White paper - Let’s talk about youDocument Transcript
Research excellenceLet’s talk about youThe key to getting accurate, actionable ideas from market researchis to help respondents tell the truth about themselvesJan Hofmeyr, Chief Researcher, Behaviour Change White paper
Let’s talk about youThe key to getting accurate, actionable ideas from marketresearch is to help respondents tell the truth about themselvesResearch cannot hope to deliver precise plans for growth unless it builds a preciseunderstanding of individuals. It seems an obvious point to make, but it’s one withwhich brand tracking surveys, in particular, struggle to come to terms. About the authorIt’s an important but often ignored truth that survey data can be valid at aggregatelevel and yet wrong about individual people. This comes about through mutually Jan Hofmeyr is TNS’s leading expert on consumer behaviour, with a career spanningcompensating error, the likelihood that for everyone who says that they used over 20 years advising many of the world’sa particular brand but didn’t there is somebody else who says that they didn’t use best-known brands.the particular brand when they actually did. Thanks to mutually compensating He invented ConversionModel whilst workingerror, brand tracking can continue to deliver topline aggregate figures that are for the Customer Equity Company (acquiredroughly correct, even if individual data is seriously compromised. by TNS in 2000), recognising a need for better quality insight on consumer motivations.This possibility ought to keep researchers awake at night, since the In 2010, following a period of five years at Synovate, Jan returned to TNS to continuerecommendations that we make about a brand’s potential and actual consumers his work in this field, updating thedepend upon individual truth and the way that each individual’s answers correlate ConversionModel methodology to cementtogether, rather than aggregate data. We require respondent-level validity – and all its position as the world’s leading measure oftoo often researchers do not push hard enough to achieve this. consumer commitment. Prior to working in market research, Jan wasTNS is developing a new approach to brand tracking that focuses clearly on a senior political advisor for the African Nationalrespondent-level validity and the adaptations that are required to achieve this. Congress during and after the first democratic elections in South Africa. He is the co-authorPut simply, we care about whether our respondents tell us the truth about their (with Butch Rice) of Commitment Led Marketinglikely actions – and we are developing new techniques to make it easier for them and the author of numerous, award-winningto do so. papers on brand equity.This approach underpins the TNS ConversionModel, a global brand tracking studythat has been built around the techniques and principles outlined in this paper.The problems with ‘big ticket’ tracking – and how to solve themThe structure of today’s brand tracking surveys makes it difficult to get toindividual truth. It’s worth pointing out early that this problem doesn’t result fromrespondents hiding the truth from us – it’s a case of survey techniques making itfrustratingly difficult for them to provide us with meaningful information. The fourmain barriers that surveys put in the way of respondents telling the truth are: Brand tracking surveys are far longer than they need to be – and asking too many irrelevant and unnecessary questions has dire consequences for data quality They ask the wrong questions and often in the wrong way, using techniques and measures that are simplistic and known to lead to false information They ask questions at the wrong time, exposing results to the fallibility of human memory and failing to deliver the real-time insights that marketers need They fail to apply enough intelligence to the analysis of data, with the result that clients do not get the information they need in time. White paper 2
Let’s talk about youFocusing on respondent-level validity, weeding out questions that don’t deliver itand developing new ways of asking questions that do, are the keys to deliveringnimbler, more effective and more actionable trackers.The TNS approach leverages available technologies and techniques to create inthe moment surveys that are able to access consumers’ instinctive responses;to apply intelligence to these surveys to ensure relevant, responsive questionsand actionable data; and to link this to data-streams such as economicconditions, sales information, marketing spend and digital behaviour to providea holistic view.Flexible, adaptive, faster: cutting survey lengthOur core proposition is that current big-budget trackers can be collapsed into oneefficient, flexible, and adaptive data stream. This data stream can be integratedwith others in a single-source approach.This new data-stream is built upon an intelligence-driven survey populated bylearning algorithms that cut survey length, drive up validity, and automaticallycreate category and brand knowledge over time. The core survey is deliberatelyand genuinely ‘thin’: