UNCONSCIOUS CUES AN EXPERIMENT INTO THE UNCONSCIOUS DRIVERS OF BRAND PERCEPTION ON TWITTER
FOREWORD As people live more of their lives online, brands find themselves adapting to new platforms and the new consumer expectations that come with them. But as the marketing industry adopts these new tactics, it’s essential we remember that keeping track of technological change is only a means to an end. Our job is to turn that change to the advantage of our clients’ business. That means taking the time to take stock. If we don’t understand how these new platforms work, we can’t use them effectively. And, of course, some things don’t change. People, for instance. Our behaviour may have taken on a new digital dimension, but our motivations and responses to the world remain as emotional – as human – as ever. We are social animals. Our brains process most of the information they receive unconsciously.
Enabled by technology, but powered by people. This has huge implications for the way people use and respond to digital experiences. Twitter is one of the most important digital experiences for our clients, and for most brands, so naturally it was something we wanted to understand more. How do people perceive brands on Twitter? How do individual features of the platform impact on those perceptions? How can we measure the unconscious? At Isobar, we run towards questions like this, and it has been a privilege to work with Twitter on this unique research experiment. We hope you enjoy reading this report.
Nick Bailey CEO & ECD, Isobar UK
UNCONSCIOUS CUES AN EXPERIMENT INTO THE UNCONSCIOUS DRIVERS OF BRAND PERCEPTION ON TWITTER
4
06 15 28
INTRODUCTION & BACKGROUND
HEADLINE RESULTS
CONCLUSIONS AND IMPLICATIONS
09 16 30
THE ‘UNCONSCIOUS CUES’ PROJECT
THE RESULTS IN DETAIL
FINAL THOUGHTS
5
INTRODUCTION “IT’S WELL ESTABLISHED THAT OUR CHOICES IN REAL WORLD SITUATIONS ARE HEAVILY INFLUENCED BY THE CONTEXT IN WHICH WE MAKE THEM .1” A large body of research, built over many years in the fields of social psychology and behavioural science, has yielded powerful insights which aid our understanding of the decision-making frameworks of people in real life situations . These frameworks are built of unconscious cues, contextual signals and heuristics - mental short cuts that make it easier to process the unconscious decisions we make every day – as well as effortful, conscious decisions. An area of particular interest within this research is ‘Social Proofs’, a mental shortcut people use to navigate unfamiliar real world social situations.2 It determines an appropriate mode of behaviour for a particular social context, and is driven by an assumption that other people have more information or knowledge about a given situation. This mode of thinking is characterised as ‘fast, automatic, frequent, subconscious and stereotypic’. We all use ‘Social Proofs’. If you are in an unfamiliar city, deciding where to eat, but with no prior knowledge of the quality of the restaurants, it’s likely you would look for somewhere busy. The assumption here is that the people inside are local and have more knowledge about the quality of the food than you do. Therefore, if it’s full, it must be good. The important thing is that these evaluation processes enter our minds automatically.
6
‘Social Proofs’ are present in digital experiences, too. Social media platforms allow us to broadcast our lives and choices, but also to be influenced by the lives and choices of others. We are irresistibly drawn to what others are doing, or what we see as popular. But the digital world has generated another, parallel body of behavioural evidence that aids our understanding of decision-making. The discipline of digital User Experience (UX) has built on practices such as A/B and multivariate testing, developed from direct marketing practices. These are vital tools that test the impact of the digital experience on users, and highlight the importance of unconscious and contextual drivers on online behaviour. These two fields of study complement each other. Each has helped brands understand their customers and create better experiences for them. Where social psychology has helped brands understand human behaviour; insights from UX has helped them optimise it. Increasingly, that behaviour takes place on social media platforms, where the user experience is defined by unconscious cues, contextual signals and heuristics, just as it is in the real world.
1 Tversky, Amos and Kahneman, Daniel, “Judgment under uncertainty: Heuristics and biases,” Science, 185 (1974), 1124-1131. 2 Sherif, M. (1935) A study of some social factors in perception, Archives of Psychology, 27(187) 3 Daniel Kahneman (25 October 2011), Thinking, Fast and Slow, Macmillan ISBN 987-1-4299-6935-2
AT ISOBAR, WE BELIEVE THESE MENTAL SHORTCUTS ARE INFLUENCING PEOPLE’S BEHAVIOUR AND WE WANTED TO FIND OUT WHAT THIS MEANS FOR BRANDS. WITH THIS IN MIND, WE HAVE CONDUCTED A SERIES OF EXPERIMENTS TO MEASURE THE IMPACT THAT UNCONSCIOUS INDICATORS HAVE ON BRAND PERCEPTION. SPECIFICALLY, WE HAVE ASSESSED THE INFLUENCE OF DIFFERENT CUES WITHIN SOCIAL MEDIA AND THE DIFFERENCE THEY MAKE TO PEOPLE’S PROPENSITY TO TRUST, RECOMMEND AND PURCHASE BRANDS.
