MFC Spotlight Note

8
February 2004

Counting (On) Your Prospective Clients: Guiding Principles in Measuring Micronance Client Satisfaction and Loyalty
Katarzyna Pawlak and Dorota Szubert1

Many Micronance Institutions (MFIs) in the Central and Eastern Europe and the New Independent States (CEE and the NIS) region have started to measure client satisfaction and loyalty in order to become more client-oriented and improve performance. This information is very useful as MFIs are working in increasingly competitive environments. As measuring client satisfaction and loyalty is a relatively new undertaking, MFI practitioners usually have to learn from their own mistakes. This Spotlight Note is based lessons learnt drawn by the Micronance Centre for CEE and the NIS2 and its partners from the client assessment work conducted under Imp-Act3 and SEEP Practitioner Learning Programme (PLP)4 projects. We intend to facilitate this new task for MFI marketing departments by presenting some principles in measuring micronance client satisfaction and loyalty using a quantitative survey.5 In this paper we present some general marketing research rules as well as new insights specic for the micronance industry. We focus on general concepts of satisfaction and loyalty and discuss the potential results one can obtain. It is important to note, however, that the list of principles and rules is not meant to be exhaustive.

Client satisfaction vs loyalty

Dont fall into the satisfaction trap. It appears from the research studies that 60 to 80 per cent of clients who switched their supplier would have declared they were very satised just before leaving (Reichheld 1996). Client satisfaction is indicative of the degree to which services and products satisfy clients preferences. However, satisfaction is not a static idea and it changes as soon as a client nds a better deal and along over-time- raising client expectations. It is client loyalty that counts. Satisfaction doesnt indicate a clients emotional connectedness, commitment to the institution, which means that a satised client is not necessarily a loyal client. A loyal clients commitment can be veried through client behaviour and attitude. A committed client recommends the institutions to relatives or friends, decides to travel a little bit further than necessary to use one service and not another or pays a higher interest rate. At the same time a loyal client will believe that their chosen MFI is the best from all the institutions on the market.

What purpose does satisfaction and loyalty informaion serve?

Satisfaction and loyalty information can help you improve your programme and develop loyaltybuilding strategies. Satisfaction is necessary but not sufcient for loyalty. However its measurement is helpful in detecting dissatisfaction factors that could be disadvantageous and costly for your MFI in the future. It will also be helpful in building greater commitment on the part of your clients. Satisfaction and loyalty studies can enable you to streamline your operations. An appropriately conducted satisfaction and loyalty study will provide you with information helpful to prioritise your

Katarzyna Pawlak and Dorota Szubert work in Research Unit at the Micronance Centre for Central and Eastern Europe and the New Independent States (MFC). More information about MFC research work can be found at www.mfc.org.pl/research. The authors are very grateful to Graham A. N. Wright for his comments on the rst draft. 2 More information on MFC can be found at www.mfc.org.pl. The brief is based on experience and lessons learnt from the research work conducted in cooperation with partner MFIs: Partner (Bosnia and Herzegovina), Inicjatywa Mikro (Poland), BosVita (Bosnia and Herzegovina), MDF Kamurj (Armenia). 3 . More information on ImpAct - the Ford Foundation sponsored global action-research programme designed to improve the quality of micronance services and their impact on poverty by developing impact assessment systems can be found at http://www.imp-act.org/. More information on the regional work is available at www.mfc.org.pl/research. 4 More information on the USAID sponsored SEEP Practitioner Learning Programme can be found at www.seepnetwork.org. More information on the regional work is available at www.mfc.org.pl/research. 5 For more information and references on other quantitative methods and techniques see the annex in the end. As the paper focuses on measurement of satisfaction and loyalty, the issue of understanding with regards to these concepts is not discussed here. A reader may nd more information on the latter in documents available at the AIMS http://www.usaidmicro.org/pubs/aims/ and MicroSave www.MicroSave.org websites.

www.mfc.org.pl/research/

Counting (On) Your Prospective Clients

improvement programme allowing you to achieve better results in a cost effective way leading to better performance. You may use the results to prompt product revision and development, improve your operations and/or inform your strategic and tactical decisions. Satisfaction and loyalty information does not replace data on impact. Although, keeping satised and loyal client is necessary for achieving impact, the fact that they are satised and loyal does not allow to claim it. This is because met client preferences dont necessarily reect the fullment of their development needs.

