SlideShare a Scribd company logo
Int. J. Nonprofi.t Volunt. Sect.Mark. 1O:43-52 (2005)
Publishedonline in Wiley InterScience
(www.interscience.wiley.com).DOI: 10.I 002/nvsm.5
Pledger modelling: Help the
Aged case study
Karen Cole,l Rachel Dinglel and Raiesh Bhayani2
t
Talking Numbers, UK
t
Help tbe Aged., uK
o Therecruitment ofpledgers (as aproryfor potential legators) to cbaritable organisations
plays a uital role in tbeir continued success,and as apercentage of allfundraising income
generated it can represent substantial proportions. Hou)euer, of all the 'donation asks'
made of supporters, askingfor a legacy is tbe most dfficult. Tberefore, it is important that
tbe target audience sbould be as utell researcbed and bigltly targeted aspossible.
o Help tbe Aged bad reacbed tbe stage ubere decisions need to be made about itsfuture
marketing in order to protect longer-term income. Tbefindings of tbis legacy targeting
project utere tofeed into communication progralnmes, direct marketing, and tbe ouerall
legacy mar keting strategJ/.
o The key objectiue u)as to identifu tlJe best prospects to mail a legaqt ask to, across tbe
supporter database, utitlt tlte likeliltood tbat tbey are going to pledge as a result.
o It uas found tbat ultilst tailored data analtsis comes at a price, tbe auerage ualue of a
legacy justifies the cost of using sophisticated targeting tools. Houteuer, because of tbe
pledge-tr>l.egaqt time lapse, tbere utill ahaays be issues uitb measuring any long-term
return on inuestment (ROI). Nonetbeless,pledgers baue to be taken on tbeir utordfor tbe
purpose of testing (and subsequent rollouts). Pledge data sbould be tested and tbe
outcomes sltould inform legaqt marketing. Houteuer, as mentioned aboue, pledgers
necessarily need tobe taken ontbeir utord and tberefore,formulatingmodekbased on tbe
type and/or ualue of pledges is not recommended.
Copyrigbt A 2OO5Jobn lYiley & Sons, Ltd.
Background.
The recruitment of pledgers (as a proxy for
potential legators) to charitable organizations
plays avital role in their continued success,and
asa percentage of all voluntary income raised,
legaciescanrepresent asubstantialproportion.
However, of all the 'donation asks' made of
supporters, asking for a legacy is the most
difficult. According to Richard Radcliffe (flill,
Correspondence to: Nigel Magson, Talking Numbers,
Iarrcure House,MarketPlace, CirencesterGL7 2NW, UK.
E-mail:nmagson@talkingnumbers.com
2OO2)'legacy income is likely to go down as
women in their 80s who are leaving legacies
now tend to be asset ignorant. . .the next
generation will be more aware of the value of
their estate and may only leave a fixed sum to
charity'.
It is important that the target audience
should be as well researched and highly
targeted as possible. According to the I-egacy
Promotion Campaign website (2OOZ)'65% of
tbe population are regular cbarity suppor-
ters. Recent researcb bas sltoutn tbat donors
baue no objection to leauing cbaritable
legacies- tbey simply neuerget around to it.'
Copyright O 2O05John Viley & Sons,Ltd. Int.J. NonprofitVolunt. Sect.Mark., February 2005
44 K. Cole et al.
25,000,000
20,000,000
15,000,000
10,000,000
5,000,000
Figure l. kgacy income by financial year received.
Ifelp tbe Aged's legacy actiuity
Help the Aged's legacyincome was S,12.8mil-
lion in zOOL/Oz,(see Figure 1.)which repre-
sents a significant part of its total voluntary
income of 3,75million. Legacymarketing activ-
ities include four direct marketing campaigns
per annum. In addition, Help the Aged offers a
'Will Advice Service'which, although indepen-
dent, doesoften leadto apledge for the charify.
The Help the Aged pledger commitment
progression is asfollows:
Enquirers .---+ Considerers .---= Pledgers
c=:> Legators
Objectiue
Help the Aged had reached the stage where
decisions needed to be made about its future
marketing in order to protect longer-term
income. The findings of this project were to
feed into communication pfogfammes, direct
marketing and the overall legacy marketing
strategy. The key proiect objective was to
identify the bestprospects to mail a 'legacyask'
to, across the supporter database, with the
likelihood that they are going to pledge as a
result. Secondaryobjectives wefe asfollows:
o Descriptive statistics of legator and pledger
data
o Centralizing data
o Comparative sociodemographic profiles of
supporters through to legators
I
I I I
I I I I
1992/93 1993/941994/951995/961996/971997/98 1998/99 1999/00 2000tO12001tO22002103
Financialyear
o Timing of 'legacy ask' and other recommen-
dations for implementation.
Metbodol.ogjt and rfltionale
Analysis steps
As with any investigative project, thefe afe a
number of analysissteps to go thfough before
any sort of modelling can be achieved. These
main proiect steps are asfollows:
r Describe data- summary statistics
o Profile data
r Reporting
o Modelling
o Reporting
The first step in this case involved under-
standing the transition - and differences -
between the stagesof legator and pledger com-
mitment by producing summary statisticson all
availabledata. This also allowed for an evalua-
tion of the existing strategy by examining any
changes shown by the sufirmary statistics bet-
ween groups or over time.
Following this analysisthe next stagewas to
profile the available data. Profiling is a techni-
que that compares gfoups of interest to some
sort of base population, for example, the life
stagesof legators compared to the GB popula-
tion as a whole. Although profiling obviously
provides insight in its own right, it is an
essential precursof to modelling because it
finds the variables that provide the most
Copyrighto 2005JohnWiley & Sons,Ltd. Int. J. Nonprofit Volunt. Sect.Mark., February 2OO5
Pledger modelling 45
discrimination befween groups. Talking Num-
bersprofiled by common databasevariables,as
well as by some of the external data variables
described later.
Modelling
The final stages of analysis revolved around
modelling the data in order to predict future
behaviour. For this project we had a large
databasebut with very few existing pledgers,
making the modelling more difficult. The
potential of modelling legators was examined
(seeCTIAIDsectionbelow) but the final models
were built for pledges. A pledge is the desired
fesponse and, as such, is a behaviour the
models should predict. The dataon pledgesdid
not allow the models to be tailored to the type
of pledge (pecuniary or residuary) or the value
of pledge.
It should be noted that even a small gain in
direct mail should be cost-effective since ave-
rage legacies(depending on type) are up to an
averageof S25,000 (residuary).
The project deliverables were two scores
across the supporter database, one for cash
donors (including committed givers who have
alsogiven acashgift) and one for all committed
givers. The scores were based on models indi-
cating propensity to become a pledger. Sup-
pofters can then be targeted appropriately
based on variables alreadv known about them
on the database.
Talking Numbers used a combination of
modelling techniques, namely CHAID and logi-
stic regression,to produce the models. Ifithin
production of a regressionmodel many models
were tested (the dataset size allowed a test
sample)in orderto pickthe most discriminatory
model to be written back to the supporter base.
Talking Numbers also allowed for an inter-
active meeting between analystsand Help the
Aged in-between the profiling and modelling
stages.This allowed the preliminary results to
be discussed and any extfa, and important,
insight to be provided by the parties actually
involved in collecting and using the data. This
meeting can be an invaluable contribution to
an analysisproiect.
CHAID Method
Chi-square Automatic Interaction Detection
(CHAID) is a form of cluster analysis.This is
where the data set is initially thought of asone
large cluster and then broken up into anumber
of clusters containing related data, aftet exam-
ining all interactions. CFIAID can be used to
find the most significant discriminating factors
for a given dichotomous variable e.g. 1/O or
yes/no. CHAID works best with dichotomous
variables.
Uses
CHAID is usedextensively in marketing. One of
the usesof CHAID isto discoverwhichvariables
are best used to describe respondents/non-
respondents.Thus segmentationmodels can be
built up for communication purposes.
Output example
Figure 2 shows atree diagram,or dendogrlm,
