listen and learn - office of the auditor general...listen and learn 2 executive summary this...
TRANSCRIPT
LISTEN AND LEARN
Using customer surveys to report performance
in the Western Australian public sector
PERFORMANCE
EXAMINATION
Report No 5 – June 1998
W e s t e r n A u s t r a l i a
A U D I T O R G E NA U D I T O R G E N E R A LE R A L
© 1998 Office of the Auditor General Western Australia. All rights reserved.This material may be reproduced in whole or in part provided the source is acknowledged.
Clipart from Microsoft Clipart.
4th Floor Dumas House
2 Havelock Street
West Perth WA 6005
Telephone: (08) 9222 7500
Facsimile: (08) 9322 5664
E-mail: [email protected]
http://www.audit.wa.gov.au/
A U D I T O R G E N E R A L
W e s t e r n A u s t r a l i a
The Office of the Auditor General is acustomer focused organisation and is
keen to receive feedback on the qualityof the reports it issues.
Through Performance Auditing enable the Auditor General
to meet Parliament’s need for independent and impartial
strategic information regarding public sector
accountability and performance.
MISSIONMISSIONof theof the
Office of the Auditor GeneralOffice of the Auditor General
VISIONVISION
Office of the Auditor GeneralOffice of the Auditor Generalof theof the
Leading in Performance Auditing
Report No 5 – June 1998
LISTEN AND LEARN
Using customer surveys to report performancein the Western Australian public sector
W e s t e r n A u s t r a l i a
A U D I T O R G E N E R A L
AUDITOR GENERAL
Western Australia
THE SPEAKER THE PRESIDENTLEGISLATIVE ASSEMBLY LEGISLATIVE COUNCIL
PERFORMANCE EXAMINATION — LISTEN AND LEARN – Using customer surveys to
report performance in the Western Australian public sector
This Report has been prepared consequent to examinations conducted under section 80 of the
Financial Administration and Audit Act 1985 for submission to Parliament under the provisions
of section 95 of the Act.
Performance examinations are an integral part of my overall Performance Auditing Program and
seek to provide Parliament with assessments of the effectiveness and efficiency of public sector
programs and activities, thereby identifying opportunities for improved performance.
The information provided through this approach will, I am sure, assist Parliament in better
evaluating agency performance and enhance Parliamentary decision making to the benefit of all
Western Australians.
D D R PEARSON
AUDITOR GENERAL
June 24, 1998
Contents
Executive Summary 1
Overall Findings and Conclusions 2
Summary of Recommendations 4
Introduction 6
Performance Indicator Trends 6
Types of Customer Feedback 7
Examination Focus and Approach 8
Survey Costs and Utilisation 10
Conclusions 10
The Cost of Customer Surveys 10
Contracting Out Customer Surveys 12
Utilising Survey Findings 13
Recommendations 14
Survey Accuracy and Reporting 15
Conclusions 15
Technical Accuracy of Customer Surveys 15
Survey Reporting and Record Keeping 24
Recommendations 26
Appendix 1: Key Steps in Conducting Customer Surveys 27
Appendix 2: Judging the Quality of Survey Research: A Checklist 29
Appendix 3: A Guide to Sample Size Requirements 31
Performance Examination Reports 32
Using customer surveys to report performancein the Western Australian public sector
1
Executive Summary
The Western Australian public sector is increasingly using the results of
customer satisfaction surveys as one of the indicators used to report on its
performance.
Excluding health service agencies, the number of agencies reporting customer
satisfaction as an effectiveness indicator has doubled in the past three years.
Customer satisfaction is now the most frequently reported indicator of
effectiveness – used by 66 per cent of all agencies in 1996–97 annual reports.
Despite the most careful procedures, all surveys involve potential errors
that can introduce uncertainty or bias. For the results to be credible, error
must be reduced whenever possible and reported results should disclose any
significant survey limitations (see Figure 1). The goal is not technical
perfection but credibility and appropriateness in relation to intended use.
Intended use within the Western Australian public sector can entail:
� An ongoing and developing accountability relationship requiring agencies
to report on their performance to Parliament – without high standards
of rigour it is easy to inadvertently mislead;
� An increasing number of policy and resource allocation decisions within
government being based on customer feedback; and
� The linking of customer satisfaction data to public sector performance
pay and Chief Executive performance agreements.
Figure 1: Achieving Survey Rigour and Credibility
If publicly reported customer satisfaction surveys are to have credibility, certain technical
standards need to be met.Source: OAG
data collectionanalysis and reporting
record keeping
sampling precisionresponse rates
unbiased sampling
reliability
validity
Auditor General WA
LISTEN AND LEARN»»
2
Executive Summary
This performance examination appraised the rigour with which public sector
agencies had conducted selected customer satisfaction surveys. Seven surveys,
each from a different agency, were examined as in-depth case studies. The
examination reveals issues that are likely to be relevant to all public sector
agencies.
Overall Findings and Conclusions
Survey Costs
Survey costs ranged from an estimated $3 000 to $74 000, the more costly
being larger and of better quality. The average survey cost per respondent
varied from just over $30 to about $160. Cost per respondent was not a
good indicator of survey rigour.
Small budget agencies with large numbers of customers may find it difficult
to meet the cost of undertaking reliable annual customer surveys. Alternatives
include sharing survey costs with other agencies that have similar customer
groups. This would also reduce the risk of ‘respondent overload’ when the
same group of customers is repeatedly surveyed by different government
agencies.
Five of the agencies used consultants to assist in their surveys. All agencies
received value for what they paid. However agencies still need to possess,
or have access to, survey knowledge and skills in order to manage consultants.
Utilising Survey Findings
The seven agencies examined were able to demonstrate, to varying degrees,
actions that they had taken as a result of feedback from customer surveys.
The degree of Chief Executive support, and the linking of surveys to other
service improvement strategies, were important factors that assisted agency
utilisation of customer satisfaction surveys.