THE PILOT EXPERIMENT: Understanding the value of a social community In 2013 Isobar collaborated with behavioural researchers at the University of Cambridge to design a controlled lab experiment. We wanted to test what impact the size of a brand’s social community had on perception of that brand. We created a fictional furniture brand called Ashwood Furnishings to test whether the size of the community might act as an unconscious cue to generate Social Proof. Each respondent was shown one of 12 different mocked-up brand visuals. The only difference was the size of the brand’s social media following. Respondents were asked to rank the brand in terms of interest, trust, consideration, preference, advocacy and value.
The influence is unconscious and immediate, in the same way that cues in the real world are. The greater the number of fans, the higher the brand perception overall, though the results suggested this effect was subject to diminishing marginal returns (see schematic graph). With this experiment we had seen that very small cues had a significant effect. But the results raised questions. What about social media experiences with multiple unconscious cues? Could we design an experiment that allowed us to test for other forms of social proof?
BRAND PERCEPTION
“THE FINDINGS DEMONSTRATED THAT THE SIZE OF THE COMMUNITY HAS A STATISTICALLY SIGNIFICANT AND POSITIVE IMPACT ON BRAND PERCEPTION”
SIZE OF COMMUNITY
To answer these questions, we wanted to work with Twitter.
4 The academic partners from Cambridge were Joe Gladstone and Jon Jachimowicz 5 The full, detailed methodology and results can be read here – ‘The Science of Social: An Experiment in Influence’
7
THE TWITTER-ISOBAR UNCONSCIOUS CUES PROJECT
Why Twitter? There are two main reasons for wanting to explore a brand’s Twitter presence, one behavioural and one business.
Testing each of these variables in isolation, as well as testing the impact of the relationship between them, would provide a wealth of insight.
From a behavioural perspective, Twitter offers multiple unconscious cues that we could explore to understand the impact of Social Proof. Each cue carries assumptions that could give rise to different user responses.
From a business perspective, brands have moved quickly to embrace Twitter, in many different ways. Some brands use social as a way to more closely engage through entertainment, or by being an active participant in the conversations users have with each other. Others provide customer service, or seek to extend their direct sales function.
1. The number of ‘Followers’: this is the equivalent to the size of community on social media. 2. The number of tweets the brand has sent: how active are they on the platform 3. The number of accounts the brand follows: how connected the brand is, and how much they mirror the activity of a ‘human’ Twitter user 4. The copy contained in the brand’s short biography: the kind of brand it is 5. The impact of a ‘promoted’ stamp on individual tweets: how is the brand perceived as a business or marketing entity
8
It’s not always been clear how to measure the impact of these different approaches, and there has been little rigorous research into how to use social platforms most effectively in pursuit of these ends. Understanding the influence of unconscious cues on different brand perceptions – such as ‘Trust’, ‘Recommendation’ and ‘Purchase Intent’ – and how these differ across different audiences can help businesses to use social platforms most effectively.
METHODOLOGY TO ENSURE THAT RESEARCH WAS DONE TO THE HIGHEST STANDARDS, IT WAS DESIGNED AND RUN IN COLLABORATION WITH THE UNIVERSITY OF CAMBRIDGE. The static panel experiment:
Variables & stimuli:
We recruited an online panel to measure individuals’ perceptions of a brand’s Twitter pages across various conditions. As a ‘between subjects’ experiment, respondents believed they were participating in a market research survey. Respondents were presented with a brand page that was experimentally manipulated to show different values of a range of different indicators. They were asked to rate the page, answering questions that explored their ‘likelihood to buy’ and their ‘brand perception’. These questions used validated scales from academic literature on Consumer Behaviour.
We identified five key variables that we believed had the most potential for influencing brand perception on Twitter.
We decided to use this methodology, adapted from behavioral and experimental economics, as it can help uncover the cues that trigger unconscious changes in brand perception. Often when using traditional research methodologies, such as surveys, respondents sometimes post-rationalise their responses, particularly to questions about external influences on their choices. People often do not readily admit to being influenced by things beyond their control. Moreover, very often people are simply not aware that they are being influenced by certain things, or if they are, they find it very hard to judge the extent to which external cues have an impact on their behaviour.