produced using CIIAID. In Figure 2 R is the
response fate. It can be seenthat the response
rate for AB males is over 80%, whereas the
response rate of DE females is only 17%.
Consequently, AB males are deemed a better
prospect than DE females.
Logistic regression
Regressiontechniques are a classof statistical
methods in which one dependent variable -
the response- is related to one or more
independent variables- the predictive or
explanatory factors. A regression model is
one in which the response variable is linearly
related to each explanatoryvariable, i.e. where
there is astraight-line relationship between the
responseandeachexplanatory variablethere is
linear regression. For example, children's
weight is often related to their height where
weight increases as height increases,this is a
straightline relationship and linear regression.
Simplelinear regressioniswhere there is only a
singleexplanatory variable and multiple regres-
sion is where there is more than one explana-
tory variable. Logistic regression is applied asa
technique when the response variable is of a
Copyright ar2O05John Viley & Sons,Ltd. Int.J. Nonprof.t Volunt.Sect.Mark., February 2005
46 K. Cole et al.
Figure 2. CHAID tree response analysis.
binary type, that is, yes,/no,hasresponded/has
not responded etc.
Available data
Data were provided by Help the Aged from
their donor and legacy databases.Many differ-
ent fields were extracted in order to allow
maximum interrogation: cash and committed
giving including Adopt a Granny database
donors, legators, pledgers, considerers and
enquirers.
Iri addition, Talking Numbers investigated
external datato allow cold modelling and add
value to the existing donor information.
MONICA (CACI)
MONICA is CACI's age classification based on
sophisticated analysisof millions of first names
taken from lifestyle questionnaires. Monica
helps to identiry the ageof people on adatabase
by looking at the likely age profile of their first
name, it alsoallows new donors to be targeted
by matching a distinct ageprof,le.
PRIZM (Acxiom)
The PRIZM postcode lifestyle-basedsegmenta-
tion and targeting tool assignsGB consumers
into one of 6O unique clusters. The data has
been aggregatedup to full postcode level. lts
core data sources include:
o Lifestyle census - information derived from
the merger of three companies: Claritas,
NDL and CMT, that includes over 40O
variables, including income, age,household
composition and lifestyle
o Company directors - supplied by Dun and
Bradstreet
o Share ownership - a list of public and
pivatized shareholders from NDL's 'Active
lnvestor File' database
o Behaviourbank - CMT's database contain-
ing details of financial products, shopping
habits, holiday preferences, appliance own-
ership and property types
o Unemployment rates- from the Depart-
ment of Employment
o Birth and death rates - from the Office of
National Statistics
o Flat, farm and house names- from the
Postcode AddressFile @AFl
o Census data-from the Office of National
Statistics.
Every GB postcode was assigned a five-char-
acter code basedon the key marketing drivers
of life stageand income. The first and second
characters indicate the following life stages.
PA: Young no children (starting out, young
singles and childless couples)
PB: Families (nursery and school-age
children)
PC: Empty nesters (and older singles)
PD: Retired seniors.
Each of the four life stageswas then broken
down into four income indicators:
Copyrighto 2005JohnViley & Sons,Ltd. Int.J. Nonprofit Volunt.Sect.Mark., February 20O5
Pledger rnodelling 4 7
1. Most affluent households
2. Mid-high affluent households
3. Mid-low affluent households
4. Leastaffluent households.
Then a unique cluster code dffierentiated
between households that had the same life
stageand income band in tems of lifestyle, for
example PA311 : Young Conservationists:
o Young married
r Own four-wheel drive
o Likely to have club cards
o Support national causes
o Support animal welfare charities.
StreetValue (CACI)
StreetValueprovides Lnavetagevalue for every
postcode, the number of properties and pre-
dominant type of property pertaining to a
postcode area. StreetValue also shows if the
properties in a particular areaare mainly own-
occupied or rented. These data combined
eventually allowed two models to be
produced:
o Cashdatabasedonors (active and lapsed)
o Committed giving database donors (active
and lapsed).
Find.ings
Summary statistics
The data exploration stage highlighted the
issueswith small ovedaps acrossthe donor and
legator databases.The different types ofdonor,
howevef, were still compared. Also discovered
at this stage was the short time span of the
enquirer to legator cycle. This meant that
there were few donors found in the mid-cycle
gfoups, such as considerers. Also, complete
life cycles were available for only a few
individuals. However, each piece of the
enquirer, considerer, pledger and legator
cycle was examined in its own right and
sufirmary information produced. For example,
Figure 3 shows this informationforthe legator
cycle.
Profiles
Legators and pledgers have a very similar life
stageprofile to each other. They are older than
the other donor groups. Other cashsupporters
(those who have not made anylegacy contact)
show a younger age profile, similar to con-
siderersand enquirers, while committed givers
are different again and areby far the youngest
group. The affluence profiles are different,
however. The legator profile most closely
matches that of the other cash supporters and
committed givers (those who have not made
any legacy contact). In terms of affluence,
pledgers are similar to considerers and
enquirers comprise the least affluent group.
Pledgers are more likely to be aged 50-59
(MONICA) and retired seniors GRIZND. The
results from the StreetValue analysis were
interesting but did not show enough differ-
entiation to use in a model,
Financial year
Figure 3. Number of legators by financial year of notification date
1992/931993/94 1994/95 1995/96 1996/97 1997/98 1998/99 '1999/0020001012001t02 2002t03
Copyrightto2005JohnWiley & Sons,Ltd. Int.J. Nonprofit Volunt.Sect.Mark., February 2OO5
48 K. CoIe et al.
Pledger profiles are interesting in terms of
their cash-giving history. Nearly half of all
pledgers (who are also cash supporters) are
in the top value band. The averagevalue of cash
gift is a very important factor in the pledger
cash model. Pledgers are the group with the
highest frequency profile and frequency of
cash gift also appears in the final pledger cash
model.
Geograpbical patterns
The geographical spread of the number of
legators was examined and this showed a
distinct Southern bias to the pattem, with
Scotland in particular providing very small
numbers of legators. A geographic analysisof
the averagelegacyvalueswas alsoundertaken.
This time, a pattern emerged whereby the
average legacy values rose in correlation to
the geographical affluence of each Lrea,
particularly in the South East.Additionally, a
numberof interesting hot spotswere identffied
including the 'retirement' areasof Torquay and
Brighton.
CHAID
After running tables and comparing profiles,
CHAID was used as an exploratory technique
to investigate the relative importance of
differentiating factors and how those factors
combine to provide the most differentiation
(and hence prediction). CHAID exploration
was carried out looking at prediction of both
pledgers and legators; rnd avalTablevariables
included cash-giving history, legacy contact
made and external data such as MONICA and
PRIZM.As there is aclear relationship between
the life rycle groups, i.e. legators are likely to
have been pledgers, etc., it was decided to
model pledgers and not a combination of
gfoups.
After the profiles were studied, it also
became apparent that two models would be
needed, one for cash donors and one for
committed givers. These two groups were
quite different in profile and it was expected
that they would behave very differently. There-
fore one model for both groups would have
been inappropriate.
Modelling: cash donors
The same modelling approach was used for
both cash donors and committed givers. The
profiling resultswere studied andvariablesthat
might become predicted factors were high-
lighted. Then both CHAID and regression
techniques were used to refine this list until
the best predictors in combination were