it takes no more than two to three minutes to complete.We do not consider ten minute surveys to be ‘thin’.The new system will not be modular. It will be adaptive. There is a difference.Modular systems are like a layer-cake that adds survey chunks to a core usingdumb criteria. The key to an adaptive system is that it learns from the respondentduring the survey what should be asked next. In other words, adaptive surveysgo where the respondent wants to go. Modular systems force respondents togo where the researcher thinks they need to go. The adaptive system becomesthe ‘conversation with consumers’, part of a tracking approach that integratesattitudes and behavior through the creation of single-source data.Smarter thinking about which questions to askBuilding intelligence into the tracker system is the key to making all aspects ofa survey more relevant to the respondent and so overcoming the problem ofboredom whilst improving data quality. At the same time, an intelligent trackersystem can reduce costs through saving time – and enabling multiple surveytrackers to be consolidated into one.The task of creating an intelligent tracker system begins with applying a rigorousapproach to sample size and covariance, asking smart questions about howmany respondents need to answer each question, and how many questions eachindividual respondent needs to be asked. White paper 3
Let’s talk about youLeveraging a database of the standard deviations of variables provides us with anopportunity to reduce the number of irrelevant questions we ask by confirminghow large the sample size for each question actually needs to be. If you knowthat a question has a small standard deviation, then you can reduce the size ofthe sample you need for that question – which in turn means you can select arandom sub-sample to answer the question and allow the rest to skip through,reducing survey time.Let’s look at a quick example of how this could work: we know from sometwenty years of doing brand equity studies, that committed users of a brand tendto be homogeneous in the image they have of that and other brands. As a result,their answers to attribute association questions hardly vary. This means that youdon’t have to force them all to respond to the attribute association question:measure a few and you will know what the rest would have said. You can allowthese few to answer the question for the others.Our approach to leveraging covariance is similar. In this case, we use a databaseof established covariance to create a skipping, interview-shortening process thatis tailored to a particular respondent. We know that some questions are highlyinter-correlated. The three questions most commonly used in loyalty studies,satisfaction, purchase intention and recommendation, happen to be greatexamples. If you know that a person’s answers to a particular question will behighly correlated with answers they have already given, then you can skip thatquestion. Again, survey length could be cut without loss of information.Intelligent pathways through heuristicsApplying heuristic (or self-educating) principles can help us to extend the idea oflearning from respondents and create intelligent pathways in surveys. The keyhere lies in adapting each survey in real-time, to reflect the way that the particularrespondent makes decisions. Once again, the key focus here is on achievingrespondent-level validity.We know, for example, that people who are uninvolved in a category behave inone of two ways: either they develop shallow habits in which they stick to onebrand because they can’t be bothered to think about what to use; or they careso little about brand choice that they’re influenced more by point-of-purchase/consumption phenomena than by brand. White paper 4
Let’s talk about youThere is very little point to asking people in this frame of mind an attributeassociation question because the results are highly predictable: their answers willbe sparse and restricted to the brand they buy by habit. What’s more pertinentis to ask them questions that measure their response to ‘in the moment’ brandstimuli: discounts, special displays, prominence on the shelves, and so on.The challenge is to develop the right, engaging virtual environments to dothis effectively.By contrast, people who are committed to a brand are less influenced by ‘in themoment’ phenomena. They could skip these kinds of questions. A more complexpathway could be built using attitudinal equity configurations.On the whole, we aren’t fans of attribute association questions. However, thesesame heuristic principles provide an opportunity to make simple changes that candramatically improve the correlation of attribute responses with actual sales atrespondent level. The four key changes that TNS has identified in this area are:1. Allow respondents to select the attributes that are most relevant to them before asking them to associate attributes with brands.2. Restrict the scope of the associations to the sub-set of brands that are relevant to each respondent.3. Replace the free form association question i.e. respondents only tick positive associations; with a binary form i.e. respondents answer ‘yes-no’.4. For driver analysis: transform the results into ‘share of mentions’ for each brand and attribute at respondent level.