Survey

Our Audience Profiler survey aims to uncover the nuanced personal circumstances, shopping behaviours, psychographics, values, and beliefs of people from all walks of life.

Data collection

The figures displayed in Audience Profiler represent results from tracking surveys conducted by Canvas8 in key markets on a quarterly or half-yearly basis. For each survey, a sample of the general population is drawn from one of our preferred panel partners. Sampling takes into account respondent age, gender, region, and race/ethnicity. Minimum targets for minority ethnic communities in each market are applied to ensure representation within the results. To ensure a sufficient sample within each generational bracket, sampling is not nationally representative of age – but is weighted after.

Weighting

Responses are weighted at the national level according to age and gender to correct for sample distribution. This allows us to collect enough data for each generation individually and display nationally representative results in the Audience Profiler platform.

Results are filtered to remove articles tagged with locations, generations, and genders that do not match the user-selected filters. We then rank the results to prioritise the most relevant articles using a combination of AI Semantic search on interests (including all interest categories they fall into, weighted by how many of them are in each group) plus boosts weighted to the results of locations, generations, and genders. Note: we treat articles untagged with any locations, generations, and genders as if tagged with all options.

These calculations test whether the size is robust enough that, if we were to rerun this exact set of questions with a different sample group, the figures provided within it, e.g. the percentage living in urban areas, would be within five percentage points (a 5% margin of error) at least 90 times out of a 100 (a confidence level of 90%). Confidence levels for a 3% margin of error and a 5% margin of error are automatically generated on filtered audiences to a minimum of 75%. If a sample size has produced a confidence level below 75% with a 5% margin of error, we indicate this, as, at this point, we would need to increase the margin of error beyond the recommended tolerance.

The minimum number of people allowed in a filtered audience is 100, which exceeds the minimum sample size required for a confidence level of 95% with a margin of error of 5% for every market. This means that if we had repeated this survey 100 times, we would expect the number of people in that audience group to be within 5% at least 95 times.

In addition to our content Library, Canvas8 provides consultancy, often requiring research survey data. The Audience Profiler survey is run by our Data & Analytics team to the same high-quality standards.

Data collection

Data is collected through an online survey via Alchemer, with pre-screened respondents invited to participate via email. To reach relevant samples for proprietary and consultancy research (B2C and B2B), we typically use one of two preferred providers – Toluna or Cint. However, we have relationships with several other sample suppliers to ensure we can conduct research with niche or otherwise difficult-to-reach audiences (e.g. through YouthSight, Dynata, or Lightspeed). Through our preferred providers, we have access to over 62 million people globally. Respondents are recruited to these panels through open enrollment, loyalty programmes, and affiliate networks. Panel composition varies by provider and market.

Quality assurance

We use a combination of human and technology-based data checks. These include:

• Duplicate IP checks: blocking survey respondents who attempt to complete the same survey multiple times
• Survey speeders: respondents who rush through the survey are identified by comparing survey and question completion times to the norm
• Straightliners: respondents who consistently answer a series of questions or rows on a grid in the same place
• Pattern responses: respondents who consistently answer a series of questions or rows on a grid in a predictable pattern
• Red herrings: respondents who provide illogical responses to questions with only one correct answer
• Respondent satisfaction: feedback from respondents is gathered and assessed to continually monitor the quality of surveys

Yes. At a minimum, any data derived from Canvas8 should cite Canvas8 and the fieldwork period as the source. We recommend one of the following formats for clarity:

• Source: Canvas8 Profiler Survey, your audience filters (n=xxx), MM-MM YYYY (accessed MM YYYY)
• Source: Canvas8 Profiler Survey, MM-MM YYYY (accessed MM YYYY)
  Base: your audience filters (n=xxx)
  

Our AI has been trained to provide summaries on key metrics, reporting on the highest number of responses within the audience sample group. Summaries have been inspected by teams at Canvas8 for inaccuracies, pressure testing, and refining the training. However, the technology is still in Beta and may occasionally summarise data in unexpected ways.