selected for the final models, and their respec-
tive levels of importance were allocated.
This methodology produced the following
factors in the final cash donor model:
o Other relationships/donor status/enquiries
made
o MONICA ageband
o Cash-givinghistory, value and frequency.
The 'gains' achieved by the model are evident
as half of all pledgers were found in the first
(righest scoring) 3O%of cash supporters, as
shown in Figure 4andTable 1.
Modelling: coflrmitted givers
The final model for committed givers included
the following factors:
o Length of relationship
o Other relationships
o PRIZM, both life stageand affluence
o MONICA.
Over half of all pledgers were found in the top
25%of committed giving supporters asshown
in Figure 5 andTable2.
Implementatiom
Writeback
Each score effectively ranked the indMduals
according to their propensity to pledge. These
two scores were appended to about 5OO,OOO
records on the donor database at Help the
Copyright O 2005 John rlfliley & Sons,Ltd. Int.l. Nonprofit Volunt.Sect.Mark., February 2005
Pledger modelling 49
60%
o
E 50o/o
E
40%
No/o
no/o
10%
0%
Figure 4. Cashpledger model gains chart.
Aged. Talking Numbers worked with Help
the Aged to plan mailing volumes using the
model - bearing in mind that certain donors
and donor groups would be excluded from
mailings due to suppressions or othef activity.
The aim for the next yearwas to test the models
within the direct marketing prografirme, com-
bined with selections based on edsting donor
segments.
Table 1. Cashpledger model
Spring 2003
The February 2OO3 legacy campaign was
mailed to 28,000 cash donors, split into eight
equal mailing segments according to standard
RFVcriteria andpledger propensity score.Two
random 3,500 samples, labelled Band 1 and
Band 2, were taken for each of the RFV
segments- Band 1 comprised those donors
NTILE* Pledgers Model (%) Random (%o) Model cumulative (%) Random cumulative (%)
I
2
3
4
5
6
7
8
9
10
l l
l 2
r3
r4
1 5
r6
t7
18
r9
20
Total
216
216
206
r93
159
101
r62
r49
r55
r45
1 1 0
77
7r
56
63
1 3
8
6
7
69
2,t82
9.90
9.90
9.44
8.85
7.29
4.63
7.42
6.83
7.ro
6.65
5.O4
3.53
1 2 q
2.57
2.49
0.60
o.37
o.27
o.32
3.16
100.oo
5.00
5.00
5.00
5.00
5.O0
5.00
5.00
5.OO
5.OO
5.00
5.00
5.O0
5.00
5.00
5.00
5.O0
5.00
5.00
5.OO
5.00
100.00
9.90
19.80
29.24
38.08
+>.t/
50.oo
57.42
64.25
71.36
78.00
83.O4
86.57
89.83
92.39
95.28
95.88
96.24
96.52
96.84
100.oo
5.00
to.o0
15.00
20.00
25.OO
30.00
35.00
40.oo
45.OO
50.00
55.00
60.00
65.OO
70.oo
75.OO
80.o0
85.00
90.00
95.00
100.00
.An NTILE is an equal sized group of ranked d^ta, in this case the data is split into 2Oequal sized groups.
Copyrighto 2O05JohnI(iley & Sons,Lrd. Int.J. Nonprofit Volunt.Sect.Mark., February 2005
50 K. Cole et al.
E oov"
8. mv"
o
p +ou
450/o 55%
Contact
Figure 5. Committed giving pledger model gains chart.
with a pledger score in the bottom 5O%of the
model, and Band 2 comprised those with a
scoreof the next bestz0%.lt hadbeen decided
previously that those with a score in the top
30% should not be tested in this the first
campaign, but 'saved' for later campaigns.
Summer 2OO3
To further test the effectivenessof the pledger
models for targeting potential legators, there
were two campaignsin Summer2003. Help the
Table2. Committed giving pledger model
Aged mailed donors with a high propensity to
pledge and those within the samedatabaseseg-
ments with a low propensity to pledge (5O/5O
split). Significant pledger results woe found for
both models ascan be seenin Tables 3 and4.
Moving forward
2OO3in summaty
The results for 2o03looked promising in terms
of the effectiveness of the pledger models. It
NTILE* Pledgers Model (%") Random (%) Model cumulative (%) Random cumulative (%) Index
I
2
J
4
)
6
7
8
9
l0
l l
t 2
7 3
14
1 5
r6
17
l8
r9
20
Total
103
62
)
27
38
37
2 l
22
r9
2a
r6
18
7
t t
l 2
r3
9
0
7
2
504
20.44
12.30
ro.32
5.36
/.>+
7.34
4.r7
4.37
) . / /
5.56
3.r7
5. >/
r.39
2.La
2.38
2.58
r.79
0.00
r.39
o.40
100.oo
5.OO
5.00
5.OO
5.OO
5.OO
5.OO
5.00
5.00
5.00
5.00
).UU
5.00
5.O0
5.O0
5.O0
5.O0
5.00
5.00
5.00
5.00
100.oo
20.44
32.74
43.06
48.41
55.95
63.29
67.46
7r.83
75.60
8 1 . 1 5
84.33
87.90
49.29
9r.47
93.85
96.43
98.2r
98.2r
99.60
100.00
5.OO
10.00
15.00
20.00
25.OO
30.00
35.00
40.oo
4r.oo
50.00
55.O0
60.00
65.O0
70.oo
75.OO
80.o0
85.00
90.00
95.00
l00.oo
4.o9
1 )'7
2.87
2.42
2.24
2 . t r
r.93
1.80
1.68
r.62
r.53
r.46
r.37
1.31
r.25
t.2 l
r . 1 6
1.09
1.05
1.00
*An NTILE is an equal sized group of ranked datt, rn this case the data is split into 20 equal sized groups.
Copyright o 2OO5John Wiley & Sons,Ltd. Int.t. Nonprofi.tVolunt.Sect.Mark., February 2005
Pledger rnodelling 51
Segment
Table 3. July 2003 (committed giving model) o Telemarketing channel as tested in 2OO3
may be included in the modelling.
There will be two models again as the
behaviourof cashdonors and committedgivers
are clearly different. The inclusion of high-
value donors and others excluded previously
may increase overall response rates. 2003's
new creative model will again be used, after
response rates and research showed its
effectiveness.
2004 actiuity
August 2OO4committed givers mailing
September 2OO4cash givers mailing (total
70,ooo)
September 2004 outbound follow-up call to
mailed committed givers.
Conclu.sions
The LegacyPromotion Campaign(2OO2)states:
'Over the past 12yearsthe percentage of those
wills going to probate containing a legacy to
charity has remained constant at arcund l3%.
This is despite all the resources that charities
have put into legacy marketing, indicating that
they only succeededin competing for ashareof
the same wills. There is evidence that whilst
the number of people making wills is increas-
ing, the proportion of those that include a
charitable legacy is in decline'.
This makes careful targeting through direct
communication even more critical. Targeting
will help ensure charities' legacy income is
maintained or even increased. Targeting
further ensufes cost-efficiencies since careful
selection means the most likely prospects are
communicated to with a proposition that
they're more likely to accept.
Thilst tailoreddataanalysis comesat aprice,
the averagevalue of alegacyiustifies the cost of
using sophisticated targeting tools. However,
because of the pledge-tolegacy time lapse,
there will always be issues with measuring
any long-terrn return on investment (ROD.
Nonetheless, pledgers have to be taken on
their word for the purpose of testing (and
Score Low High Low
No.mailed 7,5OO 7,5oO 3,500
Pledgerresponse 19 56 9
Responserate O.25% O.75% 0.26%
HiCh
2,444
) 4
0.94o/n
CG: committed givers segment.
AAG:'Adopt A Granny' committed givers make aS,l2 per
month (approximately) regular gift.
Table 4. August 2OO3(cash model)
Segment Active Active
a
O
Score
No. mailed
Pledger response
Responserate
Low
6,oo2
8
o.13%
should be noted that there were other factors
that would have influenced the results. For
example, Help the Aged tested a new creative
pack alongsidethe cashpledger model. As part
of ongoing analysis and testing of the models
there was a need to re-run the models on the
entire donor database,including new donors
and high value donors who were excluded
previously, to enablebettertargeting of donors
n2OO4. A full post-campaign meta-analysisfor
2OO3 has also been suggested so that the
number of responses across the groups could
be come mofe statistically robust. A meta-
analysisis a statistical way of combining more
than one set of test results together in order to
give the analyst latger numbers of responders
on which to basethe analvsis.
Neut model
o Z0D3post-campaign meta-analysis
o New strategy for 2OO4targeting, including
new channels (telemarketing)
o Examine the profiles and update the learning
irrz003 for new pledgers
HiCh
7,OO3
25
O.36o/o
High
7,OO3
26
o.37%
copyright o 2005Johnffiley & sons,Ltd. Int.l. Nonprofit Volunt.Sect.Mark., February 2005
52 K. Cole et al.
subsequent rollouts). Pledge data should be
tested and the outcomes should inform legacy
marketing. However, as mentioned above,
pledgers necessarilyneed to be taken on their
word and therefore, formulating models based
on the type and/or value of pledges is not
recommended.
References
Hill N. 2OO2. Q&A: legacies. Tbe Guardian,
October lL,2OO2.
LegacyPromotion Campaign Website. 2002. www.
legacypromotioncampaign.org.uk[May 2OO4].
copyright @2005 John l(iley & Sons,Ltd Int. I. Nonprortt Volunt. Sect.Mark., February 2OO5

