.N .efects re(ain

're.icta;le9(anagea;le .efects re(ain


&
#wo categories of .efects re(ain

#!ose you 6now

#!ose you .on/t 6now


2010 Holl
Consulting
021209.01 SSQ Presentation
3atent -efects
&
3atent .efects >un6nown at release ti(e?
can ;e pre.icte. using .efect .ensity an.
)ayleig! '-:
&
!at nee.s to !appen ;efore t!e
pre.ictions ;eco(e relia;le%

#!e .efect .isco*ery rate (ust flatten


d f(t)
dt
G 0
2010 Holl
Consulting
021209.01 SSQ Presentation
-isco*ery )ate
>umulati*e Defect u,mitted
0
100
200
E00
400
B00
A00
1
9
1
9
2
0
0
1
1
9
1
B
9
2
0
0
1
1
9
2
9
9
2
0
0
1
2
9
1
2
9
2
0
0
1
2
9
2
A
9
2
0
0
1
E
9
1
1
9
2
0
0
1
E
9
2
B
9
2
0
0
1
4
9
1
9
2
0
0
1
4
9
2
2
9
2
0
0
1
B
9
A
9
2
0
0
1
B
9
2
0
9
2
0
0
1
A
9
E
9
2
0
0
1
A
9
1
F
9
2
0
0
1
F
9
1
9
2
0
0
1
F
9
1
B
9
2
0
0
1
F
9
2
9
9
2
0
0
1
1
9
1
2
9
2
0
0
1
1
9
2
A
9
2
0
0
1
9
9
9
9
2
0
0
1
9
9
2
E
9
2
0
0
1
1
0
9
F
9
2
0
0
1
$ee9ly Defects u,mitted
0
10
20
E0
40
B0
A0
1
9
1
9
2
0
0
1
1
9
1
B
9
2
0
0
1
1
9
2
9
9
2
0
0
1
2
9
1
2
9
2
0
0
1
2
9
2
A
9
2
0
0
1
E
9
1
1
9
2
0
0
1
E
9
2
B
9
2
0
0
1
4
9
1
9
2
0
0
1
4
9
2
2
9
2
0
0
1
B
9
A
9
2
0
0
1
B
9
2
0
9
2
0
0
1
A
9
E
9
2
0
0
1
A
9
1
F
9
2
0
0
1
F
9
1
9
2
0
0
1
F
9
1
B
9
2
0
0
1
F
9
2
9
9
2
0
0
1
1
9
1
2
9
2
0
0
1
1
9
2
A
9
2
0
0
1
9
9
9
9
2
0
0
1
9
9
2
E
9
2
0
0
1
1
0
9
F
9
2
0
0
1
2010 Holl
Consulting
021209.01 SSQ Presentation
:itting -ata to a 2o.el
&
:itting .ata
to a cur*e
!elps us tell
t!e rate is
.ecreasing
2010 Holl
Consulting
021209.01 SSQ Presentation
Vsing +,cel
&
As ;efore, create a linear e,trapolation of
t!e last QcriteriaR .ays8
&
M:latteningN criteria e,a(ple8

#!e .isco*er rate (ust less t!an B .efects


per .ay in t!e past two wee6s of testing
2010 Holl
Consulting
021209.01 SSQ Presentation
Vsing +,cel
&
A.. linear tren. line
&
Select M-isplay e4uation on c!artN
2010 Holl
Consulting
021209.01 SSQ Presentation
Vsing +,cel
&
#!is gi*es t!e slope of t!e line (,
w!ere

yG(, X ;

"n t!is case,


( G 2.2 .efects9.ay

Iot goo. enoug!


>umulati*e Defect u,mitted
y G 2.1AE, P 1BE9B
0
100
200
E00
400
B00
A00
1
9
1
9
2
0
0
1
1
9
1
B
9
2
0
0
1
1
9
2
9
9
2
0
0
1
2
9
1
2
9
2
0
0
1
2
9
2
A
9
2
0
0
1
E
9
1
1
9
2
0
0
1
E
9
2
B
9
2
0
0
1
4
9
1
9
2
0
0
1
4
9
2
2
9
2
0
0
1
B
9
A
9
2
0
0
1
B
9
2
0
9
2
0
0
1
A
9
E
9
2
0
0
1
A
9
1
F
9
2
0
0
1
F
9
1
9
2
0
0
1
F
9
1
B
9
2
0
0
1
F
9
2
9
9
2
0
0
1
1
9
1
2
9
2
0
0
1
1
9
2
A
9
2
0
0
1
9
9
9
9
2
0
0
1
9
9
2
E
9
2
0
0
1
1
0
9
F
9
2
0
0
1
2010 Holl
Consulting
021209.01 SSQ Presentation
3atent or Unown -efect )is6
&
'ro;a;ility of occurrence in pro.uction