‘Share of mentions’ is a simple transformation: instead of using values of‘0, 1’ when performing driver analysis, use values that are based on the shareof mentions the brand gets for each attribute. So, for example, if a personassociates two brands with an attribute, then the values for that attribute for thatrespondent in a driver analysis would be ‘0.5, 0.5’.A binary response format results in much greater response stability andreliability1. And reducing both the attribute and brand lists ensures that relevantinformation is collected and reduces the tedium associated with the classicalattribute association task. White paper 5
Let’s talk about youMobile capabilities: asking questions at the right timeMobile capabilities have a vital role to play in improving brand tracking surveys,since they have the potential to solve the problem of fallible human memory andto deliver fast-turnaround results. Leveraging mobile technology enables us tokick-start all surveys at the appropriate moment.TNS has almost a decade’s experience of creating short-term panels in whichpanelists record their daily buying and consuming as it happens. These ‘In themoment’ mobile purchase and consumption diaries are less subject to memoryerrors; they can be used to collect ambient point-of-purchase or consumptioninformation; and they provide a single-source of attitude and behavioral data.The events covered by the diaries could include drinking an alcoholic ornon-alcoholic beverage, the complex and varied stages involved in planninga car purchase, exposure to an ad for the first time, and a huge range ofother occasions.In our experience, people create records of each event within an hour.In categories involving many events, people send up to eight records a day.If a panelist hasn’t sent anything for six hours since ‘waking’, they’re senta reminder. Although each record looks long, it typically takes three minutesor less; and 70 percent of the panelists complete their diaries.We validate overall consumption using external sources such as Kantar WorldPanel, Nielsen, IRI. Respondent-level validation involves setting flags tomeasure response consistency.Mobile as listening deviceTNS has developed an app called MobileBehave that leverages the mobile’spotential as a listening device for all manner of brand-consumer communicationstaking place through the mobile channel. MobileBehave data builds over timeas people become relaxed about the fact that the app is on their phone. It hasmultiple uses: A source of passive (i.e. ‘listening’) mobile behavioural data A single-source of ‘listening’ data combined with ‘in the moment’ data Can be used to recruit panelists for non-mobile ‘listening’ Enables the building of online communities based on revealed interests Can be used as a sample source for instant surveys Becomes the basis for creating causal models of behaviour over time White paper 6
Let’s talk about youAsking the right questions in the right wayWe have always known that there is a gap between what people say in surveysand what they actually do. Thanks to contemporary neuroscience, we know thevarious reasons why the gap exists – and this can help us to fix it. By basingquestions around the parts of the brain that become active when brandattachment forms, we are able to fix common mistakes that our industry makeswhen it comes to communications modeling.There are many ways in which current approaches to measuring and modelingcommunications impacts ignore reality. Here’s a short list: Over-reliance on memory to establish communications exposure. As a result, modeled effectiveness coefficients are faulty; Failure to take account of what’s already in the brain about brands, in particular pre-existing brand commitments; Failure to model communications effects holistically (for example, in the context of other information that affects brand image like competitor communications); Overly narrow focus on characteristics of the advert at the expense of measuring impacts on the person.Neuroscientists tell us that there are genuine differences between the way thebrain reacts to favoured and non-favoured brands2. All forms of exposure tobrands create neural tracks over time that link favoured brands to personal goalsand values. Favoured brands then show up in complex networks in the brain thatinclude the areas that guide decision-making, and those that deal with affectivememories. By ‘affective’ we mean more than ‘emotional’. Affective refers tofeelings with deep personal meaning.Brand connections are built in multiple ways, most notably, through direct brandexperience, through endorsements by others – most notably experts, friends, andwhat can best be called ‘the mass of humanity’, and through own-brand andcompetitor messaging.A holistic approach to communications measurement and modeling can help.This is based on the single-source approach to information that we describedearlier. We looked at the options for collecting information about brand use ina way that overcomes problems of memory; and gives access to context-relevant information. To model this information more effectively, we need to White paper 7
