Recruiting Hard-to-Reach Groups for Low-Incidence Market Research
Low-incidence audiences or populations are statistically small groups within the overall population. Low-incidence market research shows up in a wide variety of research scenarios, but the challenges are fairly constant.
For instance, a brand may want to rank or benchmark its product adoption across an entire category of competitors, including those with niche offerings that only niche audiences would adopt. The brand may want to innovate for or market to a specific niche segment. Or it may want to get to know niche segments within its existing audience to better serve them.
Low-incidence groups can also appear in larger studies, where brands must account for smaller audience segments, such as power users for a tech product, superfans for a sports or entertainment product, or specialty customers in finance.
The challenge is that small, hard-to-reach populations are needles in a haystack. Many suppliers lack sufficient audience access to reliably source such populations, making niche audience research costly at best, and unfeasible at worst.
But the importance of this research goes beyond “niche” audiences. As brands grow, their customer bases naturally diversify into distinct segments. Low incidence market research is key to understanding and serving the variety of those customers’ needs.
What Low-Incidence Audiences Look Like
Low-incidence groups are often defined by demographics, such as age, income, or household makeup. But “low-incidence group” can describe any scarce group, such as:
- People employed in niche industries (B2B vertical marketing, specialty finance)
- People earning above a certain threshold
- Anyone living with a rare medical condition or disability
- People with experience with a niche product or service (e.g., a regionalized or state-specific credit card)
Notably, groups can be low incidence because of one qualification or many. Consider a study of women who:
- are over 65,
- are still employed full-time,
- have at least one dependent living at home, and
- bank at a local credit union in specific counties across the American Southwest.
Each additional qualification further shrinks the incidence, making this a complex group. By contrast, the scarcity of people with a rare illness, or people who use a state-specific credit card, can be traced to just one quality.
Still, reaching and representing either group boils down to the same capabilities: reaching large numbers of people, and targeting individuals based on highly granular attributes.
Today, brands build audience personas with upwards of 25 dimensions, including psychographics, attitudinal data, and category experience. This is far beyond mere age, gender, and household income. But traditional market research has struggled to keep up with this level of specificity, and to produce the reach necessary for a statistically valid representation of these groups.
Why Suppliers Struggle to Recruit Low-Incidence Populations
The hallmarks of a healthy sample supplier are reach, representativity, and replicability: the ability to access enough people to produce statistically relevant samples (reach), produce a recruitment pool that reflects the composition of the population being sampled (representativity), and deliver the same sample profile again using fresh respondents (replicability), a crucial trust indicator.
For low-incidence audiences, reach is crucial. A supplier that can reach more people is more likely to reach scarce consumers, and can build a pool big enough to represent low-incidence groups within a larger population. Bigger haystack, more needles: the haystack is the overall pool, the needles are the low-incidence group, and both are proportional to the overall population being represented.
However, most providers’ pools are too small to represent low-incidence groups within the broader population. Willing survey participants are a limited resource. If one in every 200 women over 65 banks with a credit union, few suppliers can find the several hundred participants needed for a statistically valid group. Even fewer could deliver fresh sample for a tracker with this population over time.
But reach alone isn’t enough. Finding those rare audience members takes smart, relevant, and respectful screening to match the right individuals to particular projects. Suppliers are often forced to rely on blunt questioning to identify low-incidence participants. Combined with cash incentives, this is a recipe for fraud, and it drives many common data quality issues in research on these groups.
How Research For Good Recruits Low-Incidence Audiences
Our audience sourcing model is built on the premise that reaching people where they already spend their time online supercharges reach.
This approach has netted us more than 27 million verified human respondents, about whom we’ve collected 2 billion data points and counting.
We partner with communities, platforms, sites, blogs, ecommerce brands, and more for access to their users. These sites offer surveys as a way for users to earn rewards within the platforms. Respondents stay engaged in the environment they choose to spend time in, and the survey incentive fits that environment.
Traditional research forces new participants to endure a lengthy, sprawling barrage of profiling questions at the very first engagement. RFG works differently. Respondents are profiled selectively and pre-screened for in-the-moment survey matches, then profiled more deeply over time, in a friction-free experience for the person entering the system.
In subsequent engagements, the system builds on what is already known about the participant. They are offered only active research opportunities that are a fit for them. When an opportunity or exercise requires more in-depth targeting, we can ask the participant live, during that session, to confirm they are a match before routing them to a particular research activity.
This is powered by two proprietary systems: DataForce, our proprietary platform, matches people to research activities in real time, drawing on what we know about each participant and what each project requires. QuIP, our persona-based quality layer, evaluates whether that person is a genuine fit for the activity, going beyond standard pre-screening and fraud checks. DataForce finds the right person for the project, and QuIP helps make sure they belong there.
Our system can also prioritize matching a person to the study for which they are uniquely qualified when it has identified them as part of a hard-to-reach population. In other words, when someone in our database qualifies for a low-incidence audience study, that comes first.
The people we engage with are multifaceted and may qualify for a number of open surveys or research activities at any given time. To return to the earlier example: a fully employed Southwestern woman over 65 banks at a credit union and has a dependent at home. She opts into a research activity. Given a choice between a study that any average woman would qualify for and a study focused on this niche group, the system routes her to the latter.
Across multiple engagements, we gather more and more attributes about each participant. As these attributes accumulate, we’re more and more capable of finding highly qualified study participants. Whether someone is answering a survey, completing an AI training task, or joining an AI-moderated interview, the challenge is the same: reaching the right person.
Engaging with people in these spaces is how we cast as wide a net as possible. Across our pool of 27 million participants, we continuously gather people who will enthusiastically and reliably take surveys. We also learn enough about them to identify and activate needles in the haystack when brands go searching.
The Benefit for Brands
We can give brands a clear idea of research feasibility for any low-incidence audience research scenario. The matching algorithm tracks who it has seen, how often they appear, and how likely they are to return. That lets it predict feasibility for a specific project before it launches, so a client learns whether a given population can be delivered, and at what pace.
When that answer reveals a gap between the client’s incidence assumptions and what’s achievable, RFG can propose an alternative approach.
For repeat or longitudinal work, drawing from a broader base with constant refresh allows us to supply a sample frame across multiple rounds without returning to respondents from prior waves. By contrast, most suppliers typically exhaust their supply in a round or two.
Most importantly, with more than 1,100 data points for persona-based vetting and behavioral segmentation, brands can trust that these audience members are who they say they are.
And while synthetic personas should never be thought of as a replacement for genuine audience members, the real human data that we supply to AI and LLM developers will create synthetic resources that give brands high-fidelity representation of the people they have in mind when they develop new products, services, and marketing initiatives.
Different Model, Different Relationship to Audiences of All Sizes
As our industry’s views and capabilities around low-incidence audiences evolve, brands may come to think of low incidence as a measure of individuality. People’s lives are a mix of the universal and the specific. Each person can naturally be organized by common-denominator demographic attributes, but their unique choices, preferences, and circumstances make them individuals within the broader population. These are the qualities we’re increasingly able to track.
In practical terms, that means that with the right supplier, a brief can ask for the audience a project actually needs, not the closest one they can find. When a project depends on a rare audience, ask how deep a supplier’s targeting really goes, how they confirm that respondents fit the profile, whether they can tell you before launch if the audience is deliverable at the pace you need… and perhaps get curious about what else they can tell you about them.