)is6 G Se*erity D 'ro;a;ility


&
Iee. to !a*e a .ecent (easure9esti(ate
of pro;a;ility

Iee. a usage profile


&
Also, fre4uency .oes not always e4ual
*alue

" (ay not use t!e M)estoreN function *ery


often, ;ut it ;etter wor6
2010 Holl
Consulting
021209.01 SSQ Presentation
An. :inally7
&
#rac6
latent9release.
.efects
carefully
&
#!ey are your
;est in.icator of
Quality
2010 Holl
Consulting
021209.01 SSQ Presentation
)elease -efect -ensity
:ro(8
oftware Quality
Measurement#
" Aramewor9 for >ounting
Pro,lems and Defects
C2V9S+"P92P#)P022
!ttp899www.sei.c(u.e.u9pu;9.ocu(ents992.reports9p.f9tr22.92.p.f
2010 Holl
Consulting
021209.01 SSQ Presentation
Agen.a

Su((ary of 're*ious #al6

-efect 2anage(ent 2etrics

2etrics for 'ro5ect 2anage(ent


&
2etrics for 'rocess "(pro*e(ent
&
A.*ance. 2etric 2o.els
2010 Holl
Consulting
021209.01 SSQ Presentation
A #rue Story
&
A w!ile ;ac6, " loo6e. in t!e (irror an.
.eci.e. " nee.e. to loose so(e weig!t.
&
!at is t!e first t!ing "/( going to .o%
2010 Holl
Consulting
021209.01 SSQ Presentation
2y :irst Step
2010 Holl
Consulting
021209.01 SSQ Presentation
2y :irst Step Tn A Scale
&
"f " .on/t (easure (yself to.ay, " won/t
6now

"f t!e c!anges "/( planning to (a6e are ta6ing


(e t!e rig!t .irection

!ic! c!anges !a*e t!e ;est affect

How far "/*e co(e

!en "/( .one


&
Its important to estalish a aseline
2010 Holl
Consulting
021209.01 SSQ Presentation
!at Are
#!e $oals%
!Would you tell me"
please" which way I
ou#ht to #o from here?$
!%hat depends a #ood deal on where you want to #et
to"$ said the &at'
!I don$t much care where(($ said )lice'
!%hen it doesn$t matter which way you #o"$ said the &at'
2010 Holl
Consulting
021209.01 SSQ Presentation
Closure -ata Collection
Su;(it reason
& #est case faile.
& 'ass criteria unclear
& #este. wrong *ersion
& Setup9config wrong
& -ataPrelate. error
& Tt!er
Io
& Correction (a.e
& -uplicate
& C!ange not 5ustifie.
"n5ection '!ase
)oot Cause
-efect%
)esolution
Zes
Classification, 2o.e
Correcte. Co(ponent>s?
8his is where we record what we+*e learned
#otal
)esolution
+ffort
)
e
s
o
l*
e
.
2010 Holl
Consulting
021209.01 SSQ Presentation
-efect Classification 2o.els
&
"n.ustry gui.elines to stan.ar.iLe
infor(ation collection
&
#!e (ost pre*alent are fro(8

"+++

"H2

Hewlett 'ac6ar.
2010 Holl
Consulting
021209.01 SSQ Presentation
H' -efect Trigins, #ypes, an. 2o.es
2010 Holl
Consulting
021209.01 SSQ Presentation
-efect 3ea6age )easons
1. +n*iron(ental -epen.ency
2. #est Iot )un
E. #est Case +rror
4. -ata -epen.ency
B. "nconsistent Sy(pto(s
A. #est Case Success Criteria Vnclear
F. #est +,ecution +rror
1. #est Case )esults Iot Oerifie.
9. Co.e C!ange. After #est 'asse.
10. Io #est Case9Co*erage
11. Configuration 2anage(ent 'ro;le(
12. )e4uire(ents 2issing9C!ange.
2010 Holl
Consulting
021209.01 SSQ Presentation
Vsing 3ea6age -ata
&
:or eac! .efect foun. in pro.uction, log a lea6age
reason
&
Vse t!e fre4uency >*olu(e? to i(pro*e t!e relate.
process
&
)esist te(ptation to weig!t t!e fre4uency ;y
so(et!ing i(portant