The tone of voice of the copy in the brand biography
They were:
The number of followers a brand has
The number of accounts a brand follows
The number of tweets a brand has sent out
The presence of a ‘promoted’ stamp on individual tweets
We measured the impact of each of these variables in isolation while holding the other variables constant. Building on the success of our pilot experiment, we created another fictional brand: Resident, a unisex clothing brand in the style of ASOS or TopShop. By creating a fake brand we were able to control for biases in people’s experiences or perceptions of existing brands so that we could be sure that the effects being measured were driven by only the change in variables. Benchmarks for each variable number were set by looking at the numbers of ‘tweets’, ‘followers’ and ‘following’ that similar brands have on their real Twitter brand pages. 9
@RESIDENTLTD OUR FLORAL PENCIL SKIRT HAS BEEN GIVEN THE GOLD SEAL OF APPROVAL BY @GLAMOURMAGUK Promoted by ResidentLTD
10
IN TOTAL THE EXPERIMENT USED 21 DIFFERENT STIMULI SPLIT ACROSS THREE DIFFERENT AREAS OF EXPERIMENTATION. 01.Three different versions of biographical copy were written to reflect a ‘funny’ tone, ‘serious’ tone and ‘responsible’ tone.6 02. We tested each of the ‘numeric’ variables – followers, following and tweets – across four different levels; very low, low, high, very high. Each variable had its own benchmark at each level, set by looking at the numbers on real brand’s Twitter profiles.
#FUNNY “LET’S FACE IT, WE ALL LIKE A BURGER, THAT’S WHY RESIDENT JEANS ARE SUPER-STRETCH. @RESIDENTLTD, SUPPORTING YOUR EATING HABITS SINCE 1992”
03. We tested the extent to which a promoted ‘logo’ on an individual tweet changed perception by writing three different tweets - one showing ‘industry acceptance’ of the brand, another highlighting a popular industry event, and one with a competition mehcanic. For each of these tweets we had two versions, one with a promoted logo and one without.
#RESPONSIBLE “HERE @RESIDENTLTD, 10% OF OUR PROFITS GO TO A NOMINATED CHARITY EACH YEAR! FOLLOW US FOR ALL FASHION UPDATES OR TWEET US FOR QUERIES.”
6 There were original seven different biographical copies. Each of these were tested on Mechanical Turk with a random panel to make sure they could be identified clearly and differently as ‘funny’, ‘responsible’ and ‘serious’
#SERIOUS “FOUNDED IN 1992 BY DESIGNER AMY MONROE, WE HAVE GROWN INTO A LEADING FASHION CHAIN. OUR VALUES ARE SIMPLICITY, VERSATILITY AND TRUST.”
11
For this experimental treatment, all numbers were held constant at their ‘average’ level and only the biography copy changed.
VERY HIGH
FOLLOWERS FOLLOWING TWEETS
HIGH
LOW
2.7M 636,000 12,122 4,477 155,000 7,823
7,573 1,056 845
VERY LOW
358 268 343
In this part of the research, the variable we were testing (i.e. Followers, following or tweets) was changed as in the table above, while the other two variables were held constant at their ‘high’ levels. For example, while the number of Followers was changing, following and tweets were held constant at 4,477 and 7,823 respectively. By doing this we could control for interaction effects and can be sure that changes in brand perception are driven by changes in each of the variables alone.
“OUR FLORAL PENCIL SKIRT HAS BEEN GIVEN THE GOLD SEAL OF APPROVAL BY @GLAMOURMAGUK”
Industry acceptance
We recruited 4,511 Twitter users from an online panel by asking a screener question upfront. Those who claimed to use Twitter, out of a range of social and digital platforms, went through to the study. Each participant was randomly shown only one piece of stimulus. The experimental design was between subjects and approximately 215-220 individual respondents saw each piece of stimulus.