Using customer surveys to report performancein the Western Australian public sector
3
Executive Summary
Survey Accuracy
Many agencies have made a considerable effort to tackle this new and difficult
area of measuring the satisfaction of public sector customers. However, all
of the surveys examined had some weaknesses and several had many.
Sampling Methods
Three of the seven agencies were able to demonstrate that they had taken a
rigorous approach to developing a list of their customers, and then taking a
sample of people from this listing. The other four agencies had difficulties
such as defining their customer groups and constructing a customer database.
Response Rates
Low response rates (under 50 per cent) were observed in five of the seven
surveys examined. This increases the risk of survey bias, which can give
misleading results. Only one agency undertook some further analysis to try
and determine the direction and extent of this potential bias.
Sampling Precision
Sampling precision ranged from an excellent low of ± 3 per cent to an
unacceptable high of ± 17 per cent. In the latter case, a reported measure of
57 per cent customers satisfied with the program’s performance could mean
that anywhere between 40 per cent and 74 per cent were satisfied. Agencies
with a high level of sampling error in their surveys were generally unaware
of its significance. Improving the precision of agency surveys can be achieved
by taking larger samples of agency clients, and by reducing levels of non-
response.
Validity and Reliability of Measurement
Validity addresses the issue ‘Does the survey really measure what it is supposed
to measure?’. Developing a valid survey for performance reporting purposes
requires agency objectives that are clear, specific and measurable. Survey
validity has to be planned for and tested at the survey design phase. None of
the agencies had done this. Using the most basic of tests, one of the surveys
appeared to have problems with survey validity.
Auditor General WA
LISTEN AND LEARN»»
4
Executive Summary
Reliability is about consistency of measurement and needs to be considered
as part of a survey’s development. None of the agencies had done this, which
is of particular concern for the three surveys that had a limited number of
respondents. These three surveys are at more risk of being unreliable and
hence their reported performance indicators are more likely to be inaccurate.
Survey Reporting and Record Keeping
Several of the agencies have made errors in the analysis and/or interpretation
of their survey findings. In particular, they had problems in interpreting
changes in customer satisfaction ratings over time. This appeared to be a
problem of agency expertise, rather than a matter of deliberate
misrepresentation.
The reported performance indicators often contained little additional
information to assist readers assess agency performance. For example they
did not include a standard or a benchmark against which to compare agency
performance. They also did not generally include explanatory notes to inform
the reader about the survey’s technical limitations.
Agencies and/or their consultant were able to provide the examination team
with copies of all requested documents generally with minimal delay. However
managers were unsure of the retention periods that applied to these records.
Summary of Recommendations
Agencies should:
�� ensure that their surveys are conducted in a scientifically rigorous manner
so as to minimise all types of survey error;
�� present their performance indicators in conjunction with relevant
supporting information such as comparative benchmarks and the survey’s
technical limitations;
�� assess the cost-effectiveness of undertaking customer surveys,
particularly those agencies with small budgets and a large number of
customers;
Using customer surveys to report performancein the Western Australian public sector
5
Executive Summary
�� ensure that they have access to sufficient survey knowledge to effectively
manage consultants when contracting out their surveys;
�� be aware of the possibility of over burdening their clients with requests
to participate in customer surveys. Agencies with common clients should
liaise with each other to avoid this situation; and
�� use their survey findings as a tool to assist in service improvement and
as a means of demonstrating accountability.
Auditor General WA
LISTEN AND LEARN»»
6
Introduction
Public sector agencies conduct customer satisfaction surveys for two main
reasons:
� for feedback to monitor and improve service; and
� as a means of demonstrating public accountability for performance.
In addition, key customer service objectives are reflected in Chief Executive
Officer performance agreements. Similarly, customer satisfaction surveys
can also be used as staff productivity measures within agency enterprise
bargaining agreements and workplace agreements.
Performance Indicator Trends
Within the Western Australian public sector, all departments and most
statutory authorities are required to report key performance indicators in
their annual report to Parliament. These indicators are an essential component
of public sector accountability, enabling Parliamentarians and citizens to
assess the efficiency and effectiveness of public sector operations. They are
audited by the Auditor General.
Excluding hospitals and other health services, the number of agencies
reporting customer satisfaction1 as an indicator of their effectiveness has
doubled from around 40 to over 80 in the past three years. For all public
sector agencies, it was the most frequently used measure of effectiveness in
1997 Annual Reports (Figure 2).
1 The terms customer satisfaction and customer feedback are used interchangeably in this report.
Using customer surveys to report performancein the Western Australian public sector
7
Introduction
Figure 2: Types of Effectiveness Performance Indicators
Customer satisfaction is the most frequent measure of effectiveness in Annual Reports –
reported by 66 per cent of all agencies in 1997.Source: OAG
Being responsive to customer feedback is part of the Government’s public
sector customer focus strategy, launched by the Premier in 1994. The primary
aim of the strategy is to ensure that the Western Australian public sector
continuously improves service delivery, and provides value for money service
to the community of Western Australia. Agencies are required to develop
customer service charters as part of the strategy.
Types of Customer Feedback
The conclusions one can draw from customer feedback depend upon how the
information has been obtained. For example, casual comments from customers
can offer insights that help to improve service delivery, but a rigorous scientific
survey is needed to yield results that can be generalised with reasonable
certainty to all of the agency’s customers.
Customer satisfaction
Timeliness measures
Defined outputs
Independent accreditation
Client awareness and behaviour
Industry benchmarks
Output indexes
Economy activity
0 10 20 30 40 50 60 70
Percentage of Agencies using Effectiveness Indicators
Auditor General WA
LISTEN AND LEARN»»
8
Introduction
Some common approaches to obtaining customer feedback, and their
appropriate usage, include:2
Suggestion boxes
Provide a relatively non-threatening way for customers to express their
preferences and make suggestions.
Complaint handling
Establish formal systems to record customer complaints. Seek to immediately
address complaints as they arise in addition to identifying any common or
recurring patterns over time.
Focus groups
Focus groups have been used in marketing for a long time. They are used to
get direct reactions from customers on goods and services offered and to
provide an opportunity for consumers to speak. They are especially useful in
identifying needs and assessing issues relating to the introduction of a service.
