BEHAVIOUR

CHANGE GUIDE #4
Decoding human behaviour.
“The remembered self is The human mind is not well-designed to
provide an accurate report of behaviour.
Out of the petri dish and
into the real world
not a faithful reporter of Despite our best intentions, our minds are
affected by biases that mean we are not Findings from experimental
the experienced self.” accurate in sharing back our behaviours. settings, surveys and focus groups are
not always representative of real-world
Daniel Kahneman behaviours.
Our memories are not always reliable and
we can have poor recall especially when Outcomes can be affected by:
we’re tired or under time pressure.1 – Leading questions and survey bias. 5
– Researcher and research effects
We tend to be overconfident and overly (such as the Hawthorne effect or
optimistic, especially in relation to our group think). 6
own behaviour or those in our in-group. 2 – The environment in which research
takes place, including the physical
room or the digital context.7
We want to provide socially desirable
answers. 3 – The use of key words and visual
prompts. 8

Behavioural science methods and

We tend to remember the emotional
techniques can provide accurate ways
peaks and the end of an experience. 4
of capturing actual behaviour.

Consider using these approaches

instead of, or alongside, more traditional
quantitative or qualitative research
methods.

1
Catalogue of Bias Collaboration, Spencer EA, Brassey J, Mahtani
K. Recall bias. In: Catalogue Of Bias 2017.
2
The optimism bias, T. Sharot, Current Biology, 21 (23) 2011,
pp. R941-R945
3
The relationship between social desirability bias and self-reports
of health, substance use, and social network factors among urban
substance users in Baltimore, Maryland., Latkin CA, Edwards C,
Davey-Rothwell MA, Tobin KE. Addict Behav. 2017 Oct;73:133-136.
4
When More Pain Is Preferred to Less: Adding a Better End,
Kahneman, Daniel; Fredrickson, Barbara L.; Schreiber, Charles A.;
Redelmeier, Donald A. (1993). Psychological Science. 4 (6): 401–
5
A catalog of biases in questionnaires, Choi BC, Pak AW. Prev
Chronic Dis. 2005 Jan;2(1):A13. Epub 2004 Dec 15. PMID: 15670466;
6
Sackett Catalogue of Bias Collaboration, Spencer EA, Mahtani
K, Hawthorne effect. In: Catalogue Of Bias 2017
7
The influence of ceiling height: The effect of priming on the type
of processing that people use, Meyers-Levy, Joan, and Rui Zhu.
Journal of consumer research 34.2 (2007): 174-186.
8
Using Priming Techniques to Facilitate Health Behaviours,
Wryobeck, J. and Chen, Y. (2003), Clinical Psychologist, 7: 105-10 P3
Capture and understand
real-world behaviour

Before we delve into the specific methodologies,

here are the guiding principles to consider:

1 Capture real behaviour first

5 Behaviour is not locked in
Capture what people cannot explicitly tell you by looking beyond Consider opportunities for reinforcement
reported behaviour and look to capture actual behaviour. and continuous evaluation.

2 Test and trial in the real world

6 Ethical duty
Look to capture behaviour in the context in which it occurs. Adhere and respect people’s autonomy, dignity and rights,
and evaluate unintended consequences.

3 Evidence based
7 Differences in audiences
Create measurable behaviour change objectives based on Aim to understand individuals as whole people with
empirical data and cross-validate your findings with multiple different backgrounds, demographics, cultures,
data points where possible. ethnicities and affiliated groups.

4 The bigger picture of behaviour change

8 Check our own biases
Make sure the overall system supports the behaviour, Be open to the many factors that can influence behaviour
as we cannot view nudges in isolation. and be wary of our biases and preconceived ideas.

P5
Select the right tool
for the job LARGER
SAMPLE SIZE

We can unlock behavioural insights

with the right behavioural science
methodology. RCTs META-ANALYSIS
BEHAVIOURAL LONGITUDINAL
COMMUNICATIONS
Applied behavioural science is PRE AND DATA SETS STUDIES
POST TRIALS EXPOSURE IMPACT
a growing field. This list is not
exhaustive, but it does provide an
A/B TESTING ONLINE PASSIVE
overview of the current approaches. METERING

IMPLICIT RESPONSE
TESTING IN THE MOMENT
MOBILE SURVEYS

SPECIFIC BROADER
BEHAVIOURAL PATTERNS OF
RESPONSE BEHAVIOUR

BEHAVIOURAL
OBSERVATION

PHYSIOLOGICAL
MEASURES UX TOOLS
(ECG, GALVANIC INTERCEPTS
SKIN RESPONSE)
EYE TRACKING
FACIAL CODING
BEHAVIOURAL
NEUROIMAGING DIARIES
(fMRI AND EEG)

SMALLER
SAMPLE SIZE

P7
Think about whether the Think about the
behaviour in question is sample size
broad or specific

A cyber security example: Does the research approach allow for a small or large sample?

BROADER Understanding different groups of people’s SMALLER Are useful for a more in-depth understanding and
PATTERNS OF current cyber security behaviours and their SAMPLE SIZE for more complicated behaviour that you can track
motivations to stay secure online. over a full experience. Also useful for exploratory
BEHAVIOUR
research or pilot studies before scaling to a larger
sample group.