#!is assu(es cause S effect, an. t!ere is none

#!is is .ifferent fro( root cause, w!ere t!e cost to repair


is relate. to t!e type of .efect
2010 Holl
Consulting
021209.01 SSQ Presentation
Specific 'rocess "(pro*e(ent 2etrics
&
#wo types of process (etrics8
1. 2etrics t!at (onitor a process
>tell you t!e effect of c!anges?
& Hasic S'CPtype c!arts
2. 2etrics t!at tell you w!at to .o
& '!ase Contain(ent
&
)oot Cause Analysis
& #est Co*erage
& -efect 3ea6age )easons
& Static Co.e Analysis P Co(ple,ity
2010 Holl
Consulting
021209.01 SSQ Presentation
S!oul. all pro5ects ;e (easure.%
&
'ros8

2ore .ata points

+*aluate e*eryt!ing we/re .oing


&
Cons8

'ro5ects (ay use .ifferent processes

Oariation (ay ;e too great


&
Suggestions8

-on/t (easure pro*ects (easure processes

Tnly (easure your ;est pro5ects

#ailoring is o6ay, ;ut separate out t!e .ata


2010 Holl
Consulting
021209.01 SSQ Presentation
Hol. 'erio.ic 'rocess )e*iews
&
2ont!ly Quality )e*iews focuse. on

'rocess "(pro*e(ent initiati*es

2easures S Analysis9"nterpretation

3essons 3earne.
&
)ules of (etrics eti4uette8

're*iew .ata wit! sta6e!ol.ers G no surprises


2010 Holl
Consulting
021209.01 SSQ Presentation
'!ase Contain(ent
&
".entifies t!e lifecycle acti*ities t!at lea6
.efects
&
#ells you w!ere
to i(pro*e .efect
.etection efforts
Phase >ontainment
0C
10C
20C
E0C
40C
B0C
A0C
F0C
10C
90C
100C
)e4uire(ents -esign Co.e Huil. #esting
6

D
e
f
e
c
t
s

>
o
n
t
a
i
n
e
d
2010 Holl
Consulting
021209.01 SSQ Presentation
)oot Cause Analysis
&
'areto t!e types of
errors ;eing create.,
weig!te. ;y cost >effort?
2oot >ause 8otal Eours
0
B00
1000
1B00
2000
2B00
E000
)
e
4
u
i
r
e
(
e
n
t
s

A
(
;
i
g
u
o
u
s
9
*
a
g
u
e

)
e
4
u
i
r
e
(
e
n
t
s

2
i
s
s
i
n
g

3
o
g
i
c

f
l
o
w

2
i
s
s
i
n
g

l
o
g
i
c

)
e
4
u
i
r
e
(
e
n
t
s

C
!
a
n
g
e
.

r
o
n
g

*
a
r
i
a
;
l
e

u
s
e
.

T
t
!
e
r

C
a
s
e

s
t
a
t
e
(
e
n
t

#
y
p
o

S
p
e
c
i
f
i
c
a
t
i
o
n

2
i
s
s
i
n
g
$
a
r
;
a
g
e

c
o
l
l
e
c
t
i
o
n

+
,
c
e
p
t
i
o
n

!
a
n
.
l
i
n
g

'
a
r
a
(
e
t
e
r

(
i
s
(
a
t
c
!

)
e
4
u
i
r
e
(
e
n
t
s

r
o
n
g

V
n
i
n
i
t
i
a
l
i
L
e
.

*
a
r
i
a
;
l
e

H
a
.

*
a
l
i
.
a
t
i
o
n

c
!
e
c
6

2
i
s
s
i
n
g

e
n
.

s
t
a
t
e
(
e
n
t

+
n
*

n
o
t

(
a
t
c
!

t
e
s
t
S
p
e
c
i
f
i
c
a
t
i
o
n
P
r
e
4
u
i
r
e
(
e
n
t

(
i
s
(
a
t
c
!
C
o
.
i
n
g

s
t
a
n
.
a
r
.

n
o
t

f
o
l
l
o
w
e
.