Customer surveys
Scientifically rigorous surveys are particularly appropriate when precise and
unbiased information is required to support major management decisions,
and for the purposes of demonstrating accountability. For a description of
the key steps in undertaking customer surveys, see Appendix 1.
Examination Focus and Approach
The aims of the performance examination were to:
� assess the technical quality of selected agency customer satisfaction
surveys;
� identify common issues and problems and examples of good practice;
and
� overview the differing survey costs and approaches.
2 For a discussion of these and other approaches see: OECD, 1996, Responsive Government –Service Quality Initiatives, author, Paris.
Using customer surveys to report performancein the Western Australian public sector
9
Introduction
Seven agencies were purposefully selected to provide a reasonable cross-
section: large versus small budgets; service delivery agencies versus policy
advising bodies. Given the aims and timeframes for the examination, it was
considered appropriate to undertake a limited number of in-depth case
studies, as opposed to surveying a representative sample of agencies.
From each of the seven selected agencies one customer survey, included as a
performance indicator in the agency’s 1997 annual report, was selected for
examination. The seven surveys selected were:
� Customer Perception Survey – Department for Family and Children’s
Services (FCS);
� Land Operations Survey – Department of Land Administration (DOLA);
� Survey of Local Governments – Department of Local Government (DOLG);
� Customer Perception Survey – Department of Transport, Maritime
Division (Transport);
� National Tenant Satisfaction Survey – The State Housing Commission
(Homeswest);
� Survey of Community Organisations – Office of Multicultural Interests
(OMI); and
� Key Performance Indicator Survey – Rottnest Island Authority (RIA).
Any other customer satisfaction surveys conducted by the agencies selected
were not examined in detail. In addition, the examination did not attempt to
assess whether the agencies’ objectives were best measured by a customer
satisfaction survey or by an alternative methodology. It simply appraised
the rigour with which each survey was conducted. Finally, this report is not
intended as a general guide to the development and reporting of performance
indicators.3
3 See:
� Preparing Performance Indicators – A Practical Guide (joint publication Treasury Department ,Public Sector Management Office, Office of the Auditor General – April 1997);
� Under Wraps! Performance Indicators in Western Australian Public Hospitals (special report No.4, Office of the Auditor General – August 1996);
� Output Based Management (OBM) – Guidelines to Assist Agencies (Treasury Department – July1996); and
� Public Sector Performance Indicators (special report No. 7, Office of the Auditor General –December 1994).
Auditor General WA
LISTEN AND LEARN»»
10
Survey Costs and Utilisation
Conclusions
� Agencies have to make a trade-off between survey cost and obtaining
quality data – for some agencies containing costs was the primary
consideration.
� The level of Chief Executive support and the linking of surveys to other
service improvement strategies were important factors that assisted
agencies with the utilisation of survey findings.
The Cost of Customer Surveys
Survey and Was the Data Number of Total Cost perAgency Survey Collection Completed Survey RespondentNames Contracted Method Surveys Cost (1) (estimated)
Out? (estimated)
Customer Perception Yes Telephone interviews & 1 012 $39 704 $39
Survey – FCS mailed questionnaires
Land Operations No Mailed questionnaires 29 NA (2) NA
Survey – DOLA
Survey of Local No Faxed questionnaires 34 $5 464 $161
Governments
– DOLG
Customer Perception Yes Mailed questionnaires 769 $25 250 $33
Survey – Transport & telephone interviews
National Tenant Yes Face to face interviews 1 125 $74 036 $66
Satisfaction Survey & mailed questionnaires
– Homeswest
Survey of Community Yes Telephone interviews 24 $2 800 $117
Organisations – OMI
Key PI Survey – RIA Yes Telephone interviews 150 $5 500 $37
Table 1: Survey Methods and Costs
The seven surveys examined differed in size, cost and methodology.
Notes:(1) Cost estimates were based on known consultant costs plus estimated direct salary costs
of in-house staff – excepting DOLG’s estimate which was based on direct salary costs
only.(2) NA = Not available. DOLA was unable to separate its in-house staffing costs for this
particular survey from its other organisational customer focus initiatives.
Source: Agencies and OAG
Using customer surveys to report performancein the Western Australian public sector
11
Survey Costs and Utilisation
There was considerable variation across agencies in survey approaches and
costs.
OMI and RIA viewed their survey primarily as a means to obtain data for
performance reporting purposes. DOLA focused more upon the survey’s
potential for obtaining information to improve agency services. The remaining
agencies sought more of a balance between these two purposes.
The small budget agencies generally undertook a simple survey of a small
number of customers, while the larger budget agencies obtained more
comprehensive information from a large number of customers. As a result,
the total estimated cost of agency surveys varied from a low of just under
$3 000 by OMI to a high of about $74 000 by Homeswest.
Total survey costs are driven primarily by the number of respondents and
the data collection method utilised. This requires agencies to make an
informed trade-off between the cost and timeliness of their survey versus
obtaining high quality data. Small budget agencies with large numbers of
customers may find it difficult to meet the cost of undertaking reliable and
meaningful annual customer surveys. Such agencies may need to consider
other options such as sharing survey costs with other agencies that have
similar customer groups or undertaking customer satisfaction surveys less
frequently.
In addition to cost considerations, there may be other advantages in agencies
with similar customer groups coordinating their surveys. In particular it
reduces the risk of ‘respondent overload’ when the same group of customers
is repeatedly surveyed by different government agencies.
Survey cost per respondent also varied from a low of $33 for Transport
through to a high of $161 for DOLG. There were many factors influencing
the survey cost per respondent including:
� the presence of economies of scale;
� the methods of data collection used – face to face interviews are more
expensive than telephone interviews and mailed questionnaires;
� the survey response rate – lower response rates leading to higher unit
costs; and
� an agency focus upon cost minimisation rather than data quality.