Measurement basics

General satisfaction question will not tell you much! As general satisfaction level is a consequence of the client satisfaction from particular areas of institutional performance, general how are you satised question will not tell you much although will be helpful in further analysis. If you want to evaluate the forest, look at the trees. It is important to evaluate an institutions performance with regard to different areas of its operation i.e. product, service, delivery, etc. The 8Ps concept can be very helpful for unpacking the micronance product and selecting its detailed characteristics for evaluation.6 Looking at this information will allow you not only to see which areas are better or worse evaluated, but will be also helpful in constructing aggregated indices and calculating overall satisfaction rating that is more meaningful than a standing alone question on general satisfaction. Not all performance areas are equally important. There are usually only a few key aspects which improvement will lead to increased client satisfaction. The others either increase satisfaction when present but dont harm it when absent. Generally, the more signicant the area is found to be, the more its improvement inuences general satisfaction. Segmentation is key for effective operational use of the satisfaction and loyalty study results. General satisfaction and loyalty rating will not tell you much how to improve and where. Segmentation of satisfaction and loyalty data will help you to determine where improvement is most warranted and where it will offer greatest return. Thanks to the segmentation you will be able to identify prospective groups among your clients and the priority areas for the improvement action related to the preferences of those groups. How much do you need to improve to be the best? You might know your overall rating, but do you know how well you are really performing, if you lack benchmarks for comparison purposes? The problem in micronance is that it is still new and in many context clients are not nancially educated and have limited knowledge of competitors services. However, this situation is changing and there is increasing scope for comparing clients responses to the services of different service providers. Loyalty client attitude and behaviour. To measure client loyalty we need to look at the measures reecting his/her behaviour such as depth, length and breadth of loyalty, as well as referrals and intent to repurchase (Churchill, Halpern 2001). Loyalty is a consequence of different aspects, so as in the case of satisfaction, a general rating is required that is based on various aspects of behaviour and attitude. 3-D Behaviour: depth, breadth, length of Loyalty. The loyalty depth refers to exclusitivity and is reected in a share of purchase - the degree to which a customer uses your institution for all her nancial services needs. This may vary according to an institutional type; for example, in credit only institutions, the information requested could be whether a client has any other debts outstanding, or the degree to which a clients needs for additional money have been satised.7 The longevity can be measured by the average number of years a client has used your services. The breadth of loyalty reects the effectiveness of your product cross selling strategies. The more products your client purchases, the more loyal he/she is likely to be. This measure, however, is not useful in the institutions that dont offer a wider range of product. Are clients really loyal or do they have no choice? A client may stay with your institution and buy only your products without feeling any loyalty. This could be because they have no choice, are used to your service or perceive switching to another service provider to be too costly. Although they demonstrate all the behaviour of a loyal client, this doesnt mean they are loyal and that tomorrow their behaviour will not change. Only attitude and behaviour are a true indication of client loyalty.

The 8Ps stand for people, place, price, promotion, positioning, physical evidence, process, product. More information about the concept can be found in any marketing related literature (e.g. Kotler 2000); more information about ncept application in micronance can be found in the MicroSaves Market Research for Micronance Handbook. 7 This information on using other credit sources, although very valuable not only for this purpose, is difcult to obtain as clients have disincentives to provide accurate information in this regard. That is why its interpretation should be taken with care and cases identied should be considered as underestimated.

www.mfc.org.pl/research/

MFC Spotlight Note #8

Secondary behaviour counts. Referrals, endorsements, and spreading the word are examples of secondary behaviour that indicate customer loyalty. What is more, they are the most effective promotional tools in micronance.8 The way to measure secondary behaviour is to ask about the number of referrals made in a certain period or whether they would be willing to recommend your organization to their relatives and friends. This indicator is also a powerful proxy for high client satisfaction.9 Intent to repurchase is an additional sign of loyalty and can be used as an insight into a clients future behaviour. However, since this information does not guarantee the clients behaviour, it should be treated with care, especially as clients may feel that admitting they dont want to repurchase may jeopardize future opportunities to do so.

Key issues in the study design

Clarify your objectives rst. There are many approaches to measuring and understanding client satisfaction and loyalty so it is important that you choose the one that best suits your needs. First, be clear about what information you are looking for, since this will determine the approach you undertake and the methodology you choose. If you would like to better understand Design and Implementation issues related to satisfaction and loyalty, Process Steps: or have a specic loyalty or satisfactionrelated problem to explore, consider using qualitative methods. 1. Clarify your objectives If you want representative information 2. Get your sample right about the degree of your clients satisfaction 3. Design the questionnaire and loyalty, about your performance in terms of adjustment of products and 4. Pre-test the questionnaire services to the clients preferences and for 5. Collect data identifying areas for improvement, choose a quantitative survey. 6. Analyze data Last but not least, if you want quick 7. Report and use ndings operational statistics on an on-going basis you will need an easy tool for monitoring satisfaction and loyalty. A few well-selected indicators should help you maintain executional speed and remain affordable, although providing somewhat supercial information. However, satisfaction monitoring should not replace periodic satisfaction and loyalty assessments that generate in-depth analysis that can direct improvement efforts on a broad scale.

Box A.

8 According to the results of the MFC work in the region 40% - 70% of clients choose their MFI on the basis of the word-of-mouth. In East Africa, the MicroSave reports on around 60-80% clients. 9 For an interesting discussion on this issue please see The One Number You Need to Grow, Frederick F. Reichheld, Harvard Business Review, December 2003.

www.mfc.org.pl/research/

Counting (On) Your Prospective Clients

A mix of methods will provide you with a complete picture. If you use quantitative techniques, dont forget the great value-added of qualitative ones. The survey will give you quantied, representative data on your clients satisfaction and loyalty. However, to use the data effectively you will need to understand what drives it and why certain things matter or dont matter to your clients. To maximize use of the data it is necessary to complement quantitative work with qualitative information and vice versa. Qualitative methods will help you to get your questionnaire right, quantitative methods will give you representative statistics. Then probe the meaning of the statistics using qualitative methods.10 Draw your sample right! General tips on sampling11 To obtain results relevant to all your clients you do not necessarily need to interview them all.12 If your sample is correctly drawn, you will be able to use statistical tests to extrapolate your results onto the entire population of your clients. Dene your sampling framework well. You need to decide who is of interest to you and what are the characteristics of the population to be surveyed. Get a representative sample. There are different sampling methods. Random sampling guarantees that every client has an equal chance of being selected. Any differences between the sample and the population are only due to chance and not to selection bias; There are three main types of random sampling you can use: simple, stratied, and systematic sampling. Determine the right size. The bigger the sample, the more sophisticated analysis you may conduct and the more sure you can be that the results from the sample truly reect the situation in the whole population. But bear in mind, the bigger the sample, the higher the cost of the study13.