More Related Content

Viewers also liked

Qepa m4 portafolio actividad integradora
Qepa m4 portafolio actividad integradoraQepa m4 portafolio actividad integradora
Qepa m4 portafolio actividad integradora
Quetzalli De Avila
 
Luxury yacht charter croatia
Luxury yacht charter croatiaLuxury yacht charter croatia
Luxury yacht charter croatia
luxuryyacht
 
La rete wireless presso l’Universitàdi Siena: Tecnologie e servizi
La rete wireless presso l’Universitàdi Siena: Tecnologie e serviziLa rete wireless presso l’Universitàdi Siena: Tecnologie e servizi
La rete wireless presso l’Universitàdi Siena: Tecnologie e servizi
GoWireless
 
Work at Play's Franchise Hub Model for Games
Work at Play's Franchise Hub Model for GamesWork at Play's Franchise Hub Model for Games
Work at Play's Franchise Hub Model for Games
David Gratton
 
Halls xs email
Halls xs emailHalls xs email
Halls xs email
Filipe Blanco
 
APM Solutions for all Eclipsys Applications by Tevron
APM Solutions for all Eclipsys Applications by TevronAPM Solutions for all Eclipsys Applications by Tevron
APM Solutions for all Eclipsys Applications by Tevron
marktevron
 
Клетка как чудо
Клетка как чудоКлетка как чудо
Клетка как чудо
Sergey Semenov
 
FMCG 2014 in Ukraine
FMCG 2014 in UkraineFMCG 2014 in Ukraine
Сотрудники
СотрудникиСотрудники
Сотрудники
Nickolai Smirnov
 
Mobile Time is Prime Time anytime!
Mobile Time is Prime Time anytime!Mobile Time is Prime Time anytime!
Mobile Time is Prime Time anytime!
ThinkDigital
 
201404 How to boost Bank Branches in a Multichannel World
201404 How to boost Bank Branches in a Multichannel World201404 How to boost Bank Branches in a Multichannel World
201404 How to boost Bank Branches in a Multichannel World
Francisco Calzado
 
Bankenworkshop zum Thema "Was besagt ein Bestätigungsvermerk/ eine Bescheinig...
Bankenworkshop zum Thema "Was besagt ein Bestätigungsvermerk/ eine Bescheinig...Bankenworkshop zum Thema "Was besagt ein Bestätigungsvermerk/ eine Bescheinig...
Bankenworkshop zum Thema "Was besagt ein Bestätigungsvermerk/ eine Bescheinig...
KANZLEI NICKERT
 
SAPA DE LETRAS DE GASES Y ÁCIDOS Y BASES
SAPA DE LETRAS DE GASES Y ÁCIDOS Y BASESSAPA DE LETRAS DE GASES Y ÁCIDOS Y BASES
SAPA DE LETRAS DE GASES Y ÁCIDOS Y BASES
cobaep26
 
Sistema venoso
Sistema venosoSistema venoso
Sistema venoso
Andrea Celemin
 
Cross/Transmedia Story Design
Cross/Transmedia Story DesignCross/Transmedia Story Design
Cross/Transmedia Story Design
Christy Dena
 

Viewers also liked (15)

Qepa m4 portafolio actividad integradora
Qepa m4 portafolio actividad integradoraQepa m4 portafolio actividad integradora
Qepa m4 portafolio actividad integradora
 
Luxury yacht charter croatia
Luxury yacht charter croatiaLuxury yacht charter croatia
Luxury yacht charter croatia
 
La rete wireless presso l’Universitàdi Siena: Tecnologie e servizi
La rete wireless presso l’Universitàdi Siena: Tecnologie e serviziLa rete wireless presso l’Universitàdi Siena: Tecnologie e servizi
La rete wireless presso l’Universitàdi Siena: Tecnologie e servizi
 
Work at Play's Franchise Hub Model for Games
Work at Play's Franchise Hub Model for GamesWork at Play's Franchise Hub Model for Games
Work at Play's Franchise Hub Model for Games
 
Halls xs email
Halls xs emailHalls xs email
Halls xs email
 
APM Solutions for all Eclipsys Applications by Tevron
APM Solutions for all Eclipsys Applications by TevronAPM Solutions for all Eclipsys Applications by Tevron
APM Solutions for all Eclipsys Applications by Tevron
 
Клетка как чудо
Клетка как чудоКлетка как чудо
Клетка как чудо
 
FMCG 2014 in Ukraine
FMCG 2014 in UkraineFMCG 2014 in Ukraine
FMCG 2014 in Ukraine
 
Сотрудники
СотрудникиСотрудники
Сотрудники
 
Mobile Time is Prime Time anytime!
Mobile Time is Prime Time anytime!Mobile Time is Prime Time anytime!
Mobile Time is Prime Time anytime!
 
201404 How to boost Bank Branches in a Multichannel World
201404 How to boost Bank Branches in a Multichannel World201404 How to boost Bank Branches in a Multichannel World
201404 How to boost Bank Branches in a Multichannel World
 
Bankenworkshop zum Thema "Was besagt ein Bestätigungsvermerk/ eine Bescheinig...
Bankenworkshop zum Thema "Was besagt ein Bestätigungsvermerk/ eine Bescheinig...Bankenworkshop zum Thema "Was besagt ein Bestätigungsvermerk/ eine Bescheinig...
Bankenworkshop zum Thema "Was besagt ein Bestätigungsvermerk/ eine Bescheinig...
 
SAPA DE LETRAS DE GASES Y ÁCIDOS Y BASES
SAPA DE LETRAS DE GASES Y ÁCIDOS Y BASESSAPA DE LETRAS DE GASES Y ÁCIDOS Y BASES
SAPA DE LETRAS DE GASES Y ÁCIDOS Y BASES
 
Sistema venoso
Sistema venosoSistema venoso
Sistema venoso
 
Cross/Transmedia Story Design
Cross/Transmedia Story DesignCross/Transmedia Story Design
Cross/Transmedia Story Design
 

Similar to Pledger Model Paper.PDF

Rewired.earth building a sustainable future - september 2021 (4) (1) (2)
Rewired.earth   building a sustainable future - september 2021 (4) (1) (2)Rewired.earth   building a sustainable future - september 2021 (4) (1) (2)
Rewired.earth building a sustainable future - september 2021 (4) (1) (2)
Chris Skinner
 
The total impossibility of customer experience management
The total impossibility of customer experience managementThe total impossibility of customer experience management
The total impossibility of customer experience management
Digital Clarity Group
 
The total impossibility of CEM
The total impossibility of CEMThe total impossibility of CEM
The total impossibility of CEM
Tim Walters, Ph.D.
 