Auditor General WA
LISTEN AND LEARN»»
12
Survey Costs and Utilisation
Contracting Out Customer Surveys
Five of the seven agencies examined used a consultant to conduct their
survey. Consultants were generally used to initially develop the agency’s
survey, to collect data from the agency’s customers, and to then prepare a
report on the findings. The most common reasons agencies gave for using a
consultant were:
� to obtain expertise not available in-house;
� to encourage customers to respond to the survey in a full and frank
manner which could be compromised if the agency undertook the data
collection itself; and
� to give additional assurance that data collection and analysis was
undertaken in an impartial manner.
All agencies received value for what they paid. However agencies still need
to possess, or have access to, survey knowledge and skills even if they wish
to contract out their surveys. Without such skills, they risk not being able
to adequately direct, control and evaluate their consultant. For the benefit of
managers, whether using consultants or in-house staff, a checklist for judging
the quality of survey research is given at Appendix 2.
A brief review of the tendering process was undertaken within those five
agencies that had contracted out their customer satisfaction survey. In all
cases, the agencies had:
� sourced their consultants in accordance with State Supply Commission
purchasing policies; and
� made payments in accordance with the agreed contract terms.
Most agencies chose to initially test a consultant on a short-term contract
before committing to a longer-term contract. However, a series of one-off
annual contracts may result in higher costs for the agency, and a lower quality
of service over time. Agencies should therefore consider forming a somewhat
longer- term contract with their consultant to get the best value for money.
This approach, if applied appropriately, is consistent with State Supply
Commission purchasing policies.
Using customer surveys to report performancein the Western Australian public sector
13
Survey Costs and Utilisation
Utilising Survey Findings
Public sector agencies are increasingly turning to customer surveys as a tool
to help them become more responsive and effective in serving their customers.
The seven agencies examined were able to demonstrate, to varying degrees,
actions that they had taken as a result of feedback from customer surveys.
The level of Chief Executive support and the linking of surveys to other
service improvement strategies were important factors that assisted agency
utilisation of customer surveys.
DOLA in particular could demonstrate an integrated holistic approach to the
collection and utilisation of survey data. This approach is illustrated in Figure 3.
Figure 3: Utilising Customer Feedback
DOLA has an integrated organisational approach to the collection and utilisation of customer
feedback.Source: OAG & DOLA
� Training and education� general and accredited training� Process and system changes� Redesigned Job Description
Forms that now include customerservices competencies� Incentives/rewards for staff� Division/Branch awards� annual DOLA-wide awards� Staff pay linked to performance
� DOLA’s culture is becoming morefocused on customer service� Improved customer service
e.g. quality and timeliness� Improved efficiency and productivity� New products and services developed� Improved stakeholder consultation
An ongoing participative process
Customers and staff are asked to:� rate the quality of current services;� identify their own needs; and� identify their priorities for service
changes
Customer Service Charterand Standards Developed
Changes and Impact Market and CustomerResearch
Strategies ImplementedAssessment, Measurement
and Feedback
Management Action
� Strategic planning� set priorities� allocate resources� Monitor progress� Accountability reporting
Results from the customer researchare fed back to:� management� staff� DOLA’s Customer Council and� individual customers through
displays in DOLA’s offices.
Auditor General WA
LISTEN AND LEARN»»
14
Survey Costs and Utilisation
Other examples of service improvements noted during the examination
include:
� FCS became aware that Aboriginal people were under represented in
their customer perception survey. As a result they commissioned a series
of separate Aboriginal perception surveys that were conducted in a more
culturally appropriate manner.
� An area of concern identified in Transport’s survey was delays
experienced by customers when making boat registration payments.
Transport introduced a payment facility through Australia Post to address
this problem in June 1997. The new payment facility proved popular,
attracting over 40 per cent of all boat registrations in its first six months
of operation.
� Homeswest’s customer survey identified tenant dissatisfaction with time
delays in the completion of house maintenance. As a result, Homeswest
introduced new maintenance contracts in January 1998, which specified
shorter maintenance times according to the degree of urgency.
Homeswest tenants were advised of the new standards, and the contracts
provide for the payment of an incentive bonus to contractors for meeting
these new standards.
Recommendations
Agencies should:
�� ensure that they have access to sufficient survey knowledge to effectively
manage consultants when contracting out their surveys;
�� be aware of the possibility of over burdening their clients with requests
to participate in customer surveys. Agencies with common clients should
liaise with each other to avoid this situation; and
�� use their survey findings as a tool to assist in service improvement and
as a means of demonstrating accountability.
Using customer surveys to report performancein the Western Australian public sector
15
Survey Accuracy and Reporting
Conclusions
� Many agencies have made considerable effort to tackle this new and
difficult area of customer satisfaction surveying. However, all of the
surveys examined had some technical weaknesses, and several had many.
� Agencies that made greater use of customer feedback for performance
management generally also undertook better quality surveys.
� Agency surveys need not be technically perfect, but they do need to be
credible for their intended use.
� When reporting on performance, agencies should advise readers of any
significant limitations in the data presented.
Technical Accuracy of Customer Surveys
Performance reports are designed to help improve public programs, provide
accountability to the public, and guide decision makers on the allocation of
resources. If customer satisfaction surveys are to provide accurate and useful
information, they need to be properly conducted. This is illustrated in Figure 4.
Figure 4: The Importance of Survey Rigour
The level of reporting accountability, and quality of management decision making, is dependent
upon the underlying survey rigour.Source: OAG
Improved decision making andpublic services
Accurate and crediblesurvey findings
Enhanced accountability
validityt
reliabilityt
record keepingt
analysis and reportingt
sampling precision
response rates
unbiased sampling
data collection
u
u
u
u
p
Auditor General WA
LISTEN AND LEARN»»
16
Survey Accuracy and Reporting
Some of the more important technical issues that need to be addressed to
obtain accuracy when undertaking customer surveys include:
� Sampling Methods – in particular making sure the sample is randomly
selected, and the size is adequate, to increase the likelihood that the
sample accurately represents the population;
� Data Collection Methods – there is a trade-off between the quality of the
data obtained and the cost of different data collection methods;
� Response rates – low response rates can produce biased results;
� Sampling Precision – the difference between some true population value
and the estimate of this value obtained from surveying a sample of people;
� Validity – Is the survey measuring what it is supposed to?; and
� Reliability – Does the survey measure in a consistent manner?