Let’s talk about youlink it to metrics that reflect the neural connections that form around brands.There are two of these metrics: first, a quantified measure of ‘affective impact’(remembering here that ‘affective’ means more than just ‘emotional’); second,open-ended questions to create verbatims that can measure ‘affective content’. Affective impactWe can combine these two approaches to measure the affective impact of a By affective impact we mean the extent to which a piece of communications links thecommunications piece. First, we ask a simple open-ended question: what does brand to experiences that have a deeperthe advert bring to mind; in what ways has the brand become part of your life personal meaning. It’s about placing the brandand who you are? Second, we explore the sequence of emotions: The lesson in the context of personal goals and values.of most of the current emotional measurement is that ‘positive’ is good. Yetadvertising is storytelling. And we know from great storytelling that it’s themanagement of an emotional sequence that really matters. So, for example,‘negative’ need not be bad if it’s followed by ‘positive’. Examples might include:‘problem – resolution’; ‘surprise – delight’; ‘threat – victory’. And so on. Affective content By affective content we mean articulatingThe next step is to relate this view of the affective content of communications deeper motivations in words. Qualitative researchers use projective techniques and richto how people actually make decisions in the market. TNS has developed a two- stimulus material to try to link instinct andpillar model of brand equity that gets to the heart of what actually drives sales. intuition to words – so that a person can say what’s more deeply in their mind.Theories of choice based on the idea that what people do is the result ofpsychological preferences combined with situational factors, probably pre-dateancient philosophers. In modern times, they show up in the distinction betweenattitudinal and behavioral loyalty. Usually, attitudinally loyal people will buy thebrand to which they’re loyal if they can. But sometimes market (i.e. situational)factors nudge people towards an alternative; or even prevent people from buyingthe brand they want. And sometimes, when people have no strong first choice,market factors tip the scales in favour of one brand rather than another.We’ve used this simple framework for understanding brand sales for manyyears3. In our framework, sales are a function of a brand’s ‘Power in the mind’(attitudinal equity) and ‘Power in the market’ (market factors, brand presence,market equity). These two dependent variables anchor our analysis of brandequity and sales drivers.We’ve recently updated these measures using surveys on behaviour paneldata. We can show that our new metrics outperform other similar metrics atrespondent level4, and we expect to continue to improve them in the monthsto come. White paper 8
Let’s talk about youPower in the mindPower in the mind is a respondent-level measure of brand attachment thatcorrelates better than similar measures with the real (panel validated) shareof spend that each brand gets from that person. And it achieves this witha significant reduction in survey length.We measure a brand’s power in the mind in two steps. First, we identify thebrands that are relevant to each respondent. Second, we ask for two ratings foreach relevant brand. The two dimensions that have to be measured are brandperformance and brand involvement. We use scales derived from the most up-to-date neuroimaging survey measures5, and an algorithm underpinned by ouroriginal theories of brand relationship6 to calculate from these a ‘one number’measure of attitudinal brand equity. This correlates better with a person’s shareof consumption in panel data than other comparable metrics.We use this number as a dependent variable for equity modeling; and also tocreate equity segments and a brand health ‘ladder’. Because we leverage heuristicprinciples7, this measure typically takes less than 30 seconds of survey time yetresults in brand health scores for every respondent for every category and brandin a study. Continued improvements will further enhance accuracy over thecoming 12 months.Power in the marketPower in the market is a respondent level measure of the market factors thatdrive consumer behaviour. It offers a vital improvement in taking into accountthe law of double jeopardy. According to this law, bigger brands gain in twoways over smaller brands: they have more users, and their users tend to usethem more.There are important problems with the law of double jeopardy, most notablywith its assumption that individual brand preferences are stationary over time8.Nevertheless, the law highlights the benefits of scale that accrue to big brands.These drive incremental sales for locally dominant brands; and create marketbarriers for smaller brands.There are a number of important ways in which brands can pull marketing leversto drive sales: distribution, point-of-sale visibility, greater affordability, getting theproduct mix right (packs and variants), purchaser preference (leveraging the factthat the person who buys isn’t always the end-user), and creatinglocal monopolies. White paper 9