Questionnaire design basics

Questionnaire is just a tool. That is why it is important to clarify your information needs and assess capacities to satisfy those needs, and focus on your research objective throughout the process of the questionnaire design. Which type of survey to choose? There are different methods of collecting data and administering a questionnaire (mail survey, telephone survey, face to face interview, self-completion form, etc.14) and they affect the information you obtain. In this note, we will focus on a face to face survey. Stick to the KISS rule. Keep It Short and Simple to obtain precise information. Translate your objectives into the questions and avoid including questions that are not directly linked to the information you need. Ask yourself what you will do with the information from each question. If you cannot give yourself a satisfactory answer, leave it out. Make the most of existing information. The most common mistake during the questionnaire design phase is neglecting information which already exits. Using secondary data15 will aid your questionnaire design and help you formulate the right questions. Get your questions right.16 Follow some principal rules on designing questions Avoid threatening questions or statements. A respondent should not have the impression he/she is being evaluated. Use neutral questions or statements. Dont ask for too many things at once ask one question per issue. Avoid emotionally charged words or leading questions.

10

This paper focus solely on presenting a quantitative survey. However, there is a need to remember that each tool, method, technique is able to provide you with somewhat limited results. The mix of tools allows to get a wider picture of an issue. You may use focus group discussions, participatory rapid appraisal tools, individual interviews to unpack some of the complex issues underlying descriptions used in the survey. E.g. Product Attribute Ranking can be very helpful in obtain the most important characteristics of your service in a client-language as well as provide you with the understanding why they are less or more important. Relative preference ranking and Financial Sector Trend Analysis will help you identify your competitors and understand what and why make your and their competitive advantage. More information on PRA tools and other qualitative methods can be found in MicroSave Marketing Research for Micronance Toolkit and at www.MicroSave.org . 11 For more information on sampling please see the references as well as the annex in the end. 12 However, in the case of smaller programmes (or not so frequent occurrence of the phenomenon) it is also possible to interview all the clients so called census sample. 13 There are various guidelines or calculators you may use to determine your sample size taking into account your objectives, population characteristics and capacities available in the relevant literature and accessible in the Internet (e.g. at http://www.surveysystem.com/sscalc.htm). 14 See the annex for more information. 15 Secondary data is information that already exists somewhere, having been collected for another purpose. There are various sources of secondary data i.e. loan applications, client passbooks, MIS information, any previously conducted research, etc. (source: Market Research for Micronance, MicroSave). 16 The same concerns the answer categories both good phrasing and order are important (see the annex for details).

www.mfc.org.pl/research/

MFC Spotlight Note #8

Dene words clearly - make sure your questions can accurately tell you what you want to learn. The way you phrase a question can inuence the answers you get. Try to make sure the wording is clear to the respondent and does not favour one choice of answer over another. Do not use abbreviations or unfamiliar words this is particularly important for setting product attributes. In micronance, there is a lot of jargon which micronance clients may not understand. Make sure that all terms used are well understood either by pre-testing the questions or/and discussion with frontline staff. Very often using descriptions to present certain product characteristics can help clients understand what we mean. To avoid automatic responses to questions, try to use different formats or mix the questions throughout the questionnaire (see tips on questioning order).

Watch out -score or rating scale questions are a particular problem. There are different kinds of scales, and their type and sensitivity can inuence the results. You can choose from dichotomous and multi-point scale. The fewer the points, the easier it will be for a respondent to provide you with an answer. However, it will be less able to capture differences of opinion among respondents. Another decision you need to make about a scale is whether it will be of even or odd number of points. Having even number will force people to choose between negative and positive options and help you avoid the central tendency that is very natural for people to choose. However, with a good interviewer you should be able to get the real answer from the respondent and still leave the middle option for those that are really indifferent to certain issues. The questions order matters. The order of questions and answer choices can encourage people to complete your survey. It can also affect its results. Here are some general rules that will help you to use the question order in the most effective manner: Start with easy questions. These will encourage people to continue the survey and help build rapport with the interviewer. Group together questions on the same topic to make it easier for the respondent to answer.17 Move more important questions to the beginning of a series or rotate the order to counteract the habituation effect. People tend to think more about their answers to questions earlier in the series and so give more accurate answers. With later questions they tend to give the same answers.18 Leave difcult or sensitive questions for the end. Any rapport that has been built up will make it more likely that people will answer these questions. If people quit at that point anyway, at least they will have answered most of your questions.19 As a rule, move from general to specic rst ask about a general issue than move to details. Move from spontaneous to supported questions ask an open-ended question rst to get spontaneous answers and only later ask about the same issue using closed questions. Follow the logical ow - an appropriate order of questions helps you to reduce certain biases and increases the likelihood of obtaining more accurate answers on sensitive topics. For example, to learn whether your client uses other sources of credit rst ask about the extent to which his/her needs for additional money are satised by your loan and then move on to asking about how he/she provides for this additional money that is not available from your institution.