12
“OUR #STREETSTYLE EDIT FROM #LFW IS NOW UP ONLINE! OVERALL THE BUZZ WORDS ARE #RELAXED #COLOURBLOCK #CLEAN”
London Fashion Week (LFW)
“ENTER BY 4PM GMT SUNDAY, WE’LL DM 1 RANDOM WINNER WHO’LL WIN UP TO £500 & A GOOGLE CHROMEBOOK”
Competition
Two response scales To gauge levels of brand perception, each respondent was asked the same seven questions, each on a seven point Likert Scale.7 These were: Q3.1 PLEASE RATE THE TWITTER PAGE ON THE FOLLOWING 7-POINT BIPOLAR SCALE: WHERE 1 IS GOOD, AND 7 IS BAD Q3.2 PLEASE RATE THE TWITTER PAGE ON THE FOLLOWING 7-POINT BIPOLAR SCALE: WHERE 1 IS FAVOURABLE AND 7 IS UNFAVOURABLE Q3.3 PLEASE RATE THE TWITTER PAGE ON THE FOLLOWING 7-POINT BIPOLAR SCALE: WHERE 1 IS PLEASANT AND 7 IS UNPLEASANT Q2.1 I HAVE A STRONG INTEREST IN RESIDENT (PLEASE RANK THE FOLLOWING STATEMENTS ABOUT RESIDENT USING A SCALE WHERE 1 = STRONGLY DISAGREE AND 7 = STRONGLY AGREE.) Q2.2 I TRUST RESIDENT (PLEASE RANK THE FOLLOWING STATEMENTS ABOUT RESIDENT USING A SCALE WHERE 1 = STRONGLY DISAGREE AND 7=STRONGLY AGREE.) Q2.3 I LIKE THE IDEA OF BUYING A PRODUCT OR SERVICE FROM RESIDENT (PLEASE RANK THE FOLLOWING STATEMENTS ABOUT RESIDENT USING A SCALE WHERE 1 = STRONGLY DISAGREE AND 7 = STRONGLY AGREE. ) Q2.4 I WOULD RECOMMEND RESIDENT TO A FRIEND (PLEASE RANK THE FOLLOWING STATEMENTS ABOUT RESIDENT USING A SCALE WHERE 1 = STRONGLY DISAGREE AND 7 = STRONGLY AGREE. ) Using these questions we were able to create two different scales. Questions 3.1 to 3.3 were aggregated to create a ‘Likeability’ scale. Questions 2.1 to 2.4 were aggregated to create a ‘Consideration’ scale.
scale measures more reflective responses. Questions around trust, purchases intent and advocacy demand a greater level of thought and a consideration about future actions, rather than just a reflexive feeling of ‘likeability’.
We focused on these two scales as each indicates a slightly different behavioural response. The ‘Likeability’ scale measures responses that are more instinctive and momentary, such as if they thought the profile was ‘Good’ or ‘Favorable’. The ‘Consideration’
The questions that make up the ‘Consideration’ scale are also far more associated with traditional measures of brand perception and therefore may give a better indication of how likely someone is to engage with the brand on a commercial level.
IN SUMMARY, THE UNCONSCIOUS CUES EXPERIMENT CONSISTED OF THREE KEY ELEMENTS: • An online panel being shown stimulus about a fictional brand’s Twitter profile • Variable stimulus to explore the dynamics of multiple unconscious cues that form the Twitter user experience • Two response scales to measure the impact of the cues on respondents’ propensity to like and consider the brand
7 Likert, Rensis (1932). “A Technique for the Measurement of Attitudes”.Archives of Psychology 140: 1–55.
13
THE RESULTS Our findings fall into three major areas.
1. Unconscious cues on Twitter profile pages do have a significant impact on people’s perception of the brand. The cues impact both response scales – likeability and consideration are both influenced by elements of the Twitter user experience. This impact varies depending on the cue and across the range of variables within that cue.
2. ‘Likability’ doesn’t always correlate with ‘Consideration’. Stimulus that positively influenced people’s propensity to like a brand didn’t necessarily positively influence people’s propensity to consider the brand. In other words, brand’s likeability score does not predict its consideration score. In fact, there are occasions where the two scores are inversely correlated.
14
3. Unconscious cues influence different audiences in different ways Responses to cues vary across demographics and type of respondent. As all responses were unconscious, this suggests that respondents were interpreting the stimuli in accordance with existing perceptions. Individuals attached their own meaning to what they were shown. We will now explore the findings in detail. In the results section below all percentages quoted represent the proportion of people selecting a score of five or above for a given question or group of questions on the two seven point scales. This represents the percentage of people who scored 5 and above. Differences of 5% or above are considered statistically significant.
THE RESULTS IN DETAIL 1. Unconscious cues on Twitter profile pages do have a significant impact on people’s perception of the brand. The results exhibit this pattern across four out of the five unconscious cues. We will demonstrate these in turn.