S
p
e
c
i
f
i
c
a
t
i
o
n

A
(
;
i
g
u
o
u
s
+
n
*

c
a
p
a
c
i
t
y
S
p
e
c
i
f
i
c
a
t
i
o
n

r
o
n
g

r
o
n
g

(
e
t
!
o
.

c
a
l
l
e
.

-
e
s
i
g
n

l
o
g
i
c

w
r
o
n
g
-
e
s
i
g
n
P
s
p
e
c

(
i
s
(
a
t
c
!
)
e
4
u
i
r
e
(
e
n
t
s

"
n
c
o
n
s
i
s
t
e
n
t

+
n
*

!
a
r
.
w
a
r
e

f
a
i
l
u
r
e
-
o
c

t
y
p
o
-
o
c

*
a
g
u
e
+
n
*

T
p
e
r
a
t
i
o
n
a
l

+
r
r
o
r
"
n
s
t
a
l
l

(
i
s
s
i
n
g

f
i
l
e
s
)
e
4
u
i
r
e
(
e
n
t
s

T
t
!
e
r

-
o
c

i
n
f
o

(
i
s
s
i
n
g
"
n
s
t
a
l
l

w
r
o
n
g

c
o
n
f
i
g
u
r
a
t
i
o
n
"
n
s
t
a
l
l

i
n
c
o
(
p
l
e
t
e
-
o
c

w
r
o
n
g
"
n
s
t
a
l
l

w
r
o
n
g

*
e
r
s
i
o
n
)oot Cause , )ewor6 Hours
2010 Holl
Consulting
021209.01 SSQ Presentation
Agen.a

Su((ary of 're*ious #al6

-efect 2anage(ent 2etrics

2etrics for 'ro5ect 2anage(ent

2etrics for 'rocess "(pro*e(ent


&
A.*ance. 2etric 2o.els
2010 Holl
Consulting
021209.01 SSQ Presentation
-efect 'ro;a;ilities
&
ei;ull .istri;utions are useful engineering
(o.els, w!ere t!e tail approac!es Lero
asy(ptotically
&
ei;ull 'ro;a;ility -ensity :unction >'-:? is8
!ere m .eter(ines t!e s!ape of t!e cur*e, an.
c is t!e scale para(eter
t
c
m
t
m
e
t
c
m
(
f(t) +
t
c
m
c
m(,
e
t
c
m
(
+
>i6ipe.ia?
2010 Holl
Consulting
021209.01 SSQ Presentation
-ifferent ei;ull -istri;utions
$ei,ull >ur*es
0
1
2
E
4
0 1 2 E
(G0.B
(G1
(G2 (G4
(G10
ei;ull cur*es wit!
.ifferent s!apes
>*alues of (? can
;e use. to (o.el
*arious syste(s
>pro;a;ility
.istri;utions?
2010 Holl
Consulting
021209.01 SSQ Presentation
ei;ull 'ro;a;ility -istri;ution
$ei,ull >ur*es
( G 0.1 to B.0 in .1 incre(ents
0.0
0.B
1.0
1.B
2.0
0.0 0.B 1.0 1.B 2.0
2ayleigh %m;4)
2010 Holl
Consulting
021209.01 SSQ Presentation
ei;ull Cur*es
$ei,ull >ur*es
( G 0.1 to B.0 in .1 incre(ents
0.0
0.B
1.0
1.B
2.0
0.0 0.B 1.0 1.B 2.0
2ayleigh %m;4)
%almost) :ormal %m;3(5)
2010 Holl
Consulting
021209.01 SSQ Presentation
)ayleig! 'ro;a;ility -istri;ution :unction
&
ei;ull .istri;ution wit! s!ape >(? G 2
&
3et - G 2t
t
c
2
t
2
e
t
c
.
(
f(t) +
2t
c
2
e
t
c
2
(
+
-
c
2
e
-
2c
.
(
f(-) +
-
c
.
e
-
2
4c
2
(
+
2010 Holl
Consulting
021209.01 SSQ Presentation
)ayleig! 'ro;a;ility -istri;ution :unction
-
c
2
e
-
2
4c
2
(
f(-) +
2ayleigh >ur*es
c G 0.1 to 0.B
0
1
2
E
4
B
A
F
1
9
0.0 0.B 1.0 1.B
c G 0.1
c G 0.2
c G 0.E
2010 Holl
Consulting
021209.01 SSQ Presentation
+,cel :or(ula
2010 Holl
Consulting
021209.01 SSQ Presentation
Applying )ayleig! '-:
& 3et t
m
G ti(e at w!ic! t!e cur*e reac!es its pea6
&
#o fin. t!e pea6, we set G 0
& Sol*ing gi*es t
m
G
&
#o scale t!e area un.er t!e cur*e, (ultiply ;y K
& Su;stitute c = t
m
2
d f(t)
dt
c
2
2010 Holl
Consulting
021209.01 SSQ Presentation
:itting #!e )ayleig! '-:
2t
c
2
e
t
c
2
(
f(t) + K
2t
(t
m
2?
2
e
t
>t
m
2?
.
(
+ K

