10 JANUARY2021|COMPUTER SHOPPER|ISSUE
UNDER DEVELOPMENT
Questionable methods
I
’ve done acoupleof
consultancy jobs recentlythat
have broughtmeinto contact
with softwaredevelopersfrom
other organisations.Sharing
experiences offers someinteresting
insightsand confirmsthatI’mfar
fromaloneinsuffering customers
with ideas thatcould, shall wesay,
benefitfrom further work.
ButonethingIdidn’t detectthe
others havingtocope with isclients
who wantyou toproduce software
thatlies.We’re nottalkingsome
kind ofEnronfraud conspiracy,
but somecustomersthink they
haveavalid reason for having the
programshide,modifyorotherwise
be economical with the truth.
Atypicalcaseis onewherethe
managementwant tostore afalse
product costin astock-control
system.The problem they’re trying
to solvegoeslike this:
- The staff arepaid bycommission
based on profit.
- The salesorder programshows
the salespeople the salesvalue,
the costofthesales andthe
gross profit.
- Sales staff can negotiatethe
pricewith thecustomer.
TELL ME LIES
Management think the salespeople
give awaytoo muchprofit toclose
thesale,despitethefact that to
do sowould hit the salesperson’s
pocketin lost commission. So they
wantthe program to lieaboutthe
cost so that there appearstobeless
profitto‘give away’.
It wouldn’t be so bad if all you
had to do was store the false cost
in the stock item record, but it
doesn’t stop there. First, there’s the
questionofaccesstostock records.
If the user is asalesperson, youhave
to ensure the program hides all
traces of the realcost, whereas
management andaccounts users
need to see both. Then there are
the stock reports whereyou have to
do the same kind of messing about.
And the sales commission report
and payroll calculations have to use
the false costs, not the real ones.
The implications of thissimple lie
ripple on throughout the program,
making it anightmare to maintain.
QUESTIONTIME
Even betteristhe customer who
wantsto use oursoftwareto lie to
himself. Iwas dubious aboutthis
projectright from the start and
did mybest to persuade the client,
Syd, todothe job with astandard
package, buthe wasn’t havingany
of it.He wanted acustomer survey
program that asks questionssuch as
“did the salesperson approach you
in afriendlymanner?” or “please
rate our store appearance out of10,
with 10 being excellent”.
Syd sendsquestionnairesfull
of such interrogatory stuffshortly
aftereach customer makesa
purchaseand wants aprogram to
integrate with his customer records
file to store the results. Additionally,
he needs to print the results in
graphic formfor discussion at his
monthly management meeting.
He had aprogram that somebody
wrote but the authorleft, then the
program broke and nobody knows
how to fix it. Actually, discussions
about the program specification
suggestthat bits of the program
never worked in the firstplace, but
Syd’s not admitting that.
Idid aquick Google search to
see what’savailableoff theshelf.
There are hundredsofprograms to
choose from. As farasIcan see,
every conceivable angle has been
coveredbyone program or another.
The cost seems to vary from £
upto almost anypriceyou careto
imagine,so Isuggested that helook
at thespecs andsavehimselfalot
ofangstand expense. Buthe
insistedonacustomsolution.
We constructedaprogramfor
himthat stores andanalysesthe
questionnaire results. Syd’s
company has lots ofdifferent
branches andhe wants toreward
those that best satisfy their
customers. Soeach questionis
allocated ascore,usuallybetween
1and10, butnotalways. Some
questions are self-scoring,such as
the“rate ourstoreappearance out
of10” type,while othersare binary,
asin “would you shop with us
again? Yes/No”.
Syd canneverkeep things
simple, so somequestions have a
“Yes/No/Maybe” response, some
have four possibilitiesand somecan
generatenegative scores as well as
positiveones.
CHEATS NEVER PROSPER
We createdthe program and, fora
couple of months, it churned out
questionnaires andthe cust2mers
(customer’s customers) sent them
back. Theresults were analysed and
graphs generated from the data.
The graphs aresomething else
again. Ihavealways believed that if
abusiness graphic hasn’t hammered
homethe messageintwo seconds
flat, it’s failed. By that measure, the
DavidRobinsondespairsofsomeofhisclients’ideas.Andtomakemattersworse,
he’s alsostartedtoquestionthemotivebehindsomeoftheseterribleschemes
DavidRobinson
Softwareandsystems
developer
graphsfromthissystem (toSyd’s
design) failutterly. You could
stare atthem for hours andstill
be working out whatthey mean.
They’re like aJackson Pollock
painting, only busier. Oneof the
best shows 12 months’ worth of
scores onaquestion (both pluses
andminuses) foreach ofthe 24
branches witheachmonthin
adifferent colour using 3D
histograms that add even more
colours forthe barbits thatarein
shadow.There’s somuchdetail
that youneedamagnifying glassto
distinguishone bar from another.
Buteven withall that colour,
Syd isn’thappy. “Thescores are
wrong,”hesays.So we checked
the arithmeticandthey lookedOK.
Whathe’s unhappyabout isthe
Don’tKnows,as in thepeople who
can’teven answerasimpleYes/No
type question orscoresomething
between oneand 10.
Othercust2mersmessthings
up by ticking bothYes andNo. And
Maybe. Or they writeashort essay
where there should be atick to
makeachoice between options
but don’t actually choose anything
at all. If there’s away of getting
somethingwrong, people will
always find it.
We’ve rigged theprogram
so that if acust2mer answers a
questioninanambiguous way or
one that can’t be turned intoavalid
score, we discard the answer.
According to Syd, that’s not the
thing to do. What should happen,
he insists, is thatifthe program
encounters an invalid answer that
can’tbescored, thescore should be
adjusted to show the maximum
score that could be awarded.
Some questionnairesare so
badly filled in that, using this
method, they award branch
brownie points by the lorry load. If
the sales staff ever getwind of how
this works, they’llbebribingthe
customers just to mess up the
forms. It allgoes to show what a
tangled web we weave.CS
ILL
US
TR
AT
IO
N:
KE
VI
N
FE
BR
UA
RY
Originally published inShopper 225