Key technical data from the seven agency customer surveys is summarised
in Table 2.
Using customer surveys to report performancein the Western Australian public sector
17
Survey Accuracy and Reporting
Survey and Population Initial Response Sampling Validity (2) Reliability (3)
Agency Size Sample Rate Precision (1)
Names Size
Customer Perception 31 638 4090 25% ± 3% ü ü
Survey – FCS
Land Operations 253 253 11.5% ± 17% ü û
Survey – DOLA
Survey of Local 144 49 69% ± 15% ü û
Governments
– DOLG
Customer Perception 59 362 1818 42% ± 3.5% ü ü
Survey – Transport
National Tenant 34 188 2218 51% ± 3% ü ü
Satisfaction Survey
– Homeswest
Survey of Community 50 50 48% ± 14% û û
Organisations – OMI
Key PI Survey 316 940 413 36% ± 8% ü ü
– RIA
Table 2: Technical Quality of Customer Surveys
The technical quality of the surveys was variable. Common problems included: low response
rates; poor sampling precision; and not examining survey validity and reliability.
Notes:(1) Sampling precision as measured by the standard error of a proportion at the 95 per cent
confidence level with an assumed proportion of incidence in the population of 0.5.
(2) None of the agencies had tested their surveys for validity. Ticks are given for meeting the
most basic criterion of having face validity (appearing to be a sound measure of the
concept in question).
(3) None of the agencies had tested their surveys for reliability. Ticks are given to those
surveys with a larger number of survey respondents that are at less risk of producing
unreliable performance indicators.
(4) Population size is an estimate.Source: OAG and agencies
(4)
(4)
Auditor General WA
LISTEN AND LEARN»»
18
Survey Accuracy and Reporting
Sampling Methods
Customer surveys usually involve studying a sample of agency customers
with a view towards reaching conclusions about all of the agency’s customers.
Important sampling tasks to be considered include:
� Defining the customer group;
� Obtaining a complete and accurate customer list from which to select a
sample;
� Ensuring that each customer has a known non-zero probability of being
randomly selected to participate in the survey;
� Deciding whether to stratify the population in order to ensure that the
views of the full range of customer groups are accurately captured;
� Selecting the right sample size (refer appendix 3) – if the customer group
is small, say less than 200 people, it may be more appropriate to undertake
a census of the entire population.
FCS, Transport, and Homeswest were able to demonstrate that they had
taken a rigorous approach to developing a list of their customers, and then
taking a sample of people from this listing. This increases the likelihood
that their survey results will be based on a sample of individuals that
accurately matches the characteristics of their customer group.
Other agencies have experienced varying degrees of difficulty in taking a
sample of their customers, due mainly to factors such as:
� confusion concerning whether the sampling unit was an agency or
individuals within these same agencies; and
� the impossibility of constructing a customer database.
As an illustrative example of these difficulties, the customer group for the
OMI’s survey was defined as the peak community organisations that it has
the most dealings with on a regular basis. From a list of 27 organisations,
OMI provided a consultant with a list of 50 names, being individuals
associated with these 27 host organisations. The consultant then attempted
to undertake a census of the list of 50 names provided. This methodology
raises questions with regard to how OMI selected the 27 agencies and the 50
individuals to be surveyed, and whether these individuals are able to
accurately represent the complete range of views in the 27 host organisations.
Using customer surveys to report performancein the Western Australian public sector
19
Survey Accuracy and Reporting
Data Collection Methods
The nature of the data collection method to be used (i.e. mailed questionnaire,
face to face interviews, telephone interviews) will be determined by the
nature of the topics to be covered in the survey, time and cost considerations,
and the characteristics of the agency’s customers. For example, face to face
interviews are a good technique for minimising communication problems
with people from a non-English speaking background, but they are also quite
expensive. Mailed questionnaires are relatively inexpensive, but require
respondents to have well developed literacy skills.
FCS’s customers are interviewed by telephone if possible, with non-
respondents and non-telephone users being followed up with a mailed
questionnaire. FCS is aware of the potential for differing data collection
methods to produce differing survey results and has undertaken some analysis
to compare the ratings provided by clients via telephone interviews versus
mailed questionnaire. Interpreters are made available, if required, to assist
respondents from a non-English speaking background with the interview.
Homeswest collected half its survey data using face to face interviews and
the other half using mailed questionnaires. Homeswest advise that research
by public housing bodies interstate has shown that using these two different
data collection methods with rental tenants has only a minimal impact on
the survey’s results. However, the face to face interviews are not targeted at
any particular geographic area or any particular client group. Given that
Homeswest has experienced problems with response rates in the Kimberley
region and with Aboriginal people, it may be worth considering targeting
these two groups for face to face interviews exclusively.
Response Rates
A survey’s response rate is given by the percentage of sampled individuals
who actually complete the survey. For example, if 100 people are sent a
mailed questionnaire and 75 of them actually complete and return this
questionnaire, the survey’s response rate is 75 per cent. It is quite common
for individuals who do not respond to a survey to differ markedly from those
who do. The survey’s results can then be biased and misleading as a result
of this difference.
Auditor General WA
LISTEN AND LEARN»»
20
Survey Accuracy and Reporting
There is no absolute answer to the question ‘What is a minimum acceptable
response rate?’ It depends upon who is being surveyed and what methodology
is being used. As a basic guide, a response rate of at least 50 per cent is generally
considered adequate for analysis and reporting, while a response rate of 70 per
cent or more is considered to be very good. It should be borne in mind that this
is only an approximate guide and has no statistical basis. A demonstrated lack
of response bias is more important than a high response rate.4
Response rates for agency surveys varied from a high of 69 per cent for
DOLG, through to a low of 11.5 per cent for DOLA. Low response rates (less
than 50 per cent) were experienced by five of the seven agencies examined.