Followers Firstly we will look at how the number of ‘Followers’ that a brand has impacts the perception of the brand. On both the Consideration and Likeability scales, Followers have little impact at all levels. That is to say, at all Follower levels tested – 358 to 2.7 million – Consideration and Likeability scores on each of the scales were very similar. However, as Graph 1 shows, very high Follower numbers have a significant impact in how much people ‘Trust’ the brand and also how much they would ‘Like the idea of buying a product or service from the brand’. 25% of people selected five or above in terms of buying a product or service across very low, low and high follower numbers, this increases to 30% for very high Follower numbers. The percentage of people scoring five or above for Trust also increase by 5%, from 18% to 23%, for very high Follower numbers.
35
GRAPH 1 - NUMBER OF FOLLOWERS
% WHO SCORED 5 OR ABOVE
30 25 20 15 10 5 0
358
7,573
636,000
2,700,000
NO. OF FOLLOWERS THE BRAND HAS I TRUST RESIDENT
LIKE THE IDEA OF BUYING A PRODUCT OR SERVICE FROM RESIDENT 15
Tone of voice of biography copy In terms of consideration, both the ‘funny’ and the ‘serious’ biographies score highest. There is no statistically significant difference between the two for questions of overall consideration, recommendation, interest and trust. The ‘responsible’ brand biography consistently scores the lowest.
Graph 2 shows that around 13% of people scored five or above for the ‘funny’ biography, 12% for the ‘serious’ biography, but only 8% for the ‘emotional’ biography. It may be that, as this brand is unknown, people prefer messages that provide humour, information about the product or heritage of the brand rather than CSR messages. It may also be that messages on the subject of charity do not fit well with the brands designed image or natural audience.
GRAPH 2 - TONE OF VOICE CONSIDERATION SCALE 14 12 % WHO SCORED 5 OR ABOVE
10 8 6 4 2 0
FUNNY CONSIDERATION SCALE
16
SERIOUS
RESPONSIBLE
THE ‘SERIOUS’ BIOGRAPHY COMES OUT ON TOP IN TERMS OF ‘TRUST’. APPROXIMATELY 24% OF PEOPLE SCORED FIVE AND ABOVE FOR TRUST WHEN PRESENTED WITH THE ‘SERIOUS’ BIOGRAPHY, 20% WHEN PRESENTED WITH THE ‘FUNNY’ BIOGRAPHY, AND LESS THAN 15% SCORED FIVED AND ABOVE FOR TRUST WHEN THEY SEE THE ‘RESPONSIBLE’ BIOGRAPHY.
GRAPH 3 - GRAPH TONE OF VOICE, TRUST 30
% WHO SCORED 5 OR ABOVE
25 20 15 10 5 0
FUNNY
SERIOUS
RESPONSIBLE
TRUST
17
Promoted and non-promoted tweets Graph 4.1 is slightly different from the other charts in the report. Rather than showing percentage of those who scored five and above it shows the average (mean) scores, out of a total of seven, for each tweet shown to respondents. The differences represented were shown to be statistically significant at 95% confidence interval. What the chart shows is that promoted tweets consistently drive commercial awareness metrics, specifically trust, willingness to buy and recommendation more than non-promoted tweets across all three tweets. This isn’t to say that promoted tweets always drive consideration but promoted tweets prepare a brands followers for a commercial conversation with regards to those key elements of brand consideration. Consideration is driven by content and the context of the campaign.
3.4
GRAPH 4.1 - PROMOTED VS NON PROMOTED TWEETS - TRUST IN BRAND
MEAN RESPONSE
3.3
3.2
3.1
3
INDUSTRY ACCEPTANCE EXPOSED TO PROMOTED TWEET
3.4
LFW
COMPETITION
EXPOSED TO ORGANIC TWEET
GRAPH 4.2 - PROMOTED VS NON PROMOTED TWEETS - OPENNESS TO BUYING FROM THE BRAND
MEAN RESPONSE
3.25
3.1
2.95
2.8
INDUSTRY ACCEPTANCE EXPOSED TO PROMOTED TWEET 18
LFW EXPOSED TO ORGANIC TWEET
COMPETITION
3.2
GRAPH 4.2 - PROMOTED VS NON PROMOTED TWEETS - WILLINGNESS TO RECOMMEND BRAND TO A FRIEND
MEAN RESPONSE
3.1
3 2.9
2.8
INDUSTRY ACCEPTANCE EXPOSED TO PROMOTED TWEET
LFW
COMPETITION
EXPOSED TO ORGANIC TWEET
Following Interestingly the number of other accounts that the brand follows seems to have a strong influence on people’s overall consideration score. As the brand follows more accounts, consideration drops steeply from a high of around 15% at a following number of 1,056 to a low of 8% for very high following numbers (12,122).