17 As mentioning something (an idea, an issue, a brand) in one question can make people think of it while they answer a later question (event though, they might not have thought of those issues if they had not been previously mentioned), such grouping also helps to reduce the bias of inuencing the answers of one question by another one. 18 Another way to reduce this problem is to ask only a short series of similar questions at a particular point in the questionnaire. Then ask one or more different kinds of questions, and then another short series. 19 For more information on open and closed ended questions please see the annex.

www.mfc.org.pl/research/

Counting (On) Your Prospective Clients

Dont forget about demographic information. They will help you to identify people of similar characteristics. The best timing for this information (age, gender, income, education, etc.) is at the end of the questionnaire. By then the interviewer should have built a rapport with the interviewee that will allow honest responses to such personal questions. Exceptions to this rule are any demographic questions that qualify someone to be included in the survey. Obviously, these questions must come near the beginning. Some of the information you may be able to extract from the MIS instead of asking the respondent, in order to save time. Pre-testing the questionnaire will reduce the costs of the survey and increase data quality. During the pretest, make sure the questionnaire responds to the research objectives and check it is properly designed, in terms of language, length, order of questions by examining the reactions of your experimental respondents.

Box B. Example of demographic

information
Individual characteristics i.e.: Sex Age Place of residence Education Etc. Describing household situation: Number of household members Number of children Number of dependants in the household Indicators for income Etc. Describing experience with the institution: Time with the programme Number of loans taken / cycle Cross - usage Size for each loan taken Etc. Describing the business situation Type of business Place of operation Registration status Number of workers Etc.

How to Ensure Quality data

How the interviews are conducted is crucial to the quality of data. Interviewers need to be carefully chosen and properly trained.20 At the very beginning, consider the characteristics required of interviewers and analyse whether your institution has the capacity to supply the interviewers. Then decide whether to use external or internal interviewers. External interviewers are less familiar with the topic, but respondents might be more honest with them as they are not representatives of the institution. Internal interviewers will know the institution and its clients, but their links with the institution may inuence the way respondents answer the questions. Surprisingly, the costs of using external or internal interviewers may not be dissimilar. Training is a must as it is a tool increasing data credibility through building interviewers buyin and education. First of all, the research objectives have to be explained in-depth to interviewers and you must ensure they understand them clearly. Each question has to be elucidated in the context of research objectives otherwise there is a possibility that some questions will not be properly posed by interviewers. There should be common understanding of aspects so that interviewers make sure that answers provided are adequate to the researched topic. Interviewers should practice posing questions and dealing with potential problems. Visual aids and instructions. As the human memory is elusive, it is reasonable to provide the interviewers with hard copies of instructions. For face-to-face interviews it is highly advisable to present possible answers to the respondent using response cards. Such a visual aid makes the interview more eventful and makes the interviewers job easier. How useful are quality checks? You need to check with the randomly chosen respondents whether the interviewers actually visited them to conduct the survey to make sure that the answers provided are from the clients themselves, rather than the interviewers.21 In addition, look through questionnaires for exceptions and unusual answers, which can be discussed with the interviewer and veried. Avoiding mistakes in the future. After eldwork, discuss the process of collecting data with the interviewers. Use their knowledge to identify mistakes and ways of preventing them in the future.

20 21

Thanks to the good choice of interviewers you reduce likelihood of biasing situations. According to the ESOMAR guidelines, minimum 10% of questionnaires should be re-interviewed for data quality reasons. ESOMAR, founded in 1948, is an organization unites 4000 members in 100 countries, both users and providers of opinion and marketing research. The society facilitates the exchange of experiences between suppliers and users of research in order to optimise the integration of research results into the decision making process. ESOMARs mission is to promote the use of Opinion and Market Research for improving decision making in business and society worldwide. More information on ESOMAR can be found at www.esomar.org.

www.mfc.org.pl/research/

MFC Spotlight Note #8

What to do with the data?

Get the most from your data. If your survey design was driven by your research questions and you ensured the data quality, the data analysis will be rewarding.23 It is wise to prepare the data analysis framework before the research implementation so that you will get all the data you need (Table 1). There are several tools you may use for the analysis, starting with Excel and ending with more specialized statistical software, such as SPSS, Statistica, Quantum. How satised are your clients? Questions on overall satisfaction only indicate the rst reaction to the product or service and does not provide you with accurate information on overall satisfaction.24 One of the ways to obtain meaningful Table 1. Example of a Data Analysis Framework information on Research objective Survey questions Analyses overall satisfaction it is to construct Frequencies * To determine the 1. About overall satisfaction Descriptive statistics overall and detailed a satisfaction index 2. About detailed satisfaction Satisfaction index satisfaction level (see chart 2).25 How loyal are your clients? As with the satisfaction, Questions covering all aspects of loyalty: we can look at Breath different aspects Depth of loyalty.26 To nd Frequencies To determine the level Length * Descriptive statistics out about client Secondary behaviour: intent of loyalty Loyalty indices loyalty we need to to repurchase, likelihood to take into account recommend Type of loyalty (emotional its different attitude) dimensions, using loyalty indices. Descriptive statistics In the presented To identify groups of * (means and standard example, we have The same questions as above lower satisfaction and deviations) constructed a loyalty Pivot tables secondary loyalty index. The positive Factor analysis * To segment clients answers provided The same questions as above Cluster analysis based on loyalty and by respondents satisfaction to the questions To segment * Proling on client intent to clients based on the Factor analysis MIS repurchase, past protability and target Cluster analysis recommendation characteristics and likelihood to recommend were * To benchmark 1. About overall satisfaction Frequencies summarized and satisfaction and loyalty 2. About detailed satisfaction Descriptive statistics allocated to the performance against 3. Questions on relative performance Satisfaction indices other players on the dened category of the competition Correlation analysis market22 of secondary loyalty. To get a loyalty rating it is important to construct an index that will be a combination of all the mentioned variables.
* To identify strengths and weaknesses 1. 2. The same questions as above The question about importance of aspects Frequencies Descriptive statistics Correlation analysis