These low response rates suggest that the survey’s findings are likely to be
biased, but the exact direction and size of this bias cannot be known without
undertaking further analysis. Good survey practice suggests that such analysis
should be routinely conducted for all surveys.5 FCS was the only agency that
attempted to address this issue.
Agencies therefore need to do better in this aspect of their surveys, both by
raising their response rates and by routinely analysing the extent of bias in
their surveys arising from non-response.
Sampling Precision
Survey findings obtained from a sampled group of customers are subject to
a degree of sampling variability or error. That is, these findings may differ
from the figures that would have been obtained if a different sample had
been selected or if a census of the entire population had been undertaken. A
large sample is more likely than a small sample to produce results that
closely resemble those that would have been obtained from a census. The
difference between survey estimates and expected census results can be
measured statistically using the concept of a ‘standard error’.
4 Adapted from: Babbie, E. 1990, Survey Research Methods, Wadsworth Publishing, California.
5 See: United States General Accounting Office, 1993, Developing and Using Questionnaires,author, Washington, D.C.; and Groves, R. 1989, Survey Errors and Survey Costs, John Wiley,New York.
Using customer surveys to report performancein the Western Australian public sector
21
Survey Accuracy and Reporting
For example, if a survey found that 60 per cent of agency customers were
satisfied with an error of ± 20 per cent at a 90 per cent confidence level, we
could be 90 per cent certain that in the population of agency customers,
somewhere between 40–80 per cent of all customers were satisfied. In this
case there would be uncertainty as to whether an apparently large increase
in customer satisfaction from 35 per cent to 60 per cent over three years
really did reflect a ‘true improvement’ or simply ‘sampling variability’6 This
is illustrated in Figure 5.
Figure 5: The Effect of Sampling Variability
This figure illustrates how sampling variability can be mistaken for a true change in the
population, particularly for a short time series with a large sampling error.Source: OAG
For the purpose of analysing trends over time, particularly when measuring
small changes, agencies should consider obtaining samples large enough to
keep their sampling errors to less than ± 5 per cent with a high degree of
confidence. See Appendix 3 for a guide to selecting samples of an appropriate
size.
6 The above two paragraphs have been adapted from Australian Bureau of Statistics (Victoria),1993, An Introduction to Sample Surveys.
100
90
80
70
60
50
40
30
20
10
01994–95 1995–96 1996–97 1997–98
p
pp
lower limits ofsampling error
p
p
True situation: 50 per cent satisfied no trend
Apparent upward trend due to sampling variability
upper limits ofsampling error
Auditor General WA
LISTEN AND LEARN»»
22
Survey Accuracy and Reporting
Sampling precision ranged from a low of ± 3 per cent for FCS and Homeswest
(excellent) to a high of ± 17 per cent for DOLA. Of the seven agencies
examined, four had error rates greater than ± 5 per cent. Agencies with a
high level of error in their surveys were generally unaware of the significance
of sampling error. Improving the precision of agency surveys can be achieved
by taking a larger sample of agency clients, and reducing levels of non-
response.
Validity of Measurement
Validity concerns the extent to which a survey is really measuring what it is
supposed to measure. An example of a lack of measurement validity would
be to give an Australian I.Q. test to a non-English speaking migrant. The
test results will be a reflection of their limited familiarity with English, not
their level of intelligence.
Developing a valid survey for performance reporting purposes requires agency
objectives that are clear, specific and measurable.
Establishing a survey’s validity is an important matter, and should be a
standard part of planning and developing a survey. If it is not done it can
cause systematic survey errors that lead to results that are consistently
biased in one direction.
None of the seven agencies formally assessed the validity of their surveys.
The most basic approach for establishing the validity of a survey is to ask a
knowledgable person, ‘Does this survey question appear to be a good measure
of this concept or outcome objective? Does it look right, does it make
sense?’. In the absence of formal validity testing and assessment by the
seven agencies, the examination applied this most basic of tests to determine
the face validity of the seven surveys. This test raised issues regarding two
of the surveys examined.
The relevant effectiveness indicators addressed by FCS’s Customer Perception
survey are: ‘Proportion of family and community support customers who
have increased their knowledge and skills; and extent to which family and
community support customers develop their own solutions and
Using customer surveys to report performancein the Western Australian public sector
23
Survey Accuracy and Reporting
independence’. The relevant questions in FCS’s survey are ‘On your last
contact, how much did you learn that was useful to you?’; and ‘How confident
are you that you will be able to handle a similar situation in the future?’.
FCS realises that these two questions are likely to only be a partial measure
of the relevant indicators, and they have managed this issue by reporting a
suite of indicators from several different surveys to counter the limitations
of using a single survey question in isolation.
OMI’s 1997 objective was “to promote a harmonious community and equitable
access for people from diverse cultural, linguistic and religious backgrounds
by assisting the State Government in the development of its policies and by
supporting community initiatives”. This objective is not clear, specific and
measurable, making it difficult to develop valid survey questions.
The OMI objective could suggest that their customers are individual persons
in the community. However, OMI view their customers as being a number of
peak community organisations but not individual persons. The validity of
OMI’s performance indicators would therefore be enhanced if its objective
stated the outcomes OMI is seeking to achieve in relation to these
organisations.
OMI’s survey asks selected representatives from these peak community
organisations to rate the impact of OMI’s activities on the wider community.
Given that these representatives are unlikely to have adequate data on this
impact, and that OMI acts as a resource centre for these peak community
organisations (the respondent’s host organisation), this raises further doubts
as to the validity of the survey.
Reliability of Measurement
Reliability concerns the extent to which a survey would give similar results
if it were given more than once to the same group of people. Reliability is a
function of random measurement error in the data. Determining a survey’s
reliability is another basic part of survey planning and development.
An example of reliability is weighing oneself on the bathroom scales. If you
got on your bathroom scale and it read 70 kilograms, you got off and on
again and it read 75 kg, repeated the process and it read 65 kg, your scale
Auditor General WA
LISTEN AND LEARN»»
24
Survey Accuracy and Reporting
would not be a very reliable measure. Thus a lack of survey reliability is of
concern as it leads to inconsistent averaged findings, particularly for groups
with fewer than 60–80 respondents.
