Measuring whether and how brand marketing changes people’s perception of a brand often remains more art than science. Tools to track brand perception can easily cost millions across multiple markets but frequently come with serious challenges in data quality. One thing that can greatly contribute to ambiguity in the data authenticity and accuracy is the sampling source. 

Research panels often consist of “professional respondents”, not everyday consumers. 

Only a tiny fraction of a country's population – often as little as 0.1% – is signed up to research panels and actively participates in them. In addition, certain demographics are often overrepresented in research panels.

A small group of survey takers (aka "professional respondents") completes a disproportionate number of surveys, sometimes hundreds or even thousands per year. When only a tiny fraction of consumers are actively involved in research, the insights gathered can be fairly unrepresentative and biased to the type of person who willingly signed up to participate in regular surveys. This sampling bias is likely reflected in skewness and inaccuracies in the resulting brand data.

The data quality challenges for reaching everyday consumers

A reliance on “professional respondents” can create significant challenges for data quality in brand tracking research: 

1. Self-selection bias

Joining research panels involves extensive vetting and pre-qualification steps, attracting only those highly motivated to participate. This creates a self-selection bias, as panel respondents tend to represent individuals with a strong inclination toward research participation. Self-selection bias can also occur when people choose to participate based on specific interests, and when individuals are more aware of and/or knowledgeable about a topic and decide to partake in the survey as a result.

Motivation and interest might sound like positives when the initial aim is to gather many survey completions. However, self-selection bias causes a big problem for brand tracking, strongly impacting the reliability of data. Brand awareness scores can be inflated, market share insights skewed, and consumer preferences misrepresented. The data ends up reflecting the habits and perceptions of a niche group rather than the broader population.

2. Limited sampling in small markets

In small markets, the respondent pool for online panels is often too small to ensure new participants for each wave of data collection. For example, in Switzerland, with a population of nearly 9 million, only a few thousand individuals actively participate in research panels. This scarcity makes it challenging to implement exclusion rules effectively and limits the diversity of respondents.

The alternative approach to reaching smaller markets is to subcontract other panels to work simultaneously to achieve the sample size required to be reliable enough. This is an expensive solution with sampling costs that could be three or four times that of a larger market. 

3. Demographic skews

Panel respondents often overrepresent certain demographic groups (e.g. middle-aged, stay-at-home individuals) while underrepresenting others (e.g. older and younger groups, people with above-average income). When tracking requires segmenting underrepresented groups or maintaining stable sample compositions, this bias can lead to larger margins of error, seemingly random data fluctuations, and poor reliability of insights. 

‍Solution: reaching the 99% of consumers who aren’t signed up for panels

To reach the vast majority of consumers who aren’t signed up for research panels, an innovative solution researched and implemented by Latana leverages interactive ads that allow people to answer survey questions directly within digital advertisements. These surveys are embedded into full-page ads, seamlessly displayed across apps and mobile websites, allowing respondents to provide answers without leaving the platform or registering for a research panel.

This method takes advantage of the global ubiquity of smartphones, offering access to a far more representative sample compared to traditional research panels when it comes to data collection. Traditional panels typically sample between 1,000 and 12,000 people per year per brand, whereas the ad-based sampling approach can engage up to 100,000 respondents, providing a larger and more diverse pool. Importantly, it samples anyone with a smartphone, significantly reducing the self-selection bias present in traditional panels that rely on self-selected participants.

Moreover, this method spans 200+ countries, offering global insights across regions, while traditional methods are limited to the top 50 markets. With the ability to bid on over 50 billion ad impressions that reach over three billion people daily, this approach ensures seamless, instant access to any geography, avoiding the sample shortages that often occur in traditional methods.

The methodology not only scales efficiently, but also addresses critical challenges like demographic skews. For example, it enables easy access to harder-to-reach groups such as 18–25-year-olds and those aged 66+, groups that traditional panels often miss.

Final thoughts

The vast majority of online survey research today is conducted using panels, which severely underrepresent the majority of the world population. Reaching everyday people is crucial to achieving more representative and accurate brand data. We use ad-based sampling to reach everyday people, enabling us to:

1 - Better represent the population being measured

With a much wider pool of respondents available to us, we can collect large and representative samples that do not require heavy weighting to account for underrepresented groups of people. 

2 - Reach more of the world’s population

We are also able to collect larger samples even in smaller countries that are traditionally difficult to reach using panel-based methodologies. We never have to reduce our standard sample sizes or use supplementary methods to reach respondents. 

‍3 - Avoid self-selection bias

Our ad-based samples are opt-in and opt-out, without an incentive to complete the entire survey. This approach removes any skew towards highly motivated survey participants and ensures that the answers we collect are genuine and authentic.