The sample should cover your clients and your competitors clients. The satisfaction and loyalty questionnaire has to be the same for the two groups. For more information on statistical analysis, see the annex and relevant references in the end. As we discussed earlier to get a satisfaction rating score we need to look at the satisfaction with detailed aspects of performance areas. The way to do it is to calculate a satisfaction index using satisfaction results from the detailed characteristics evaluated. 25 There are different ways to construct an index. In the case presented on chart 2, points were allocated only for top answers allocated to each aspect and then summed. 4 categories were dened : not satised (0 max.value/4), poorly satised (max.value/4-max.value/2), satised (max.value/2 *max.value), very satised (* max.value max.value). 26 As we discussed, loyalty is considered through such dimensions like breadth, depth, width and secondary behaviour.
23 24

22

www.mfc.org.pl/research/

Counting (On) Your Prospective Clients

How well are you performing? To assess this, you will need to compare your results to those of your competitors, using benchmarking analysis as it was described earlier on. How do clients evaluate your performance in different areas? Using top answers (the percentage of denitely satised in each category) to the question on client satisfaction with various aspects of your performance and calculating the adjustment threshold27 will allow you to identify two main categories: programme elements that clients are satised with and those that they are less satised with.

Clients expectations. Clients do not attach equal value to all aspects there are aspects of high importance and aspects that no attention is paid to. High satisfaction with the former leads to higher overall client satisfaction and in case if improvements are needed, these are the attributes the change of which will bring the highest return. There are two ways to identify important aspects. First approach (chart 6): is to rank the aspects of performance by asking clients about the importance of each. Using a multi-scale question we are able to calculate frequencies and nd out which aspects were rated as the most important ones (top answers). The calculated adjustment threshold28 enables the division of different attributes into categories of less importance and more importance for clients. The limitation of this approach is that a questionnaire becomes too long and thus tiring for a respondent, and consequently the quality of results may decrease.

27

28

The adjustment threshold is a sum of all denitely good answers divided by the number of all aspects. The adjustment threshold is calculated as the mean value of percentage of denitely important answers for all aspects.

www.mfc.org.pl/research/

MFC Spotlight Note #8

Second approach (chart 7): work out the level of each aspects inuence on the overall satisfaction level through a correlation analysis. Using the general satisfaction question and a multi-scale question about satisfaction with detailed aspects, we run a correlation analysis to identify the inuence of each aspect on the overall satisfaction. The limitation of the second approach is that answers have to be diversied (different respondents have to choose different answers) in order to differentiate between the aspects relative importance. Performance vs. client expectations. To nd out how this evaluation of institutional performance in certain aspects of its operations (client satisfaction) relates to client expectations, it is necessary to compare satisfaction and importance results. This will enable you to identify priority areas for an organization to improve. To do this a multiple-scale question on satisfaction detailing various aspects of the programme, as well as a question on client expectations, should be used. The ndings can be presented as in chart 8.29 The crossing point divides the chart into 4 areas: 1. Improve (low satisfaction and high inuence) an institution should prioritise those aspects, as they are important for clients. 2. Maintain or improve (high satisfaction and high inuence) an institution is doing well in these areas, so further improvements will not have much inuence on overall client satisfaction. 3. No action needed (high satisfaction and low inuence) this area is highly evaluated and not so important. If there is a need to look for cost savings, then look here.

29 We use axis X to present inuence on overall satisfaction of each aspect (or a declared importance of each aspect), and axis Y to present the mean value (or percentage of top answers) of satisfaction.

www.mfc.org.pl/research/

Counting (On) Your Prospective Clients

4. Less important (low satisfaction and low inuence) this is an area of low priority for improvement. Even though the aspects are not highly evaluated, they are not important. Institutions can ignore those aspects and focus on priority areas. Short-cuts to segmentation. You should already have an idea of client overall satisfaction and loyalty rating; which aspects are better or worse evaluated, less and more important to clients and where improvements are needed. However, it is important to consider who will benet from these improvements, since MFI services serve different people with different needs. Simple descriptive statistics can be used to get this information - mean and standard deviation can be calculated for each aspect using the multi-point question on satisfaction with detailed aspects. This analysis will help you to understand how consistent opinions are across your client base. The ndings can be presented as illustrated in chart 10. The point of crossing of the axes divides the chart into four areas:30 1. H o m o g e n o u s satisfaction (high mean of satisfaction with the aspect and low standard deviation) aspects in this area are consistently highly evaluated by a majority of clients; 2. H o m o g e n o u s dissatisfaction (low mean and low standard deviation) aspects in this area are consistently given a low by a majority of clients; 3. Dispersed satisfaction (high mean and high standard deviation) even though a majority of clients are satised with the aspects, this opinion is not consistent. There are some respondents less satised with the aspect; 4. Dispersed dissatisfaction (low mean and high standard deviation)- even though a majority of clients are not satised with the aspects, this opinion is not consistent. There are some respondents satised with the aspect. Identifying aspects of dispersed satisfaction or dissatisfaction gives you an indication that further client segmentation is needed. However, further exploration of segments will be possible only if the sample is big enough (about 300). The analyses are also supportive in verifying information on the priority areas to improve. For example, when you analyse the previously presented chart 9, you nd out that the reactions to suggestions and complaints is an aspect that doesnt require improvement. However, when you look at chart 10, this information is characterized by dispersed satisfaction. This tells us that there are some groups dissatised even though overall, the majority of respondents are positive. Being equipped with this information, you may identify who those people are. Client proles. To identify more or less satised groups, you may use pivot tables to study the characteristics differentiating the groups. In such an analysis, you may notice e.g. that people who take smaller loans are generally less satised with ofce location and loan ofcers professionalism than those taking medium size and higher loans. Further in-depth analysis can provide you with additional information upon which you may decide to change your priority areas for improvement. As we can see, analysis of the entire database will provide you with a rather general, supercial picture. To get more precise information you will need to conduct analysis in sub-groups of respondents. Such proling as presented above will provide you with more in-depth information on different client preferences. However, it is not powerful enough to identify client groups taking into account several indicators simultaneously. To obtain such information you will need more sophisticated statistical methods for segmentation.