We have provided our AI with a detailed dataset of key metrics. The summary is generated based on this rather than on the charts displayed. Many of the category lists are long, and including them in full would significantly reduce readability. Therefore, we have set a maximum number of ten categories to be displayed. If you'd like to see an expanded dataset, get in touch.

AI-powered semantic searches influence the results in related articles by inspecting articles that initially surfaced via the demographics and interests of your audience. Article content that is conceptually similar is also included, as long as it isn’t specifically filtered out (for instance, suppressing articles specific to Britons when your selected audience is American). For example, if you were trying to find out about parents or people interested in parenting in Britain, many articles will be about British parents and their children. However, you might also see content related to, for example, careers, where insights and background impact, or are impacted by, parents.

Schwartz values, also known as the Schwartz Value Theory, refer to a model of human values developed by social psychologist Shalom H. Schwartz. This theory posits that there are ten basic values recognised across cultures and that these values represent universal motivations that guide individuals' behaviours and attitudes. The ten values identified by Schwartz are organised into four higher-order categories:

Self-transcendence:

• Universalism: Concern for the welfare of all people and for nature
• Benevolence: Preserving and enhancing the welfare of those with whom one is in frequent personal contact

Conservation:

• Tradition: Respect, commitment, and acceptance of the customs and ideas that one's culture or religion provides
• Conformity: Restraint of actions, inclinations, and impulses likely to upset or harm others and violate social expectations or norms

Self-enhancement:

• Power: Social status and prestige, control, or dominance over people and resources
• Achievement: Personal success through demonstrating competence according to social standards

Openness to Change:

• Stimulation: Excitement, novelty, and challenge in life
• Self-direction: Independent thought and action, creativity, exploration of new ideas and experiences.

Hedonism: Pleasure and sensuous gratification for oneself

These values are believed to be universal, although their importance and prioritisation may vary among individuals, cultures, and contexts. Schwartz's theory has been widely used in social psychology, cross-cultural studies, organisational behaviour, and consumer research to understand human behaviour, attitudes, and decision-making processes.

Confidence levels are calculated based on the filtered population, representing an entire sample set, rather than calculating confidence in the proportion of the filter out of the total sample size. This is done so confidence intervals can be established for the responses given by that filtered audience group. As more data is added, these audience groups will also grow, and higher confidence levels can be used.

The minimum number of people allowed in a filtered audience is 100, which exceeds the minimum sample size required for a confidence level of 95% with a margin of error of 5% for every market. This means that if we had repeated this survey 100 times, we would expect the number of people that fall into that audience group to be within 5% at least 95 times.

Audience Profiler will be in beta mode until September 2024. During this period, Members can test the tool for free. Beta mode includes access to a dataset of five core markets and the ability to create and track up to five audience profiles.

This is a preview of the dataset, multi-wave views will be available in upcoming iterations later in 2024.

We understand the importance of integrating custom data sets for audience understanding. While the beta mode doesn’t currently support custom data, we’re actively exploring this feature for future updates. We encourage you to share your needs and suggestions to help us prioritise this feature in our development roadmap.

Data is updated every quarter. Waves (periods of data collection) are completed twice a year, with planned pauses in fieldwork allowing us to add or remove custom questions. When new data is added to a wave that is live in the Audience Profiler tool, all graphs, statistical calculations and summaries will incorporate new entries automatically.

Countries currently included for 2024 are the UK, the US, France, Germany and Australia. Markets were prioritised according to membership demand. As such, if additional markets are needed please get in touch to discuss requirements.

Our filter limitations are in place to ensure a minimum of 100 people per profile, i.e. a sufficiently robust sample size for statistical testing.

Yes. If you require additional or different custom filters or data in addition to the core profiler, please get in touch to discuss your requirements.

While we are in Beta, our Data & Analytics and Development team are working to roll out audience comparison views later in 2024 H2, which will allow users to apply additional series in charts, as well as comparison to market total. In the meantime, audiences can be compared by saving individual profiles and checking their results against each other.

While Audience Profiler is in beta mode, we’ll be working on improving the technical architecture of the tool so that we have better custom controls to manage licensing and access in line with your Membership license tier. We’ll update you on access levels after beta mode has ended, or you can discuss it with your account manager.