GRAPH 5 - FOLLOWING
% WHO SCORED 5 OR ABOVE
20 15 10 05 0 268
1,056
4,477
12,122
NO. OF ACCOUNTS THE BRAND FOLLOWS CONSIDERATION SCALE
19
The pattern represented on Graph 5 – scores falling as the number of accounts the brand follows increases – is replicated across all four components of the consideration scale. Scores for trust, recommendation, interest in and willingness to purchase from all drop off steeply at ‘very high’ following numbers. This may be because a brand with very high ‘following’ numbers is seen as indiscriminant in their use of the platform or that they only follow others to gain Followers themselves, a practice known as ‘Follow Back’. Interestingly, the likability of the brand does not have the same pattern. This will be discussed in the next section.
2. ‘Likeability’ doesn’t correlate with ‘consideration’. Often it is assumed or even expected that the ‘likeability’ of a brand should go hand-in-hand with levels of ‘consideration’. What this study has shown is that this is not always the case. We have found that in many cases a change in one indicator will lead to an increase in likeability but a decrease in consideration. Take graph 6 for example. This graph shows that the responsible biography scored the highest in terms of likeability, but the lowest for consideration.
%WHO SCORED 5 OR ABOVE
GRAPH 6 - TONE OF VOICE, LIKEABILITY VS CONSIDERATION 18 16 14 12 10 8 4 2 0 FUNNY CONSIDERATION SCALE
SERIOUS
RESPONSIBLE
LIKEABILITY
This pattern is repeated across many of the variables tested. For example, average scores for the non-promoted tweets are 8% higher on the ‘likeability’ scale than on the ‘consideration’ scale (see Graph 7). When we compared the average scores of the promoted vs. non-promoted tweets we found that promoted tweets scored higher for ‘consideration’ while the non-promoted tweets scored higher for ‘likeability’. This is perhaps because the ‘promoted’ badge is a signal that makes people automatically associate that tweet with something commercial, generating a higher score on the ‘consideration’ scale. Organic tweets may be more liked simply because they are received on an ‘opt-in’ basis, as the user has made a conscious choice to follow tweets from the brand. 20
GRAPH 7 - PROMOTED VS NON-PROMOTED TWEETS (SCORING 5 AND ABOVE) 20 18 16 14 12 10 8 6 4 2 AVG. NON-PROMOTED %
AVG. PROMOTED % CONSIDERATION SCALE
LIKEABILITY
As can be seen in graph 8 this pattern persists. The profiles with very low and very high following numbers are the most liked but the least considered.
GRAPH 8 - FOLLOWING, CONSIDERATION VS LIKEABILITY
% WHO SCORED 5 OR ABOVE
20 15 10 5 0 268 CONSIDERATION SCALE
1,056
4,477
12,122
LIKEABILITY 21
3. Unconscious cues influence different audiences in different ways There are some clear demographic differences across all variables and on both the likeability and the consideration scale. Looking firstly at the scores for likeability and consideration overall – across all 21 of the variables tested – graph 9 clearly illustrates this difference. The graph shows the overall scores for likeability and consideration cut by the claimed level of confidence in using the Internet; 1 being a complete novice and 7 being an expert.
GRAPH 9 - OVERALL SCORES BY CONFIDENCE IN USING THE INTERNET
%WHO SCORED 5 OR ABOVE
30 25 20 15 10 5 0
NOVICE
02
03
04
05
06
EXPERT
INTERNET USAGE CONFIDENCE CONSIDERATION SCALE
LIKEABILITY
It shows a clear linear relationship between confidence in using the Internet and consideration scores. This indicates that those who are most comfortable with the Internet are most likely to trust, recommend, be interested in or be willing to buy a product or service. However, conversely those who feel that they are less confident online are least likely to score high on ‘consideration’ but most likely to score high in terms of likeability.
22
There is also a variation in age across the two scales. Older people are more likely to score any piece of stimulus higher on the ‘likeability’ scale compared to the ‘consideration’ scale. The age group most likely to score things highly on the consideration scale is 25-34s. Around 18% of all 25-34s score any piece of stimulus five or higher. This may be driven by the differences in Internet confidence across the age groups. The data shows that older people are more likely to score themselves closer to the novice end of Internet confidence usage.