30

In the presented graph the X axis represents standard deviation and the Y axis - mean.

www.mfc.org.pl/research/

10

MFC Spotlight Note #8

Comprehensive segmentation. To be able to further prioritise and better direct your improvement programme, it is important to distinguish groups of clients with relatively similar characteristics, behaviour and attitudes. This involves more sophisticated statistical methods such as cluster or factor analysis. In the example presented, cluster analysis was conducted using three previously calculated indices31: satisfaction, loyalty depth and secondary behaviour. Based on the analysis, four segments were distinguished: 1. Lost clients: relatively low satisfaction, low secondary loyalty, low loyalty depth 2. Searching clients: relatively high satisfaction, low secondary loyalty, high loyalty depth 3. Waiting clients: relatively low satisfaction, high secondary loyalty, high loyalty depth.32 4. Happy faces: relatively high satisfaction, high secondary loyalty, high loyalty depth.

31

For calculating loyalty depth index - % needs satised with loan , using other source of credit, for secondary loyalty index - intent to take a follow on loan, willingness to recommend to others were taken into account. 32 The identied client segments were named in a way to characterize each segment clients potential behaviour based on the information on satisfaction and loyalty. Lost clients are those clients that are dissatised and disloyal so keeping them will be the hardest task; searching clients are those that are relatively satised but also relatively disloyal (they are probably already searching for a better solution) and are likely to exit the programme in the future; waiting clients can be characterized by low satisfaction but high loyalty those clients are bound to our institution today (maybe because they have no alternative or they give us a chance to improve, etc.) even though they are dissatised; the happy faces category includes all satised and loyal clients.

11

www.mfc.org.pl/research/

Counting (On) Your Prospective Clients

For further prioritisation the loyalty-satisfaction segmentation was combined with information on client protability and target status. This is where demographic information or MIS client data is very useful33. Such segmentation allows identication of priority groups that organization wants to focus on. Running previously discussed analyses for those groups helps to identify areas for improvement that will pay off best.

Box C. Main Difculties Faced in Satisfaction and Loyalty

Studies in Micronance
Financial products can be difcult for clients to understand as they are intangible. This particularly causes problems in getting client opinions on product attributes, which is why getting an appropriate list and wording of product aspects is a must before undertaking any client satisfaction and loyalty research (qualitative tools will be of a great help here). Using descriptive phrases to characterize certain aspect can be of help but its effectiveness depends on the environment you work in decide what is the most effective way for your situation. Information on pricing is difcult to obtain. Clients are interested in obtaining servicing at the lowest possible price. That is why this research will not help you much to explore this issue. Results on pricing are neither meaningful nor helpful for decision-making. You will probably face problems with obtaining information on your competition and other sources that your clients use. This is because the majority of clients dont know much about existing alternatives, although this varies according to context. Secondly, they might be afraid of revealing this information to you so as not to be suspected of using other sources of credit, which is as a rule, restricted. You may try to use less detailed questions to compare your performance to your competitors, but it will be difcult to gain more precise answers on the relative advantages and disadvantages. For more in-depth information on the competition you may use qualitative tools. Information on secondary loyalty should be taken with caution. These are only insights, not facts about future client behaviour. It has been realized that client tend to claim that they plan to take a follow-on loan, but then fail to do so. This is probably because they want to jeopardize the possibility of getting loan, should they need one.

Closing Remarks

Is it worth all this effort? With properly designed satisfaction and loyalty research, the results obtained will provide you with sufcient condence to make difcult decisions, which are important to build long-term relationships with your clients. Your improvement programme will be well directed, rewarding your organisation with the loyalty of your priority clients and a highly competitive performance level. Beware of traps, follow the safe path. Go step by step and pay attention to detail. Avoid mistakes from the very beginning. Understand the big picture rst and then design, implement and use your study ndings following a systematic process. Do not play the hero outsource some activities if needed. You dont have to be an expert on every single issue. If there is no capacity in your organization to undertake the required tasks, look for external possibilities. In any case, make sure that you clearly understand your objectives and know all options possible to be able to choose the best one for your organization. Most important of all dont only measure appropriately, but use the results to benet your client and your institution.