2t
2t
2
m
e
( t
2
2t
2
m
+ K

1
t
m
te
( t
2
2t
2
m
f(t) + K
2
"f we can esti(ate !ow (any .efects we/ll fin. >K? an. w!en
we/*e !it t!e pea6 >t
m
?, we can create a '-: for our pro5ectW
2010 Holl
Consulting
021209.01 SSQ Presentation
Huil.ing Zour +sti(ate
2010 Holl
Consulting
021209.01 SSQ Presentation
're.icting 3atent -efects
&
At pea6, t!e area un.er t!e )ayleig! cur*e is E9.EBC
&
#!is (eans A0.ABC of t!e .efects are yet to fin.
&
+,a(ple8
At pea6 you foun. a cu(ulati*e total of B1 .efects.
#otal G >B19E9.B?D100G14F.4
14F P B1 G 19 re(aining.
3atent -efects + K dt
1
t
m
te
( t
2
2t
2
m
2
)elease -ate
1 year%
2010 Holl
Consulting
021209.01 SSQ Presentation
+asier t!at Calculus
&
"n +,cel, run a colu(n
for cu(ulati*e total an. a
colu(n for re(aining
.efects
&
:ro( your release .ate
you can easily esti(ate
latent .efects
2010 Holl
Consulting
021209.01 SSQ Presentation
Cu(ulati*e )ayleig! 2o.el
t
c
2
(
C-: G 1 P e
2010 Holl
Consulting
021209.01 SSQ Presentation
)ayleig! Assu(ptions
&
#!e .efect rate o;ser*e. .uring t!e
.e*elop(ent process is positi*ely
correlate. wit! t!e .efect rate in t!e fiel..
&
$i*en t!e sa(e error in5ection rate, if
(ore .efects are .isco*ere. an. re(o*e.
earlier, fewer will re(ain in later stages.
2010 Holl
Consulting
021209.01 SSQ Presentation
+,ponential 2o.els
&
+,ponential 2o.el >ei;ull, (G1?
,
c
e
t
c
(
f(t) +
1G!onential
ei;ull, ( G 1
0.0
0.2
0.4
0.A
0.1
1.0
1.2
0.0 0.B 1.0 1.B 2.0
& +rror .etection rate
& "nstantaneous failure rate
& HaLar. rate
& "n general, t!is is a
reliaility #rowth model
2010 Holl
Consulting
021209.01 SSQ Presentation
:itting A 2o.el to -ata
&
+asy:it fro( www.(at!wa*e.co(
2010 Holl
Consulting
021209.01 SSQ Presentation
2o.el Su((ary
&
)ayleig! is ei;ull .istri;ution wit! (G2
&
Tt!er *alues of ( pro.uce ot!er
pro;a;ility .istri;utions
&
#!ere are (any (o.els= try se*eral t!at fit
&
/)ll models are wron#' Some models are
useful'0

$eorge +. Ho,
2010 Holl
Consulting
021209.01 SSQ Presentation
Su((ary
&
A .efect (anage(ent pro#ram
inclu.es

A ;alance. solution

#raining S e,plaining

2etrics t!at are use.


Questions
?
C!ris Holl
C!ris[Holl\!ot(ail.co(
:eGt >lass# Tools for Metrics

You might also like