None of the seven agencies attempted to assess the reliability of their survey,
which is of concern for the agencies with a small number of respondents
(i.e. DOLA, DOLG, and OMI). This means that the performance indicator
figures reported for these agencies are likely to contain significant amounts
of random measurement error, which means that their findings are more
likely to be inaccurate.
Survey Reporting and Record Keeping
In addition to technical rigour, two further issues of importance are:
� Standard of reporting – clearly presenting the survey’s results and
indicating any significant limitations in the methodology; and
� Record keeping – the availability of supporting documentation.
Standard of Reporting
Survey data should be objectively analysed using standard scientific methods.
Agencies need to explain their results, particularly any unusual or unexpected
findings. Results should be interpreted with the appropriate level of precision,
and expressed with the proper degree of caution about the conclusions drawn.
Performance data needs to be presented in a manner and form that enables
agency staff and external audiences to assess the current level of performance
and whether it is improving or worsening, and to what extent. Charts and
graphs present findings clearly, but should include a description of how the
survey was conducted, explanatory notes about the findings, and a discussion
of the limitations of the survey’s methodology.
Using customer surveys to report performancein the Western Australian public sector
25
Survey Accuracy and Reporting
A related issue is the matter of independence of data collection and analysis.
The use of an independent consultant to conduct an agency’s customer survey
adds to the perceived impartiality of the survey, and also encourages
customers to respond to the survey in a frank manner. This is especially
relevant for those agencies that:
� regulate or fund their customers; or
� provide services upon which their customers are highly dependent.
All of the seven agencies examined analysed their survey data using simple
descriptive statistics such as percentages or average scores. Most agencies
then presented their findings in a simple tabular format (FCS, DOLA,
Transport, OMI, and RIA), with or without supporting explanatory notes.
Homeswest made use of coloured graphics to present their findings, while
DOLA relied on lengthy written descriptions of their findings.
Several of the agencies examined made errors in the analysis and/or
interpretation of their performance indicators based upon survey findings
(DOLA, DOLG, Homeswest, and OMI). In particular, agencies had problems
in interpreting changes over time in satisfaction ratings due to ignoring the
impact of sampling precision. This seems to be a problem of agency expertise,
rather than a matter of deliberate misrepresentation.
Many agencies made no attempt to interpret the meaning of their performance
indicator data, simply presenting the data and leaving the reader to make
their own conclusions about agency performance (FCS, Transport, OMI and
RIA). There was a general lack of explanatory notes to inform the reader
about any limitations in the methodology used for the survey. No agency
reported their findings along with the relevant standard error figure.
The reported performance indicators often contained little additional
information to assist readers assess agency performance. For example they
did not include a standard or a benchmark against which to compare agency
performance. They also did not generally include explanatory notes to inform
the reader about the survey’s technical limitations.
Agencies such as FCS and Transport were well aware of the importance of
the independence of survey data collection and analysis, while other agencies
such as DOLG need to further consider this aspect of their customer survey.
Auditor General WA
LISTEN AND LEARN»»
26
Survey Accuracy and Reporting
Record Keeping
Agencies and/or their consultant were able to provide the examination team
with copies of all requested documents, generally with minimal delay. DOLA
was noteworthy for using a consultant to check on their audit trail as part of
a review of their customer survey process.
Agencies were generally unsure how long the paper copies of survey forms
should be kept for audit purposes, although the Market Research Society of
Australia recommends hard copies be kept two years. Managers were unaware
as to whether these records were covered by a Retention and Disposal Schedule
under the Library Board Act of Western Australia 1951. This Schedule is the
means by which agencies can legitimately dispose of public records or transfer
records of permanent evidential value to archives.7
Recommendations
Agencies should:
�� ensure that their surveys are conducted in a scientifically rigorous manner
so as to minimise all types of survey error;
�� present their performance indicators in conjunction with relevant
supporting information such as comparative benchmarks and the survey’s
technical limitations; and,
�� assess the cost-effectiveness of undertaking customer surveys,
particularly those agencies with small budgets and a large number of
customers.
7 Refer to OAG Report No 6 of 1996 title ‘For the Public Record’ for further information aboutpublic sector records management.
Using customer surveys to report performancein the Western Australian public sector
27
Appendix 1: Key Steps inConducting Customer Surveys
Designing, implementing, and utilising customer surveys is best thought of
as an ongoing iterative process characterised by the following key steps:8
1. Determine who the agency’s customers are
� Identify customers for all relevant agency products and services,
e.g. internal and external clients, direct and indirect clients
2. Determine agency objectives for seeking customer feedback
� Determine why customer feedback information is required, e.g. to
assist with the planning of a new service, to further improve current
services, for accountability purposes, etc.
� Identify specific information needs in relation to each purpose, and
in relation to each client group
� Clarify how the agency will use the survey information once it has
been obtained
3. Develop the agency’s measurement strategy
� Develop institutional structures and plans to clarify who will have
responsibility for managing the customer survey, who will decide
upon the utilisation of any findings, and who will then implement
these actions
� Clarify how frequently the agency will need to obtain feedback from
its customers, e.g. on an ongoing basis versus periodically
� Identify the most appropriate methodology for data collection given
the agency’s information needs, budgetary resources, and the
characteristics of the client group
� Calculate an appropriate sample size for the survey based upon the
degree of precision required in the survey’s findings
� Consider obtaining input from technical specialists to assist with
the development and pilot testing of the survey
8 Adapted from: Treasury Board of Canada, 1992, Measuring Client Satisfaction.
Auditor General WA
LISTEN AND LEARN»»
28
Appendix 1: Key Steps in Conducting Customer Surveys
4. Gather and analyse information
� Adopt standard scientifically valid methodologies to minimise errors,
ethical concerns, and other potential problems
� Rigorously analyse the data and make appropriate comparisons, e.g.
over time, over geographical areas, over different customer groups,
against service standards, against industry benchmarks
� Interpret the survey’s results with the appropriate level of precision,
and express the proper degree of caution about the conclusions that
can be drawn from the results
� Attempt to explain any unexpected or unusual results
� Document procedures followed in the course of the survey. Edit and
archive the data/findings to allow independent confirmation of the
results
5. Report and utilise the survey’s results
� Ensure that published findings are consistent with the survey’s results
� Share the results with service delivery staff, agency managers, and
customer groups
� Set agency priorities, and develop strategies to address service areas
in need of improvement
� Set service standards that are both challenging and realistic
� Establish monitoring procedures to assess performance
improvements over time
6. Review agency’s measurement practices
� Are internal agency lines of responsibility and accountability clear?