%WHO SCORED 5 OR ABOVE
25
GRAPH 10 - CONSIDERATION AND LIKEABILITY BY AGE
20 15 10 5 0 18-24
25-34
CONSIDERATION SCALE
35-44
45-54
55-64
65+
LIKEABILITY
Looking at the demographic differences within specific variables we see that younger people – 18-34s – are most likely to prefer higher Follower numbers (the percentage for 55-64s is high here too, but there is a low sample size in this group when broken down to this level of granularity).
23
APPROXIMATELY 10% MORE 18-34S SELECT FIVE OR ABOVE FOR VERY HIGH FOLLOWER NUMBERS COMPARED TO VERY LOW FOLLOWER NUMBERS. GRAPH 11 - FOLLOWERS CONSIDERATION SCALE BY AGE 35
%WHO SCORED 5 OR ABOVE
20 15 10 5 0 VERY LOW
LOW
HIGH
FOLLOWER NUMBERS 18-24
24
25-34
35-44
45-54
55-64
65+
VERY HIGH
MEN ARE MORE INFLUENCED BY HIGH FOLLOWER NUMBERS COMPARED TO WOMEN This may be because of confidence in interacting with the fashion sector. Women are perhaps more confident when considering a brand such as Resident, and therefore are less likely to look for social validation in the form of greater Follower numbers.
GRAPH 12 - FOLLOWERS CONSIDERATION SCALE BY GENDER 20 18
%WHO SCORED 5 OR ABOVE
16 14 12 8 6 4 2 0 VERY LOW MALE
LOW
HIGH
VERY HIGH
FEMALE
25
GRAPH 13 - TONE OF VOICE OF BIOGRAPHY BY GENDER 16 14 %WHO SCORED 5 OR ABOVE
12 10 8 6 4 2 0 FUNNY MALE
SERIOUS
RESPONSIBLE
FEMALE
WOMEN MOST PREFERRED THE ‘FUNNY’ BIOGRAPHY, BUT WERE ALSO SLIGHTLY MORE RESPONSIVE TO THE ‘RESPONSIBLE’ BIOGRAPHY COMPARED TO MEN. Younger people clearly had a preference for the ‘funny’ biography in terms of consideration while older people preferred the ‘serious’ biography. Women most preferred the ‘funny’ biography, but were also slightly more responsive to the ‘responsible’ biography compared to men.
26
CONCLUSIONS AND IMPLICATIONS 1. Be clear on what you’re trying to achieve What have we found? The scores for likeability and consideration do not respond in the same way to changes in the variables. This means that higher levels of likeability do not lead to higher levels of consideration. Why do we think it’s happening? The questions relating to likeability can be answered quickly and instinctively as an unconscious, emotional response. Though still capturing an unconscious response, consideration is a more reflective, considered scale – respondents are being asked to think through and predict their likely actions. What are the implications for brands? Think about what kind of response you are looking to generate with your Twitter presence. What does your brand biography say about you? Is your content strategy more focused on building brand affinity or on future purchase intent? Think about how you can structure followers’ experience to emphasise the approach and make the intended response more likely.
2. Be clear on who you’re trying to reach it with What have we found? Higher levels of Likeability correlate with older respondents and respondents who are less experienced in Internet and/or social media. Consideration scores are higher amongst expert Internet users and younger people. Why do we think it’s happening? Older and inexperienced Internet users have less to compare the stimulus to as they may not have
much knowledge of online shopping or commercially engaging with brands on Twitter. They may also be more risk averse when it comes to consideration. Younger and more experienced users are more familiar with brands on Twitter and are perhaps less easily impressed. What are the implications for brands? Understand the comfort levels of your audience and which approach is most likely to nurture your relationship with them. When it comes to less experienced users, don’t overestimate ‘liking’ metrics and make sure you’re doing enough to build consideration. For younger audiences, don’t be afraid to close the deal – they’re far more comfortable with the idea of entering into a transactional relationship. It’s important to note that this could be a function of our fictional brand and its category appeal. Make sure you understand which audience segments these behavioural approaches might apply to.
3. Strike the right tone for your brand What have we found? Of the three biographies created, the ‘funny’ and ‘serious’ biographies were the most considered. When it comes to likeability, however, the emotional biography scored highest. Why do we think it’s happening? The questions on the Likeability scale make it likely that the responsible biography is the most immediately appealing. We think the ‘serious’ biography may lead to consideration because it is the most informative – it appeals to a more reflective response and consideration of how a respondent might feel about the brand.