33

Protability was estimated based on average loan size, average loan term, effective interest rate and average total cost per client. Target group status was dened reecting the institutional development focus on women and rural clients.

www.mfc.org.pl/research/

12

MFC Spotlight Note #8

References

General Background Churchill F.C., Halpern S.S. (2001), Building Customer Loyalty, Technical Guide No.2, Micro nance Network. Fink A., The Survey Kit (2002), SAGE Publications. Gitomer J. (1998), Customer Satisfaction is Worthless, Customer Loyalty is Priceless, Bard Press. Market Research for Micronance Toolkit (2000), MicroSave Africa, www.MicroSave.org. Reichheld F. (1996), The Loyalty Effect, Harvard Business Scholl Press. Reichheld F. (2003), The One Number You Need to Grow, Harvard Business Review, on-line edition, December issue. Questionnaire design: Walonick D. S. (2000), Survival Statistics , StatPac. Converse J.M., Presser S. (1986), Survey Questions: Handcrafting the Standardized Questionnaire, Sage Publications. Fowler F.J. (1995), Improving Survey Questions: Design and Evaluation, Sage Publications. Data analysis and presentation: Pallant J. (2003), SPSS Survival Manual: A Step by Step Guide to Data Analysis Using SPSS for Windows, Open University Press. Nicol A.A.M., Pexman P.M. (1999), Presenting Your Tables: A Practical Guide for Creating Tables, American Psychological Association. www.micro nancegateway.org/impact/method/iss4_3.htm

13

www.mfc.org.pl/research/

Counting (On) Your Prospective Clients

Annex I. Techniques of satisfaction and loyalty surveys

Technique Face to Face Interview Short description A trained interviewer conducts a structured interview at the respondents place. Answers are noted on the paper and then entered into a data-base. 1. 2. 3. Pros A questionnaire can be longer (up to 30 minutes); A questionnaire can contain open-ended questions; Observing the situation of an interview (body language, etc.) an interviewer is able to say whether a respondent is telling truth; The interviewer can verify that the right person is answering the questions; The response rate is higher than in the case of other techniques; A respondent gives more honest answers as he/she is not under interviewers pressure; The questionnaire can be lled out at the respondents convenience; There is also possibility to verify some answers; A questionnaire can contain open-ended questions; 6. Cons The process of eldwork is time-consuming and consequently costly; A respondent may be biased and over-report his/her situation because of the interviewers presence. Therefore an interviewer has to be very sensitive; A respondent may feel very uncomfortable with the circumstances of the interview.

7.

4.

8.

5.

Self-completed. Interview left by the Interviewer

A trained interviewer gives a respondent a questionnaire to ll out with some instructions on how to do it. A questionnaire is completed either in the interviewers presence or the interviewer comes back to pick it up the same day after some time.

1.

1.

2.

2.

3. 4.

3. 4.

A questionnaire cannot be too long and should be as simple as possible; There may be a lot of missing answers as respondents ll out only those questions they want; Time-consuming and costly process; If the questionnaire is left to be lled out, the danger is that someone other than the designated respondent may ll in the answers;

Phone Interview

A trained interviewer conducts a structured interview by phone. The interview technique is similar to that of the ]face-to-face interview. There are two alternatives for entering answers into the data-base. The rst method, which is more timeconsuming but less vulnerable to mistakes, is where answers are noted on the paper and than entered into a data-base. The second one is to enter answers at once into a data-base during the interview.

1.

2. 3. 4. 5.

A respondent may give more honest answers because he retains more anonymity outside the face-to-face situation; The situation of interviewing is more convenient; Process is time-saving and relatively cheap; Interviewers may be less sensitive to biasing; Another person can monitor the interviews.

1. 2. 3.

4.

Very limited number of open-ended questions; A questionnaire cannot be too long (10-20 minutes); Response rate tends to be lower than in the case of face-to-face interviews; The sample is limited only to those who have phones;

Self-completed. Interview sent by mail

There are several alternatives for conducting this interview: 1. The classic one is that an interview is sent by mail to an interviewer with the return stamp and an envelope and sent back by mail. 2. An interview is sent by mail to an interviewer with the information that the interviewer will come to get a completed questionnaire after some time. 3. The interviewer leaves a respondent a questionnaire to ll out with the return stamp and an envelope. A questionnaire is sent back by mail.

1.

2. 3.

A respondent gives more honest answers as he is not under the interviewers pressure; The situation of interviewing is more convenient; A questionnaire can contain open-ended questions;

1.

2.

3.

4. 5. 6.

A questionnaire cannot be too long and should be as simple as possible; There may be a lot of missing answers as respondents ll out only those questions they want; Another person than the respondent can ll a questionnaire out; There is no possibility to verify the answers; Time-consuming and costly process; This has the lowest response rate (about 5%) of all available techniques;

www.mfc.org.pl/research/

14

MFC Spotlight Note #8

Annex II. Sampling, Questionnaire Design and Analysis Further Tips

1) Basic Sampling methods (Walonick 2000) Simple Random sampling is the purest form of probability sampling. Each member of the population has an equal and known chance of being selected. It is akin to pulling names out of a hat. Essentially, you start with a list of clients in the population and the number of clients you want to sample. The real question is how to select the clients randomly. You could mix the names up in a hat and pull them out until you have the number you want. You also could assign a number to each name on the list and select respondents using the random number function found on calculators or in spreadsheet programmes or by using the random number table located in the back of many statistics textbooks. Systematic sampling is often used instead of random sampling. After the required sample size has been calculated, every Nth record is selected from a list of population members. As long as the list does not contain any hidden order, this sampling method is as good as the random sampling method. In stratied sampling, the researcher rst identies the relevant stratums (subsets of the population that share at least one common characteristic) and their actual representation in the population. Random sampling is then used to select individuals from each stratum until the number of individuals in that stratum is proportional to its frequency in the population. Stratied sampling is often used when one or more of the stratums in the population have a low incidence relative to the other stratums. When a population consists of many subgroups, simple random sampling may not ensure that all subgroups are represented equally in the sample. Say you want to draw a random sample of 100 clients using agricultural loans to represent all of the clients that use this product. You know that 75% of the clients are served by branch A and 25% by branch B. If you picked the names of 100 clients out of a hat, it is possible thatby chance alonemore of the clients in the sample are from branch B. Using stratied sampling, you separate clients into strata, their subgroups, and randomly sample clients from each strata. You select either the same number of clients from each strata (disproportionate stratied sampling) or different numbers proportional to the size of the groups in the population (proportionate stratied sampling). Randomly selecting 50 clients from branch A and 50 from branch B would be an example of disproportionate stratied sampling. Randomly selecting 75 of branch A and 25 of branch B would be an example of proportionate stratied sampling because agricultural clients from branch A make 75% of all agricultural clients.

