A global food and beverage manufacturer, working through its United States coffee portfolio, wanted to understand how people use coffee at different moments of the day and how brand choice changes between those moments. The research ended in an in-home use test, but the hard part came first: finding the right people. This case study describes the in-home use test recruitment Research and Metric delivered, and how a tightly layered sample was filled cell by cell.
The client had already built its segmentation. Our role was fieldwork: recruiting to the segments, incidences, demand spaces and typing battery the client supplied.
The challenge
The client needed to hear from two distinct consumer mindset segments, both defined in advance, and to see how each used coffee across different occasions. That meant one recruitment programme had to deliver three things at once:
- a general qualifying audience of US coffee purchasers and drinkers;
- two pre-defined mindset segments, typed using the client's own battery;
- eight separate demand-space cells inside one of those segments.
The audience was specific. Respondents were aged 18 and over, had primary or shared responsibility for buying coffee in the household, drank coffee at least three times a week, had a household income of $35,000 or above and were college graduates or above. The category covered coffee and coffee whiteners, so whitener behaviour mattered as well as coffee behaviour.
Why the numbers were difficult
On paper, the segment incidences looked workable: 17 percent for Segment 1 and 14 percent for Segment 2. The difficulty came from stacking requirements. Once a segment qualifier was crossed with a demand-space allocation, a brand non-rejection filter and a brand usage requirement, the effective incidence on the smallest cells fell into low single digits.
Both segments carried the same brand conditions: non-rejection of a named brand, and usage of a minimum number of that brand's products. Each condition is reasonable alone. Together they narrow the pool sharply.
The sample structure made this more demanding. The target was N=100 completes, with general criteria balanced 50 percent female and 50 percent male:
- Segment 1: 24 completes across eight demand-space cells, with targets of 5, 2, 3, 3, 2, 3, 3 and 3.
- Segment 2: 20 completes in one demand space.
- General qualifying sample: the remaining 56 completes.
Several cells needed only two or three completes. That sounds easy, but a small cell at very low incidence is harder to fill reliably than a larger cell at a higher incidence. With so few completes required, there is no room for a single dropout. One respondent who fails to complete the in-home test leaves a cell short, and finding a replacement who fits every layer is slow.
If you are planning a similar study, our guide to survey incidence rate and how it shapes a fieldwork quote explains why stacked criteria affect feasibility so strongly.
Our approach to in-home use test recruitment
This was segment-screened recruitment followed by an in-home use test with guided participant activities. The screener was the foundation, so we designed it carefully.
A four-layer screener
Research and Metric built a four-layer screener, with each layer doing a separate job:
- Category qualification: confirming the respondent bought and drank coffee at the required frequency and met the household responsibility, income and education criteria.
- Segment typing battery: placing qualified respondents into one of the two pre-defined mindset segments, using the battery supplied by the client.
- Demand-space allocation: assigning respondents to the relevant demand-space cell.
- Brand non-rejection and usage: checking the brand conditions that both segments carried.
Coffee whitener behaviour and brand usage were captured at the screening stage, so the data was recorded once, up front, rather than reconstructed later.
Behaviour first, demographics for balance
Respondents qualified on behaviour throughout. The demand-space cells were defined by occasion and behaviour, not by demographics. Demographics were captured for quota balance, including the 50/50 gender split, rather than used as filters. This kept the screener focused on how people actually buy and drink coffee, and avoided excluding people who fitted the behavioural definition.
Over-recruitment set cell by cell
A single blended over-recruitment ratio would have been the simplest option, but it would not have reflected the problem. A cell at a comfortable incidence and a cell at low single digits do not behave alike. We therefore set over-recruitment for each cell individually, and revised it as real incidence emerged during fieldwork. Planning assumptions gave way to observed performance as soon as we had it.
Flexible sourcing
Sourcing was not fixed at the start. As each cell filled, we shifted between four routes:
- Research and Metric's own consumer panel;
- its proprietary database;
- social recruitment;
- referrals.
Different routes suit different cells. Using several, and moving between them, is a core part of how we handle hard-to-reach and specialty audiences, including consumers who are not hard to reach in general but become so once the criteria are stacked.
Fieldwork and quality
Quality in a study like this depends on control at cell level. A healthy total can hide a cell that has stalled, so the total was not the measure we managed to.
Daily, cell-level monitoring
Cells were monitored daily and individually, not against a single total. This allowed us to see which cells were progressing and which needed a change of source or an adjusted over-recruitment target. For low incidence recruitment, early warning matters more than any later correction.
A phased delivery
Delivery was phased. The first phase was an initial 50 completes, and the balance followed. The initial 50 gave a real read on how each cell performed, rather than relying on projected incidence alone. It also let us adjust the replacement protocols before the second half began. Replacement is where small cells are most exposed, so tuning it on real evidence reduced the risk for the remainder of the sample.
Screening discipline
Because every cell depended on several layers of qualification, the screener had to be applied consistently. Respondents were typed and allocated using the client's supplied battery and definitions, and the brand conditions were checked at screening. Our work follows the ICC/ESOMAR International Code and ISO 20252, and our information security practices follow ISO 27001. Data protection requirements, including the CCPA for US respondents, apply to how respondent information is handled.
Once the sample was in place, the in-home use test was fielded with guided participant activities. For background on how these tests are run well, see our IHUT playbook, or our page on product placement and in-home use tests.
What the client received
The deliverable was a full sample across every segment and demand-space cell. In practical terms, the client received:
- the full N=100 completes, delivered in two phases;
- Segment 1 completes in each of the eight demand-space cells;
- Segment 2 completes in its single demand space;
- the general qualifying sample, balanced 50 percent female and 50 percent male;
- screening data on coffee whitener behaviour and brand usage;
- an in-home use test fielded with all participant activities completed.
Every cell was filled, including those that needed only two or three completes. That was the central requirement of the brief, because a missing cell would have left a gap in the comparison between mindsets and moments of the day.
What this project shows
Recruitment to an existing segmentation is a fieldwork discipline in its own right. The client supplied the segmentation and the research questions. Our contribution was to turn those definitions into a screener, a sourcing plan and a monitoring routine that could deliver the people the design needed.
If your study stacks several qualifiers and needs small cells filled with certainty, speak to us about hard-to-reach audiences or talk to an expert about your design.
Client project. Client details are anonymised.