The Total Impossibility of Customer Experience Management (CEM)
The Total Impossibility of Customer Experience Management (CEM)The Total Impossibility of Customer Experience Management (CEM)
The Total Impossibility of Customer Experience Management (CEM)
Digital Clarity Group
 
Business Risk Case Study Ba 32
Business Risk Case Study  Ba 32Business Risk Case Study  Ba 32
Business Risk Case Study Ba 32
Sandip Sen
 
Bank churn with Data Science
Bank churn with Data ScienceBank churn with Data Science
Bank churn with Data Science
Carolyn Knight
 
Insurance Churn Prediction Data Analysis Project
Insurance Churn Prediction Data Analysis ProjectInsurance Churn Prediction Data Analysis Project
Insurance Churn Prediction Data Analysis Project
Boston Institute of Analytics
 
IRJET - Online Donation based Crowdfunding using Clustering and K-Nearest Nei...
IRJET - Online Donation based Crowdfunding using Clustering and K-Nearest Nei...IRJET - Online Donation based Crowdfunding using Clustering and K-Nearest Nei...
IRJET - Online Donation based Crowdfunding using Clustering and K-Nearest Nei...
IRJET Journal
 
Deloitte UK / Facebook : a people based telecom business - breathing new lif...
Deloitte UK / Facebook  : a people based telecom business - breathing new lif...Deloitte UK / Facebook  : a people based telecom business - breathing new lif...
Deloitte UK / Facebook : a people based telecom business - breathing new lif...
yann le gigan
 
New segmentation strategies with Big Data
New segmentation strategies with Big DataNew segmentation strategies with Big Data
New segmentation strategies with Big Data
Amalist Client Services
 
Success in the management ofcrowdfunding projects in the.docx
Success in the management ofcrowdfunding projects in the.docxSuccess in the management ofcrowdfunding projects in the.docx
Success in the management ofcrowdfunding projects in the.docx
picklesvalery
 
Canback and D'Agnese - Where in the World Is the Market?
Canback and D'Agnese - Where in the World Is the Market?Canback and D'Agnese - Where in the World Is the Market?
Canback and D'Agnese - Where in the World Is the Market?
Tellusant, Inc.
 
how_the_digitalchannels_shape_the_future_of_shopping
how_the_digitalchannels_shape_the_future_of_shoppinghow_the_digitalchannels_shape_the_future_of_shopping
how_the_digitalchannels_shape_the_future_of_shopping
Karl Fredrik Lund
 
Retail Banking: In tech we trust
Retail Banking: In tech we trustRetail Banking: In tech we trust
Retail Banking: In tech we trust
The Economist Media Businesses
 
Reply to DiscussionsD1 navyaA bank failure is the ending of.docx
Reply to DiscussionsD1 navyaA bank failure is the ending of.docxReply to DiscussionsD1 navyaA bank failure is the ending of.docx
Reply to DiscussionsD1 navyaA bank failure is the ending of.docx
chris293
 
Time Series Analysis
Time Series AnalysisTime Series Analysis
Time Series Analysis
Amanda Reed
 
BCI Segment Intelligence
BCI Segment IntelligenceBCI Segment Intelligence
BCI Segment Intelligence
Steve Abbott
 
AR - Applying Big Data to Risk Management
AR - Applying Big Data to Risk ManagementAR - Applying Big Data to Risk Management
AR - Applying Big Data to Risk Management
Valentine Seivert
 
John Gutfranski, CFP, AIF, CRPC & Debra White Stephens, CFP – Proactive Advis...
John Gutfranski, CFP, AIF, CRPC & Debra White Stephens, CFP – Proactive Advis...John Gutfranski, CFP, AIF, CRPC & Debra White Stephens, CFP – Proactive Advis...
John Gutfranski, CFP, AIF, CRPC & Debra White Stephens, CFP – Proactive Advis...
Proactive Advisor Magazine
 
How Banks Can Close the 'Value Gap' and Regain Customer Trust
How Banks Can Close the 'Value Gap' and Regain Customer TrustHow Banks Can Close the 'Value Gap' and Regain Customer Trust
How Banks Can Close the 'Value Gap' and Regain Customer Trust
Joseph M Bradley
 

Similar to Pledger Model Paper.PDF (20)

Rewired.earth building a sustainable future - september 2021 (4) (1) (2)
Rewired.earth   building a sustainable future - september 2021 (4) (1) (2)Rewired.earth   building a sustainable future - september 2021 (4) (1) (2)
Rewired.earth building a sustainable future - september 2021 (4) (1) (2)
 
The total impossibility of customer experience management
The total impossibility of customer experience managementThe total impossibility of customer experience management
The total impossibility of customer experience management
 
The total impossibility of CEM
The total impossibility of CEMThe total impossibility of CEM
The total impossibility of CEM
 
The Total Impossibility of Customer Experience Management (CEM)
The Total Impossibility of Customer Experience Management (CEM)The Total Impossibility of Customer Experience Management (CEM)
The Total Impossibility of Customer Experience Management (CEM)
 
Business Risk Case Study Ba 32
Business Risk Case Study  Ba 32Business Risk Case Study  Ba 32
Business Risk Case Study Ba 32
 
Bank churn with Data Science
Bank churn with Data ScienceBank churn with Data Science
Bank churn with Data Science
 
Insurance Churn Prediction Data Analysis Project
Insurance Churn Prediction Data Analysis ProjectInsurance Churn Prediction Data Analysis Project
Insurance Churn Prediction Data Analysis Project
 
IRJET - Online Donation based Crowdfunding using Clustering and K-Nearest Nei...
IRJET - Online Donation based Crowdfunding using Clustering and K-Nearest Nei...IRJET - Online Donation based Crowdfunding using Clustering and K-Nearest Nei...
IRJET - Online Donation based Crowdfunding using Clustering and K-Nearest Nei...
 
Deloitte UK / Facebook : a people based telecom business - breathing new lif...
Deloitte UK / Facebook  : a people based telecom business - breathing new lif...Deloitte UK / Facebook  : a people based telecom business - breathing new lif...
Deloitte UK / Facebook : a people based telecom business - breathing new lif...
 
New segmentation strategies with Big Data
New segmentation strategies with Big DataNew segmentation strategies with Big Data
New segmentation strategies with Big Data
 
Success in the management ofcrowdfunding projects in the.docx
Success in the management ofcrowdfunding projects in the.docxSuccess in the management ofcrowdfunding projects in the.docx
Success in the management ofcrowdfunding projects in the.docx
 
Canback and D'Agnese - Where in the World Is the Market?
Canback and D'Agnese - Where in the World Is the Market?Canback and D'Agnese - Where in the World Is the Market?
Canback and D'Agnese - Where in the World Is the Market?
 
how_the_digitalchannels_shape_the_future_of_shopping
how_the_digitalchannels_shape_the_future_of_shoppinghow_the_digitalchannels_shape_the_future_of_shopping
how_the_digitalchannels_shape_the_future_of_shopping
 
Retail Banking: In tech we trust
Retail Banking: In tech we trustRetail Banking: In tech we trust
Retail Banking: In tech we trust
 
Reply to DiscussionsD1 navyaA bank failure is the ending of.docx
Reply to DiscussionsD1 navyaA bank failure is the ending of.docxReply to DiscussionsD1 navyaA bank failure is the ending of.docx
Reply to DiscussionsD1 navyaA bank failure is the ending of.docx
 
Time Series Analysis
Time Series AnalysisTime Series Analysis
Time Series Analysis
 
BCI Segment Intelligence
BCI Segment IntelligenceBCI Segment Intelligence
BCI Segment Intelligence
 
AR - Applying Big Data to Risk Management
AR - Applying Big Data to Risk ManagementAR - Applying Big Data to Risk Management
AR - Applying Big Data to Risk Management
 
John Gutfranski, CFP, AIF, CRPC & Debra White Stephens, CFP – Proactive Advis...
John Gutfranski, CFP, AIF, CRPC & Debra White Stephens, CFP – Proactive Advis...John Gutfranski, CFP, AIF, CRPC & Debra White Stephens, CFP – Proactive Advis...
John Gutfranski, CFP, AIF, CRPC & Debra White Stephens, CFP – Proactive Advis...
 