� Are agency practices technically sound?
� Is the agency’s survey producing useful information in a cost-
effective manner?
Using customer surveys to report performancein the Western Australian public sector
29
Appendix 2: Judging the Qualityof Survey Research: A Checklist
1. Does the survey report contain a list of specific issues or questions the
survey is intended to address?
2. Do the research questions posed by the investigators appropriately and
adequately address the topic of the survey?
3. Are the research questions posed by the investigators well organised
and well structured?
4. Does the report identify the target population to which generalisation
was desired?
5. Does the report describe the sampling frame used and the rationale for
its use?
6. Does the report indicate a close match between the target population
and the operational population?
7. Does the report describe the sampling procedure used? Were probability
procedures used?
8. Are non-response rates reported for the entire survey, and for individual
questions?
9. Were non-response rates low enough to avoid substantial bias errors?
10. Are any analyses of potential sampling bias reported (including patterns
of non-response)?
11. Are sample sizes sufficient to avoid substantial sampling error? Are
standard errors of estimate reported?
12. Is the primary mode of data collection (mail questionnaire, face to face
interviews, telephone interviews) consistent with the objectives,
complexity, and operational population of the survey?
13. Is a copy of the survey instrument provided in the survey report?
14. Was the instrument thoroughly pre-tested as a part of its development?
15. Are instructions for completing the survey clear and unambiguous?
16. Are questions in the instrument clear and unambiguous?
17. Do questions in the instrument encourage the respondent’s honesty in
admitting a lack of knowledge or uncertainty?
Auditor General WA
LISTEN AND LEARN»»
30
Appendix 2: Judging the Quality of Survey Research: A Checklist
18. Are questions in the instrument free from obvious bias, slanting or
loading?
19. Are the survey questions ordered appropriately?
20. Was the survey consistent with ethical research practices? Have the
issues of anonymity and confidentiality been handled adequately?
21. Does the survey report contain a description of field and data management
procedures?
22. Were these field/data management procedures adequate and appropriate?
Is it likely that major sources of bias error have been avoided?
23. Are the data analyses clearly described?
24. Are data analyses appropriate to the purposes of the survey? Were all
relevant statistical assumptions satisfied?
25. Did the survey provide answers to the research questions posed by the
investigators?
26. Are the researcher’s conclusions sound, or are alternative interpretations
of the findings equally plausible?
27. Does the survey report contain a description of any deviations from the
survey’s implementation plan, and the likely impact of these deviations?
28. Does the survey report contain an analysis of the quality of the survey,
including its reliability and validity?9
9 See: Jaeger, R. 1988, Complementary Methods for Research in Education, American EducationalResearch Association, Washington DC; and Babbie, E. 1990, Survey Research Methods,Wadsworth, California.
Using customer surveys to report performancein the Western Australian public sector
31
Appendix 3: A Guide to SampleSize Requirements10
Required sample sizes at the 95 per cent confidence level to yield a given
level of sampling error (standard error of a proportion), assuming a proportion
of incidence in the population of 0.5.
Required final sample size for an error of:
± 3% ± 5% ± 10%
Population Size
50 48 45 33
100 92 80 49
250 203 152 70
500 341 217 81
750 441 254 85
1 000 516 278 88
2 500 748 333 93
5 000 880 357 94
10 000 964 370 95
25 000 1 023 378 96
50 000 1 045 381 96
100 000 1 056 383 96
1 000 000 1 066 384 96
Infinitely large 1 067 385 96
Note: The figures presented in this table refer to the number of completed,
useable questionnaires that the agency needs to get back, not the starting
sample size. For example, if 278 respondents are required from a population
of 1 000 to yield a precision of ± 5 per cent and the expected response rate is
60 per cent, the starting sample size would be 464 people (i.e. 278 x 100/60).
10 Adapted from: Salant, P. and Dillman, D. 1994, How to Conduct Your Own Surveys, John Wiley& Sons, New York.
Auditor General WA
LISTEN AND LEARN»»
32
Performance Examination Reports
Tabled
1996
Improving Road Safety May 1, 1996
The Internet and Public Sector Agencies June 19, 1996
Under Wraps! – Performance Indicators of Western Australian Hospitals August 28, 1996
Guarding the Gate – Physical Access Security Management within the
Western Australian Public Sector September 24, 1996
For the Public Record – Managing the Public Sector’s Records October 16, 1996
Learning the Lessons – Financial Management in Government Schools October 30, 1996
Order in the Court – Management of the Magistrates’ Court November 12, 1996
1997
On Display – Public Exhibitions at: The Perth Zoo, The WA Museum and
The Art Gallery of WA April 9, 1997
Bus Reform – Competition Reform of Transperth Bus Services June 25, 1997
Get Better Soon – The Management of Sickness Absence in the WA Public Sector August 27, 1997
Waiting for Justice – Bail and Prisoners in Remand October 15, 1997
Public Sector Performance Report 1997 November 13, 1997
Private Care for Public Patients – The Joondalup Health Campus November 25, 1997
1998
Selecting the Right Gear – The Funding Facility for the Western Australian
Government’s Light Vehicle Fleet May 20, 1998
Weighing up the Marketplace – The Ministry of Fair Trading June 17, 1998
On request these reports may be made available in an alternate format for those with visual impairment.