Let’s talk about youLike our ‘power in the mind’ measure, our ‘power in the market’ measureleverages heuristic principles to cut survey time while increasing the validity of theresults. It typically takes less than 30 seconds and gives granular, respondent-levelinformation about the market drivers of sales for brands.Put the two together and you have a powerful system of core metrics that takesless than a minute of survey time to deliver equity and market information aboutall brands at respondent level.Gamification: a better way to ask questionsThe gamification methods pioneered by Puleston and others can help us to solvethe problems of length, irrelevance, and boredom; and tap more effectivelyinto less conscious motivations by engaging the parts of the brain that are notactivated by classical, word-driven surveys.Even when gaming methods aren’t very game-like, tests show that respondentsare much more engaged by these devices than they are by classical surveymethods. Mobile can play an important contributory role in applying gamificationmore widely, since mobile devices provide a channel for incorporating thisapproach into face-to-face interviews.Intelligent, pro-active systems for ‘just-in-time’ informationBesides making surveys shorter, more relevant and more responsive, intelligentsystems can also be used to deliver actionable information and insight morepro-actively. We are skeptical about the use of ‘early warning’ systems that relyon single trend analysis such as moving averages, Bollinger bands, and the like.Our reason is: a single trend doesn’t contain enough information to provideintelligent alerts. We set more store by the analysis of anomalous gaps acrosstrends. By analyzing multiple trends gaps, we should be able to identify thatstresses are developing in the system. These stresses can be a powerful indicatorof opportunity or threat.An example of a potentially anomalous gap would be when sales are under-supported by equity. Over twenty years of brand health modeling, we’ve seensuch under-support often enough to know that it’s a sign that the brand’s saleswill come under pressure. Similarly, when equity exceeds sales, it’s a sign ofpotential opportunity. White paper 10
Let’s talk about youHow can we build anomalous gaps into analytical systems? A database ofrelationships between the key variables in data streams can help to establishthe key anomalous gap values between such data points as marketing spend,attitudinal equity and sales. We can then build intelligence into the trackingsystem by automating the discovery of values in the data. This is a three-foldprocess: automating the collection of instances such as turning points in marketshare; populating a database with relevant instances that can trigger analysis;and automating the updating process so that the the data-stream deliversnew instances.Putting it all together: survey architecture for intelligent adaptive trackingThe TNS ConversionModel has been redeveloped along the principles set down inthis paper, to deliver respondent-level validity within an adaptive tracking approachand reduced survey time. This approach enables the model to deconstruct marketshare precisely and provide clear guidance on opportunities for brand growth.The core ConversionModel study will now form the basis of future trackingthat is able to leverage an adaptive, heuristic architecture to ensure fewer, morerelevant questions and respondent-level validity around individual behaviour.ConversionModel takes into account that people care about some decisions morethan others – and that this prioritisation varies by individual as well as by category.In further developing the ConversionModel, and applying a new approachto tracking more generally, we will develop survey architecture along thefollowing lines:‘In the moment’ tracking activationBy taking measurement close to behavioral events, we can measure three keythings, with no more than three to four minutes for each event, diminishing overtime as machine learning kicks in: What people actually buy; Basic context information: where were they, what were they doing; Complete brand equity and market barrier information at a situational level White paper 11