How Banks Can Close the 'Value Gap' and Regain Customer Trust
How Banks Can Close the 'Value Gap' and Regain Customer TrustHow Banks Can Close the 'Value Gap' and Regain Customer Trust
How Banks Can Close the 'Value Gap' and Regain Customer Trust
 

Pledger Model Paper.PDF

  • 1. Int. J. Nonprofi.t Volunt. Sect.Mark. 1O:43-52 (2005) Publishedonline in Wiley InterScience (www.interscience.wiley.com).DOI: 10.I 002/nvsm.5 Pledger modelling: Help the Aged case study Karen Cole,l Rachel Dinglel and Raiesh Bhayani2 t Talking Numbers, UK t Help tbe Aged., uK o Therecruitment ofpledgers (as aproryfor potential legators) to cbaritable organisations plays a uital role in tbeir continued success,and as apercentage of allfundraising income generated it can represent substantial proportions. Hou)euer, of all the 'donation asks' made of supporters, askingfor a legacy is tbe most dfficult. Tberefore, it is important that tbe target audience sbould be as utell researcbed and bigltly targeted aspossible. o Help tbe Aged bad reacbed tbe stage ubere decisions need to be made about itsfuture marketing in order to protect longer-term income. Tbefindings of tbis legacy targeting project utere tofeed into communication progralnmes, direct marketing, and tbe ouerall legacy mar keting strategJ/. o The key objectiue u)as to identifu tlJe best prospects to mail a legaqt ask to, across tbe supporter database, utitlt tlte likeliltood tbat tbey are going to pledge as a result. o It uas found tbat ultilst tailored data analtsis comes at a price, tbe auerage ualue of a legacy justifies the cost of using sophisticated targeting tools. Houteuer, because of tbe pledge-tr>l.egaqt time lapse, tbere utill ahaays be issues uitb measuring any long-term return on inuestment (ROI). Nonetbeless,pledgers baue to be taken on tbeir utordfor tbe purpose of testing (and subsequent rollouts). Pledge data sbould be tested and tbe outcomes sltould inform legaqt marketing. Houteuer, as mentioned aboue, pledgers necessarily need tobe taken ontbeir utord and tberefore,formulatingmodekbased on tbe type and/or ualue of pledges is not recommended. Copyrigbt A 2OO5Jobn lYiley & Sons, Ltd. Background. The recruitment of pledgers (as a proxy for potential legators) to charitable organizations plays avital role in their continued success,and asa percentage of all voluntary income raised, legaciescanrepresent asubstantialproportion. However, of all the 'donation asks' made of supporters, asking for a legacy is the most difficult. According to Richard Radcliffe (flill, Correspondence to: Nigel Magson, Talking Numbers, Iarrcure House,MarketPlace, CirencesterGL7 2NW, UK. E-mail:nmagson@talkingnumbers.com 2OO2)'legacy income is likely to go down as women in their 80s who are leaving legacies now tend to be asset ignorant. . .the next generation will be more aware of the value of their estate and may only leave a fixed sum to charity'. It is important that the target audience should be as well researched and highly targeted as possible. According to the I-egacy Promotion Campaign website (2OOZ)'65% of tbe population are regular cbarity suppor- ters. Recent researcb bas sltoutn tbat donors baue no objection to leauing cbaritable legacies- tbey simply neuerget around to it.' Copyright O 2O05John Viley & Sons,Ltd. Int.J. NonprofitVolunt. Sect.Mark., February 2005
  • 2. 44 K. Cole et al. 25,000,000 20,000,000 15,000,000 10,000,000 5,000,000 Figure l. kgacy income by financial year received. Ifelp tbe Aged's legacy actiuity Help the Aged's legacyincome was S,12.8mil- lion in zOOL/Oz,(see Figure 1.)which repre- sents a significant part of its total voluntary income of 3,75million. Legacymarketing activ- ities include four direct marketing campaigns per annum. In addition, Help the Aged offers a 'Will Advice Service'which, although indepen- dent, doesoften leadto apledge for the charify. The Help the Aged pledger commitment progression is asfollows: Enquirers .---+ Considerers .---= Pledgers c=:> Legators Objectiue Help the Aged had reached the stage where decisions needed to be made about its future marketing in order to protect longer-term income. The findings of this project were to feed into communication pfogfammes, direct marketing and the overall legacy marketing strategy. The key proiect objective was to identify the bestprospects to mail a 'legacyask' to, across the supporter database, with the likelihood that they are going to pledge as a result. Secondaryobjectives wefe asfollows: o Descriptive statistics of legator and pledger data o Centralizing data o Comparative sociodemographic profiles of supporters through to legators I I I I I I I I 1992/93 1993/941994/951995/961996/971997/98 1998/99 1999/00 2000tO12001tO22002103 Financialyear o Timing of 'legacy ask' and other recommen- dations for implementation. Metbodol.ogjt and rfltionale Analysis steps As with any investigative project, thefe afe a number of analysissteps to go thfough before any sort of modelling can be achieved. These main proiect steps are asfollows: r Describe data- summary statistics o Profile data r Reporting o Modelling o Reporting The first step in this case involved under- standing the transition - and differences - between the stagesof legator and pledger com- mitment by producing summary statisticson all availabledata. This also allowed for an evalua- tion of the existing strategy by examining any changes shown by the sufirmary statistics bet- ween groups or over time. Following this analysisthe next stagewas to profile the available data. Profiling is a techni- que that compares gfoups of interest to some sort of base population, for example, the life stagesof legators compared to the GB popula- tion as a whole. Although profiling obviously provides insight in its own right, it is an essential precursof to modelling because it finds the variables that provide the most Copyrighto 2005JohnWiley & Sons,Ltd. Int. J. Nonprofit Volunt. Sect.Mark., February 2OO5
  • 3. Pledger modelling 45 discrimination befween groups. Talking Num- bersprofiled by common databasevariables,as well as by some of the external data variables described later. Modelling The final stages of analysis revolved around modelling the data in order to predict future behaviour. For this project we had a large databasebut with very few existing pledgers, making the modelling more difficult. The potential of modelling legators was examined (seeCTIAIDsectionbelow) but the final models were built for pledges. A pledge is the desired fesponse and, as such, is a behaviour the models should predict. The dataon pledgesdid not allow the models to be tailored to the type of pledge (pecuniary or residuary) or the value of pledge. It should be noted that even a small gain in direct mail should be cost-effective since ave- rage legacies(depending on type) are up to an averageof S25,000 (residuary). The project deliverables were two scores across the supporter database, one for cash donors (including committed givers who have alsogiven acashgift) and one for all committed givers. The scores were based on models indi- cating propensity to become a pledger. Sup- pofters can then be targeted appropriately based on variables alreadv known about them on the database. Talking Numbers used a combination of modelling techniques, namely CHAID and logi- stic regression,to produce the models. Ifithin production of a regressionmodel many models were tested (the dataset size allowed a test sample)in orderto pickthe most discriminatory model to be written back to the supporter base. Talking Numbers also allowed for an inter- active meeting between analystsand Help the Aged in-between the profiling and modelling stages.This allowed the preliminary results to be discussed and any extfa, and important, insight to be provided by the parties actually involved in collecting and using the data. This meeting can be an invaluable contribution to an analysisproiect. CHAID Method Chi-square Automatic Interaction Detection (CHAID) is a form of cluster analysis.This is where the data set is initially thought of asone large cluster and then broken up into anumber of clusters containing related data, aftet exam- ining all interactions. CFIAID can be used to find the most significant discriminating factors for a given dichotomous variable e.g. 1/O or yes/no. CHAID works best with dichotomous variables. Uses CHAID is usedextensively in marketing. One of the usesof CHAID isto discoverwhichvariables are best