2) What Kind of Question to Ask? Closed and Open Ended Questions. In open-ended questions there are no predetermined set of responses, and the respondent is free to answer however he/she chooses. Open ended questions are good for soliciting subjective data, capturing unexpected and insightful suggestion, or when the range of responses is not tightly dened. However, those are more difcult and costly to analyse, open to the inuence of the reader and tiring for the interviewer. Generally, we dont use them much in this kind of research most often the questionnaire ends with open ended questions to solicit some unabashed ideas for changes and improvements. Closed ended questions usually take the form of a multiple-choice question. By restricting the answer set, they easily allow to calculate percentages and other hard statistical data over the whole group or over any subgroup of respondents. They make it easier to track opinion over time by administering the same questionnaire to different but similar respondent groups at regular intervals. Finally closed format questions allow you to lter out useless or extreme answers that might occur in an open ended question. Multiple choice questions can allow for one or multiple answers to choose from. While multiple answer questions help you to save paper and avoid a long series of very repetitive question and answer choice lists, they are also a bit harder than the repeated lists for some people to understand.

3) Designing Answer Categories Make sure that the categories are understood by the client this particularly concerns product attributes and characteristics areas you are looking at. Make sure they are in a respondent-friendly language. If necessary use descriptions instead of phrases/words. Answer categories need to be adequate to the question make sure that that the way you ask for answer will really lead a client to provide you with the information you require.

15

www.mfc.org.pl/research/

Counting (On) Your Prospective Clients

Make sure you include all the relevant alternatives as answer choices. Leaving out a choice can give misleading results. Add the category other (specify) in case a respondent provides an answer not included in the provided list or none of the above in case you want to limit the choice to the listed answers. If you expect a single answer, make sure that the choices you have provided do not overlap with one another. Keep a logical or natural order of presenting answer categories. Answer choice order can make individual questions easier or more difcult to answer. Always present agree-disagree choices, positive to negative and excellent to poor scales in that order. When using numeric rating scales higher numbers should indicate a positive answer and low numbers should signify more negative responses. Remember that people tend to pick the last answer when they hear a list of choices read to them and the rst ones when they see them (e.g. on a respondents card). The answer choice order is more likely to affect which choice is picked in the questions about preference or recall or questions with relatively long answer choices that express an idea or opinion. That is why it is good to rotate the answers while reading it to the respondent.

3. Data Analysis General Tips Before starting the analysis make sure that your data is representative of your studied population. If you used the sampling methods that do not reect the variations in clientele, you will need to weight your data to make it representative. Clean the data base by looking at distributions and leaving out the outlying cases, recode the missing values. When analysing questions exclude exceptions. These answers are of interest to you as they may indicate those clients that face some phenomena not characteristic for the majority, however, to obtain meaningful statistical results, rst focus on your typical clientele. To analyse satisfaction and loyalty data you dont need to know any sophisticated analytical techniques. It is enough to be able to use descriptive statistics such as mean, standard deviation, maximum and minimum value, frequencies, quartiles, mode, etc. Simple correlation analysis will be of help. However, to segment your clients you will need to be able to apply more sophisticated methods such as cluster or factor analysis. You also need to learn about statistical tests that you may use to test signicance of your results. Signicance tests enable you to extrapolate with certain likelihood the ndings from the study sample onto the whole client base from which the sample was drawn.34 When you analyse satisfaction questions with multi-point scales, use the frequency results with caution. As discussed earlier people tend to over-report or underestimate certain answers. It is better to be very conservative in your interpretations, which is why it is better to consider only top answers (answers that pinpoint the highest possible evaluation or importance, i.e. denitely satised) as indications for client satisfaction. If the distribution of data is skewed use rather positional statistics such as mode, median and quartiles, as mean and standard deviation will not be very meaningful in such cases Pay attention to the percentage of hard to say or/and missing answers. They should be excluded from analysis. However, if their percentage is big (exceeds 20%) they may indicate the respondent had a problem with answering the question (either didnt feel comfortable with answering the question, could have been dissatised, the question could have been not clear/understood, interview was badly conducted). Statistical signicance. To extrapolate results obtained through your analysis on the sample of clients, you need to test them for statistical signicance. If this is not done, you cannot claim that the results are representative of all your clients As the sample is heterogonous and clients differ in their preferences and evaluation, the results should be analysed in sub-groups in order to capture characteristics of groups of satisfaction and loyalty. When comparing current with previous studies remember about including background information on the internal and external changes that have happened in the meantime as those may have inuence on your client satisfaction and loyalty. Relating results to competitors will help you to incorporate external changes factor to some extent. Keeping record of internal changes is easy and will give you good insights for your interpretation. Be objective and rely on gures and facts rather than your subjective opinion and prejudices while analysing data.

This paper was published with the support of the Open Society Institute.
34

The following test would be of use: for checking differences between groups nominal and ordinal variables (chi square); for checking statistically signicant differences in means - numerical variables (independent samples t test); Spearman correlation analysis to explore correlations between variables; non-parametric tests, etc.

www.mfc.org.pl/research/

16