SPECIFIC Understanding people’s responses to LARGER Have greater statistical power and can help
BEHAVIOURAL incentives that prompt them to use a long SAMPLE SIZE detect meaningful differences or relationships
and strong online password. between variables.
RESPONSE

P9
Understanding
the different types LARGER
SAMPLE SIZE

of behavioural
data that will be Experimental Data Sets
collected RCTs META-ANALYSIS
LONGITUDINAL
BEHAVIOURAL
PRE AND COMMUNICATIONS STUDIES
DATA SETS
POST TRIALS EXPOSURE IMPACT

A/B TESTING ONLINE PASSIVE

METERING

IMPLICIT RESPONSE
TESTING IN THE MOMENT
MOBILE SURVEYS

SPECIFIC BROADER
BEHAVIOURAL PATTERNS OF
RESPONSE BEHAVIOUR

BEHAVIOURAL
OBSERVATION

PHYSIOLOGICAL
MEASURES UX TOOLS
(ECG, GALVANIC INTERCEPTS
SKIN RESPONSE)
EYE TRACKING
FACIAL CODING
BEHAVIOURAL
NEUROIMAGING DIARIES
(fMRI AND EEG)

Physiological Observation

SMALLER
SAMPLE SIZE

P11
Experimental methodologies Best suited for: Behavioural data sets – such Best suited for:

involve changing one or more Larger sample sizes, specific behaviour. as census data, engagement Larger sample sizes, broad patterns of behaviour.
variables to measure the effect. metrics, purchase and spend
Experimental research can be Example: data and CRM data – can help Example:
useful when trying to establish Testing to see the effectiveness of a social norm
us understand broader patterns Traffic data, location tracking data and public
a cause-and-effect relationship message versus a loss aversion message when of behaviour. The data is often transport data (locations where people tag on and
between different variables. it comes to motivating people to start using a captured in a non-experimental tag off), can tell us the most common trips and
password manager. destinations and when they were taken.
setting and therefore is often
Limitations:
representative of the real-world Limitations:
behaviours.
Generalising findings to the real-world context can Data sets do not always provide the underlying
be difficult, depending on the conditions of the motivations behind the data or the important
study. Groupings of people and experiments need context in which the behaviour occurred. Watch out
to have a large and diverse sample to account for for data validity and how representative the data is
differences between specific groups. of the population of interest.

Physiological methodologies Best suited for: Observing behaviour can be Best suited for:

capture our physiological Smaller sample sizes, specific behaviour. done in the environment in which Smaller sample sizes, broad patterns of behaviour.
or unspoken System 1 the behaviour occurs, making
responses to a stimulus, such Example: it useful for understanding real Example:
as the electrical rhythms of It can often be hard for people to articulate their
behaviour in the real world. Observing people trying to navigate through a
our heart, facial reactions reactions. Facial coding and reaction times can This includes observing how train station they’ve never been to and using eye-
or skin conductance. This is help us understand people’s System 1 responses the broader environment can tracking glasses to understand what navigation
to particular messaging or communications. cues they are using.
useful for when responses or affect behaviour.
behaviours can be hard Limitations: Limitations:
to detect.
Some of these methods require people to wear Observation often only captures visible behaviours,
clunky or invasive technology which can limit the not the underlying motivations. It can also
real-world test scenarios of the research. Findings be affected by observer bias and subjective
tend to give one specific behavioural response interpretation. Watch out for self-reporting
rather than a complete picture of the behaviour. measures as they be influenced by biases such
as social desirability bias, leading questions and
memory bias.

P13
GLOSSARY OF BEHAVIOURAL SCIENCE METHODOLOGIES

Experimental Physiological Data sets Observation

RCTs (Randomised Control Trials): Experiments EEG: Electroencephalogram that measures Behavioural data sets: Data and information that Behavioural observation: Capturing and
in which participants are randomly assigned into electrical activity in the brain. It can be used captures peoples behaviour. recording people’s behaviour in the real world.
different groups (e.g. a treatment group versus a to test which areas of the brain respond to stimulus,
control group) to test interventions. and to what extent. Some examples include: Some examples include:

Pre and post trials: Experiments that measure ECG: Electrocardiogram measures the electrical – Census data – Observers watching a behaviour over time,
outcomes or changes in behaviour before and activity of the heart in response to different – Purchase and spend data capturing frequency, duration and other
after an intervention. stimulus. measures of the behaviour
– Engagement metrics
– Reviewing CCTV footage or video capture
A/B Testing: Two (or more) versions of a message, fMRI: Functional magnetic resonance imaging – CRM data
website, product or service are tested against measures brain activity by detecting – Location tracking data Intercept interviews: Interviews conducted with
each other to determine which is more effective. blood flow. people in real-world settings to capture and
Meta-analysis: Merging and analysing findings understand the context and attitudes around a
Communications Exposure Impact: Measures the Galvanic Skin Response: Measures electrical from independent studies using statistical behaviour.
impact messaging or communications in the real conductance of the skin which can indicate methods.
world has on people’s behaviours and attitudes. changes in emotional states. UX tools: Capture people’s responses to websites,
Longitudinal studies: Collecting data from the mobiles apps and other digital products and
Implicit response testing: Measures unconscious Facial coding: Using software and video capture to same participants over time to understand the services. This includes mouse tracking, screen
attitudes and biases based on people’s reaction analyse facial expressions and movements to infer change in behaviour and other measures, such as and user video capture, Google analytics and eye
times. emotional responses and measure engagement. physiology and attitudes, over time. tracking.

Eye tracking: Glasses that capture eye movement In the moment mobile surveys: Short surveys Behavioural diaries: Self-reported accounts
to measure what people are looking at. delivered to people’s phones in real-time, typically of people’s behaviour. This can be of natural
in response to an event or around the time of a behaviour as and when it occurs, or a behavioural
particular behaviour. mission where participants are completing a new
or different behaviour.
Passive metering: Tracking people’s online
behaviour across digital devices.

Subscribe to our newsletter

CONFIDENTIAL © TR A 2023 P15

TheResearchAgency.com