used to describe respondents/non- respondents.Thus segmentationmodels can be built up for communication purposes. Output example Figure 2 shows atree diagram,or dendogrlm, produced using CIIAID. In Figure 2 R is the response fate. It can be seenthat the response rate for AB males is over 80%, whereas the response rate of DE females is only 17%. Consequently, AB males are deemed a better prospect than DE females. Logistic regression Regressiontechniques are a classof statistical methods in which one dependent variable - the response- is related to one or more independent variables- the predictive or explanatory factors. A regression model is one in which the response variable is linearly related to each explanatoryvariable, i.e. where there is astraight-line relationship between the responseandeachexplanatory variablethere is linear regression. For example, children's weight is often related to their height where weight increases as height increases,this is a straightline relationship and linear regression. Simplelinear regressioniswhere there is only a singleexplanatory variable and multiple regres- sion is where there is more than one explana- tory variable. Logistic regression is applied asa technique when the response variable is of a Copyright ar2O05John Viley & Sons,Ltd. Int.J. Nonprof.t Volunt.Sect.Mark., February 2005
  • 4. 46 K. Cole et al. Figure 2. CHAID tree response analysis. binary type, that is, yes,/no,hasresponded/has not responded etc. Available data Data were provided by Help the Aged from their donor and legacy databases.Many differ- ent fields were extracted in order to allow maximum interrogation: cash and committed giving including Adopt a Granny database donors, legators, pledgers, considerers and enquirers. Iri addition, Talking Numbers investigated external datato allow cold modelling and add value to the existing donor information. MONICA (CACI) MONICA is CACI's age classification based on sophisticated analysisof millions of first names taken from lifestyle questionnaires. Monica helps to identiry the ageof people on adatabase by looking at the likely age profile of their first name, it alsoallows new donors to be targeted by matching a distinct ageprof,le. PRIZM (Acxiom) The PRIZM postcode lifestyle-basedsegmenta- tion and targeting tool assignsGB consumers into one of 6O unique clusters. The data has been aggregatedup to full postcode level. lts core data sources include: o Lifestyle census - information derived from the merger of three companies: Claritas, NDL and CMT, that includes over 40O variables, including income, age,household composition and lifestyle o Company directors - supplied by Dun and Bradstreet o Share ownership - a list of public and pivatized shareholders from NDL's 'Active lnvestor File' database o Behaviourbank - CMT's database contain- ing details of financial products, shopping habits, holiday preferences, appliance own- ership and property types o Unemployment rates- from the Depart- ment of Employment o Birth and death rates - from the Office of National Statistics o Flat, farm and house names- from the Postcode AddressFile @AFl o Census data-from the Office of National Statistics. Every GB postcode was assigned a five-char- acter code basedon the key marketing drivers of life stageand income. The first and second characters indicate the following life stages. PA: Young no children (starting out, young singles and childless couples) PB: Families (nursery and school-age children) PC: Empty nesters (and older singles) PD: Retired seniors. Each of the four life stageswas then broken down into four income indicators: Copyrighto 2005JohnViley & Sons,Ltd. Int.J. Nonprofit Volunt.Sect.Mark., February 20O5
  • 5. Pledger rnodelling 4 7 1. Most affluent households 2. Mid-high affluent households 3. Mid-low affluent households 4. Leastaffluent households. Then a unique cluster code dffierentiated between households that had the same life stageand income band in tems of lifestyle, for example PA311 : Young Conservationists: o Young married r Own four-wheel drive o Likely to have club cards o Support national causes o Support animal welfare charities. StreetValue (CACI) StreetValueprovides Lnavetagevalue for every postcode, the number of properties and pre- dominant type of property pertaining to a postcode area. StreetValue also shows if the properties in a particular areaare mainly own- occupied or rented. These data combined eventually allowed two models to be produced: o Cashdatabasedonors (active and lapsed) o Committed giving database donors (active and lapsed). Find.ings Summary statistics The data exploration stage highlighted the issueswith small ovedaps acrossthe donor and legator databases.The different types ofdonor, howevef, were still compared. Also discovered at this stage was the short time span of the enquirer to legator cycle. This meant that there were few donors found in the mid-cycle gfoups, such as considerers. Also, complete life cycles were available for only a few individuals. However, each piece of the enquirer, considerer, pledger and legator cycle was examined in its own right and sufirmary information produced. For example, Figure 3 shows this informationforthe legator cycle. Profiles Legators and pledgers have a very similar life stageprofile to each other. They are older than the other donor groups. Other cashsupporters (those who have not made anylegacy contact) show a younger age profile, similar to con- siderersand enquirers, while committed givers are different again and areby far the youngest group. The affluence profiles are different, however. The legator profile most closely matches that of the other cash supporters and committed givers (those who have not made any legacy contact). In terms of affluence, pledgers are similar to considerers and enquirers comprise the least affluent group. Pledgers are more likely to be aged 50-59 (MONICA) and retired seniors GRIZND. The results from the StreetValue analysis were interesting but did not show enough differ- entiation to use in a model, Financial year Figure 3. Number of legators by financial year of notification date 1992/931993/94 1994/95 1995/96 1996/97 1997/98 1998/99 '1999/0020001012001t02 2002t03 Copyrightto2005JohnWiley & Sons,Ltd. Int.J. Nonprofit Volunt.Sect.Mark., February 2OO5
  • 6. 48 K. CoIe et al. Pledger profiles are interesting in terms of their cash-giving history. Nearly half of all pledgers (who are also cash supporters) are in the top value band. The averagevalue of cash gift is a very important factor in the pledger cash model. Pledgers are the group with the highest frequency profile and frequency of cash gift also appears in the final pledger cash model. Geograpbical patterns The geographical spread of the number of legators was examined and this showed a distinct Southern bias to the pattem, with Scotland in particular providing very small numbers of legators. A geographic analysisof the averagelegacyvalueswas alsoundertaken. This time, a pattern emerged whereby the average legacy values rose in correlation to the geographical affluence of each Lrea, particularly in the South East.Additionally, a numberof interesting hot spotswere identffied including the 'retirement' areasof Torquay and Brighton. CHAID After running tables and comparing profiles, CHAID was used as an exploratory technique to investigate the relative importance of differentiating factors and how those factors combine to provide the most differentiation (and hence prediction). CHAID exploration was carried out looking at prediction of both pledgers and legators; rnd avalTablevariables included cash-giving history, legacy contact made and external data such as MONICA and PRIZM.As there is aclear relationship between the life rycle groups, i.e. legators are likely to have been pledgers, etc., it was decided to model pledgers and not a combination of gfoups. After the profiles were studied, it also became apparent that two models would be needed, one for cash donors and one for committed givers. These two groups were quite different in profile and it was expected that they would behave very differently. There- fore one model for both groups would have been inappropriate. Modelling: cash donors The same modelling approach was used for both cash donors and committed givers. The profiling resultswere studied andvariablesthat might become predicted factors were high- lighted. Then both CHAID and regression techniques were used to refine this list until the best predictors in combination were selected for the final models, and their respec- tive levels of importance were allocated. This methodology produced the following factors in the final cash donor model: o Other relationships/donor status/enquiries made o MONICA ageband o Cash-givinghistory, value and frequency. The 'gains' achieved by the model are evident as half of all pledgers were found in the first (righest scoring) 3O%of cash supporters, as shown in Figure 4andTable 1. Modelling: coflrmitted givers The final model for committed givers included the following factors: o Length of relationship o Other relationships o PRIZM, both life stageand affluence o MONICA. Over half of all pledgers were found in the top 25%of committed giving supporters asshown in Figure 5 andTable2. Implementatiom Writeback Each score effectively ranked the indMduals according to their propensity to pledge. These two scores were appended to about 5OO,OOO records on the donor database at Help the Copyright O 2005 John rlfliley & Sons,Ltd. Int.l. Nonprofit Volunt.Sect.Mark., February 2005