Let’s talk about youThe development of smart mobile devices and gamification survey techniques,will improve compliance and validity of responses. Among respondents fromwhom we get permission to install the MobileBehave app, we will enablea three-fold integration of event-based behavior, situational brand equity,and mobile ‘listening’ over time. Analysis and deliveryWe apply two levels of near real-time reporting and analysis: Basic: feeds back trend information e.g. buying, consuming; that can be disaggregated according to ‘who’, ‘where’, ‘when’, ‘for what purpose’ Analytic: feeds back information that requires algorithms based on trend changes and, more importantly, gaps across trendsExamples of basic feedback include ongoing, real-time trend information aboutwhat people are buying and consuming, where, and why. Basic feedback alsoincludes real-time information about category/brand situational equities andsituational drivers.The analytic components of the system will be programmed to learn fromexperience, identifying when positive or negative equity stresses develop. As anexample: when equity is high and consumption is low, this suggests a failureof marketing. When equity is low and consumption is high, this suggests thatconsumption is unsupported by psychological demand.Intelligent, adaptive follow up surveysThe ‘in the moment’ survey process is the thin core. It gives us basic purchase andconsumption information coupled to situational brand equities and market barrierinformation. As the diary builds, fewer questions will need to be asked. Questionsabout situational equities, for example, only need to be asked once.The follow-up survey happens after a set time period that could be daily, weeklyor monthly. Respondents will be channeled into questions that are relevant tothe way they make decisions, with different subsets for people with strong brandpreferences and people without, for example. We will know this from our analysisof patterns of attitudes and behavior revealed in the diary survey.By creating live adaptive questioning that is tailored to each respondent,we can integrate big-ticket trackers into one system that combines allrelevant measurement areas: actual behaviour, brand equity, market factors,communications influences, path-to-purchase, and point-of-consumption. White paper 12
Let’s talk about youThe future of tracking conversationsA lot is spoken about the need for brands to engage consumers in meaningfuldialogue. Tracking surveys are no exception. The measures outlined in this paper You mayleverage what we know about consumers, markets and the human brain in order be interested in...to conduct conversations that are relevant and meaningful for each respondent.It makes for a more stimulating and enjoyable experience for those involved in The trouble with tracking by Jan Hofmeyr >our surveys. And it makes for more valid, holistic and actionable information for ConversionModel >our clients. Commitment Economy >The early results of this new approach can be seen in the insights delivered bythe 2012 TNS ConversionModel. However, evolving trackers to reflect consumerdecision-making more closely is an ongoing process. We are passionate aboutdelivering questions and answers that are valid for individual respondents in oursurveys. And we will continue to explore and apply new techniques in order todo so. White paper 13
Let’s talk about youAbout TNSTNS advises clients on specific growth strategies around new market entry,innovation, brand switching and stakeholder management, based on long-established expertise and market-leading solutions. With a presence in over Sources80 countries, TNS has more conversations with the world’s consumers thananyone else and understands individual human behaviours and attitudes 1: Dolnicar, Sara, Bettina Grun, and Friedrich Leisch (2011) ‘Quick, simple, and reliable: Forceacross every cultural, economic and political region of the world. binary survey questions,’ International Journal of Research in Marketing, 53:2TNS is part of Kantar, one of the world’s largest insight, information and 2: Plassman, Hilke, Peter Kenning, and Dieterconsultancy groups. Ahlert (2007), ‘Why Companies Should Make Their Customers Happy: The Neural CorrelatesPlease visit www.tnsglobal.com for more information. of Customer Loyalty,’ Advances in Consumer Research, 34:2Get in touch 3: Hofmeyr, Jan H. and Butch Rice (2000), Commitment-Led Marketing, John WileyIf you would like to talk to us about anything you have read in this report, and Sons, Chichesterplease get in touch via enquiries@tnsglobal.com or via Twitter @tns_global 4: Hofmeyr, Jan, Victoria Goodall, Marting Bongers, and Paul Holtzman (2008), ‘A new measure of brand attitudinal equity based on the Zipf distribution,’ International Journal of Marketing Research, 50:2; Keiningham, Timothy L., Lerzan Aksoy, Alexander Buoye, and Bruce Cooil (2011), ‘Customer Loyalty isn’t Enough. Grow your Share of Wallet,’ Harvard Business Review, October 5: Reimann, Martin, Requel Castano, Judith Zaikowsky, and Antione Bechara (2011), ‘How we relate to brands: Psychological and Neurophysiological insights into Consumer- Brand Relationships,’ Journal of Consumer Psychology, (forthcoming) 6: Hofmeyr, Jan H. and Butch Rice (2000), Commitment-Led Marketing, John Wiley and Sons, Chichester 7: Gigerenzer, Gerd, Peter M. Todd, ABC Research Group (2000), Simple Heuristics That Make Us Smart, Oxford University Press, USA. 8: Hofmeyr, Jan, Victoria Goodall, Marting Bongers, and Paul Holtzman (2008), ‘A new measure of brand attitudinal equity based on the Zipf distribution,’ International Journal of Marketing Research, 50:2; White paper 14