27
What are the implications for brands? Where humour is a key part of a brand’s values and architecture, a humorous tone with Twitter can add value. However, it may inhibit users’ propensity to enter into a more intimate relationship with you. Perhaps this mirrors the shopping and offline experience of a clothing retailer. Credibility is gained through valuable and effective interactions – these create expectations that may be carried into the online experience. This research shows that the brand biography can be used as a tool to conduct particular types of conversation with your audience. It is a variable that audiences use to make judgments about the character of the brand. When launching a brand advertising campaign, reframing the biography in a more human or humorous tone may help drive likability as people discover the brand on Twitter. Conversely, a more serious tone in the biography would help drive commercial consideration during a large DR campaign. On Twitter, be clear on the kind of tone a customer would expect you to strike, and make sure you consistently meet those expectations.
4. Use the right tools for the job What have we found? Unpromoted tweets were liked more than the promoted tweets, as we might expect. However, promoted tweets appear to trigger a better response on the ‘consideration’ scale. This was observed at a population level, but the effect was even more pronounced among more experienced users. Why do we think it’s happening? Users ‘opt-in’ to see unpromoted tweets, but promoted tweets suggest credibility and scale to a brand. Perhaps a presence on paid formats suggests a brand being sufficiently well established to be able to afford activity of this kind. 28
What are the implications for brands? Promoted tweets offer a chance to build credibility and consideration. They could be particularly useful to support a campaign focused on sales or purchase intent.
5. Think about who and what the users will compare you to What have we found? Very high numbers of Followers build levels of trust and purchase intent. There is a gender difference, however – women tend to be interested in lower numbers of Followers while for men the numbers need to be higher Why do we think it’s happening? We would expect to see higher levels of Followers create a norming effect. This wouldn’t explain the gender difference, however. Our hypothesis is that within the fashion category, greater popularity might not always result in increased purchase intent. Exclusivity is a category motivator, and this might be expected to apply more visibly to informed and confident female respondents. Male respondents might be expected to look for reassurance in higher numbers and greater perceived popularity. What are the implications for brands? In social, it is important to balance scale with intimacy. Our pilot experiment showed that larger communities generated positive responses in diminishing returns. The results of this experiment suggest building your follower count isn’t enough on its own to drive a brands commercial objectives on Twitter. Category norms can also be played out on social media so it’s important to think about these as part of an experience that users share with everyone else. Therefore, followers are an important part of the equation, but as part of a wider strategy around who you are trying to reach and how you are going to reach them.
FINAL THOUGHTS Many clients ask agencies, ‘what should our Twitter strategy be?’
A high score on the ‘consideration’ scale is characterised by:
The frustrating answer that clients sometimes receive is, ‘it depends’.
• An audience likely to be 25-34 • With a high level of confidence using the internet • Motivated by high follower numbers – especially men
Twitter is a public sphere, a social medium. People use it for many different reasons. All, or just merely most of, human life is here. It is subject to the rules of human behaviour, just like any other area of life. That makes it complicated. There isn’t necessarily a single, ‘right’ way to use Twitter. Throughout this report we have recorded the different effects that cues had on different types of user. There are many different drivers of perception open to brands on Twitter. Understanding the way these drivers relate to your brand or business challenge can help identify the right way, based on a desired outcome. Consider these two scenarios, developed by drawing on the drivers that correlate most strongly with high scores on our two different scales:
A high score on the ‘likeability’ scale is characterised by: • An audience of people likely to be older than 45 • With a low level of confidence using the Internet • Not motivated by Follower numbers, apart from men, who are more likely to like a brand if it has fewer Followers Stimulus that caused people to consider our brand resonated with a very different audience to the one that found the brand to be likeable. That has significant implications for brands using the platform to talk to everyone at the same time, or for brands unsure of what they’re trying to achieve with their Twitter presence. Targeting is becoming an increasingly important part of the Twitter experience. While brands have become adept at creating content that resonates with their target audience, we think the cues within the user experience offer another more nuanced targeting opportunity. We hope this research offers brands the chance to do that.
29
AUTHORS Joe Gladstone & Jon Jachimowicsz PhD Researcher in Behavioural Economics, University of Cambridge Judge Business School
Nick Siantonas Behavioural Strategist, Isobar
James Caig Head of Strategy, Isobar
Stephen Donajgrodzki Senior Partner, Isobar
30