  • 7. Pledger modelling 49 60% o E 50o/o E 40% No/o no/o 10% 0% Figure 4. Cashpledger model gains chart. Aged. Talking Numbers worked with Help the Aged to plan mailing volumes using the model - bearing in mind that certain donors and donor groups would be excluded from mailings due to suppressions or othef activity. The aim for the next yearwas to test the models within the direct marketing prografirme, com- bined with selections based on edsting donor segments. Table 1. Cashpledger model Spring 2003 The February 2OO3 legacy campaign was mailed to 28,000 cash donors, split into eight equal mailing segments according to standard RFVcriteria andpledger propensity score.Two random 3,500 samples, labelled Band 1 and Band 2, were taken for each of the RFV segments- Band 1 comprised those donors NTILE* Pledgers Model (%) Random (%o) Model cumulative (%) Random cumulative (%) I 2 3 4 5 6 7 8 9 10 l l l 2 r3 r4 1 5 r6 t7 18 r9 20 Total 216 216 206 r93 159 101 r62 r49 r55 r45 1 1 0 77 7r 56 63 1 3 8 6 7 69 2,t82 9.90 9.90 9.44 8.85 7.29 4.63 7.42 6.83 7.ro 6.65 5.O4 3.53 1 2 q 2.57 2.49 0.60 o.37 o.27 o.32 3.16 100.oo 5.00 5.00 5.00 5.00 5.O0 5.00 5.00 5.OO 5.OO 5.00 5.00 5.O0 5.00 5.00 5.00 5.O0 5.00 5.00 5.OO 5.00 100.00 9.90 19.80 29.24 38.08 +>.t/ 50.oo 57.42 64.25 71.36 78.00 83.O4 86.57 89.83 92.39 95.28 95.88 96.24 96.52 96.84 100.oo 5.00 to.o0 15.00 20.00 25.OO 30.00 35.00 40.oo 45.OO 50.00 55.00 60.00 65.OO 70.oo 75.OO 80.o0 85.00 90.00 95.00 100.00 .An NTILE is an equal sized group of ranked d^ta, in this case the data is split into 2Oequal sized groups. Copyrighto 2O05JohnI(iley & Sons,Lrd. Int.J. Nonprofit Volunt.Sect.Mark., February 2005
  • 8. 50 K. Cole et al. E oov" 8. mv" o p +ou 450/o 55% Contact Figure 5. Committed giving pledger model gains chart. with a pledger score in the bottom 5O%of the model, and Band 2 comprised those with a scoreof the next bestz0%.lt hadbeen decided previously that those with a score in the top 30% should not be tested in this the first campaign, but 'saved' for later campaigns. Summer 2OO3 To further test the effectivenessof the pledger models for targeting potential legators, there were two campaignsin Summer2003. Help the Table2. Committed giving pledger model Aged mailed donors with a high propensity to pledge and those within the samedatabaseseg- ments with a low propensity to pledge (5O/5O split). Significant pledger results woe found for both models ascan be seenin Tables 3 and4. Moving forward 2OO3in summaty The results for 2o03looked promising in terms of the effectiveness of the pledger models. It NTILE* Pledgers Model (%") Random (%) Model cumulative (%) Random cumulative (%) Index I 2 J 4 ) 6 7 8 9 l0 l l t 2 7 3 14 1 5 r6 17 l8 r9 20 Total 103 62 ) 27 38 37 2 l 22 r9 2a r6 18 7 t t l 2 r3 9 0 7 2 504 20.44 12.30 ro.32 5.36 /.>+ 7.34 4.r7 4.37 ) . / / 5.56 3.r7 5. >/ r.39 2.La 2.38 2.58 r.79 0.00 r.39 o.40 100.oo 5.OO 5.00 5.OO 5.OO 5.OO 5.OO 5.00 5.00 5.00 5.00 ).UU 5.00 5.O0 5.O0 5.O0 5.O0 5.00 5.00 5.00 5.00 100.oo 20.44 32.74 43.06 48.41 55.95 63.29 67.46 7r.83 75.60 8 1 . 1 5 84.33 87.90 49.29 9r.47 93.85 96.43 98.2r 98.2r 99.60 100.00 5.OO 10.00 15.00 20.00 25.OO 30.00 35.00 40.oo 4r.oo 50.00 55.O0 60.00 65.O0 70.oo 75.OO 80.o0 85.00 90.00 95.00 l00.oo 4.o9 1 )'7 2.87 2.42 2.24 2 . t r r.93 1.80 1.68 r.62 r.53 r.46 r.37 1.31 r.25 t.2 l r . 1 6 1.09 1.05 1.00 *An NTILE is an equal sized group of ranked datt, rn this case the data is split into 20 equal sized groups. Copyright o 2OO5John Wiley & Sons,Ltd. Int.t. Nonprofi.tVolunt.Sect.Mark., February 2005
  • 9. Pledger rnodelling 51 Segment Table 3. July 2003 (committed giving model) o Telemarketing channel as tested in 2OO3 may be included in the modelling. There will be two models again as the behaviourof cashdonors and committedgivers are clearly different. The inclusion of high- value donors and others excluded previously may increase overall response rates. 2003's new creative model will again be used, after response rates and research showed its effectiveness. 2004 actiuity August 2OO4committed givers mailing September 2OO4cash givers mailing (total 70,ooo) September 2004 outbound follow-up call to mailed committed givers. Conclu.sions The LegacyPromotion Campaign(2OO2)states: 'Over the past 12yearsthe percentage of those wills going to probate containing a legacy to charity has remained constant at arcund l3%. This is despite all the resources that charities have put into legacy marketing, indicating that they only succeededin competing for ashareof the same wills. There is evidence that whilst the number of people making wills is increas- ing, the proportion of those that include a charitable legacy is in decline'. This makes careful targeting through direct communication even more critical. Targeting will help ensure charities' legacy income is maintained or even increased. Targeting further ensufes cost-efficiencies since careful selection means the most likely prospects are communicated to with a proposition that they're more likely to accept. Thilst tailoreddataanalysis comesat aprice, the averagevalue of alegacyiustifies the cost of using sophisticated targeting tools. However, because of the pledge-tolegacy time lapse, there will always be issues with measuring any long-terrn return on investment (ROD. Nonetheless, pledgers have to be taken on their word for the purpose of testing (and Score Low High Low No.mailed 7,5OO 7,5oO 3,500 Pledgerresponse 19 56 9 Responserate O.25% O.75% 0.26% HiCh 2,444 ) 4 0.94o/n CG: committed givers segment. AAG:'Adopt A Granny' committed givers make aS,l2 per month (approximately) regular gift. Table 4. August 2OO3(cash model) Segment Active Active a O Score No. mailed Pledger response Responserate Low 6,oo2 8 o.13% should be noted that there were other factors that would have influenced the results. For example, Help the Aged tested a new creative pack alongsidethe cashpledger model. As part of ongoing analysis and testing of the models there was a need to re-run the models on the entire donor database,including new donors and high value donors who were excluded previously, to enablebettertargeting of donors n2OO4. A full post-campaign meta-analysisfor 2OO3 has also been suggested so that the number of responses across the groups could be come mofe statistically robust. A meta- analysisis a statistical way of combining more than one set of test results together in order to give the analyst latger numbers of responders on which to basethe analvsis. Neut model o Z0D3post-campaign meta-analysis o New strategy for 2OO4targeting, including new channels (telemarketing) o Examine the profiles and update the learning irrz003 for new pledgers HiCh 7,OO3 25 O.36o/o High 7,OO3 26 o.37% copyright o 2005Johnffiley & sons,Ltd. Int.l. Nonprofit Volunt.Sect.Mark., February 2005
  • 10. 52 K. Cole et al. subsequent rollouts). Pledge data should be tested and the outcomes should inform legacy marketing. However, as mentioned above, pledgers necessarilyneed to be taken on their word and therefore, formulating models based on the type and/or value of pledges is not recommended. References Hill N. 2OO2. Q&A: legacies. Tbe Guardian, October lL,2OO2. LegacyPromotion Campaign Website. 2002. www. legacypromotioncampaign.org.uk[May 2OO4]. copyright @2005 John l(iley & Sons,Ltd Int. I. Nonprortt Volunt. Sect.Mark., February 2OO5