When Global Research Is Precise in the Wrong Places.

A global study can have a substantial sample in every country and still be poorly designed for the decision the business ultimately needs to make.

The instinct to allocate an equal number of interviews to every market is understandable. It creates symmetry, makes country comparisons straightforward, and can feel like the fairest way to represent a global footprint.

But equal allocation is not automatically the most useful allocation. The right sample depends on what the research needs to tell the business, which audiences need to be understood, and at what level the resulting decisions will be made.

The starting point for global sample design, therefore, should not be the map alone. It should be the decision.

Start with the decision, not the geography

Consider an industrial technology company researching eight markets with 1,600 interviews available. The business wants to understand the opportunity for a specialist application such as predictive maintenance and determine where to focus future product and commercial investment.

Study A: 200 interviews per country

One approach would be to allocate 200 interviews to every country. This creates a balanced design and provides a strong basis for understanding broad market-level patterns.

But suppose only one in five respondents is directly involved in the relevant purchase or decision. The specialist audience then represents around 320 respondents across the entire study, approximately 40 per country, before considering further differences such as industry, company size, buyer role, or existing technology.

The study still contains 1,600 interviews. On paper, that looks substantial. But the sample available to answer the specialist business question is much smaller.

A sample can be perfectly appropriate for one analytical purpose while being insufficient for another.

Study B: What happens when the sample follows the decision?

An alternative design might allocate 100 broad interviews to each market, retaining coverage across all eight countries, while directing the remaining 800 interviews toward the relevant buyer, industry, or application groups.

The two studies contain exactly the same total number of interviews. What changes is where the analytical strength sits.

The first design emphasises country-level analysis. The second creates greater depth among specialist audiences, which will influence product and investment decisions.

Country findings based on 100 interviews would need to be treated appropriately as directional. But the larger specialld allow the business to explore whether needs are concentrated within certain industries or use cases, whether buyer roles have different priorities, and whether an application has potential across multiple markets.

Neither architecture is inherently better. The right one depends on the decisions the research needs to support.

Beware of false precision

One risk in global research is that a large overall sample can create a sense of precision that does not extend to the questions that matter most.

A study with 200 respondents in Germany, France, and Japan may produce robust-looking country tables. But if the recommendation ultimately depends on 35 aerospace decision-makers, 42 users of a particular application, or 50 senior purchasing executives spread across those markets, the apparent precision can be misleading.

This is particularly relevant in specialised B2B research, where samples may be divided across countries, industries, applications, company sizes, and decision-making roles.

Every additional cut reduces the base available for analysis.

So rather than considering total sample size in isolation, ask: What will the bases look like when we analyse the groups that will actually drive the recommendation?

Design backward from the analysis

One of the most useful exercises when developing a sample plan is to think about the analysis before finalising the sample.

What tables will you want to see? Which groups will you want to compare? Which differences could change the recommendation?

If the business wants to compare four industries, three applications, different buyer roles, and eight countries, those requirements should inform the sample architecture before fieldwork begins. Otherwise, a seemingly large sample can quickly fragment into dozens of small bases.

This does not mean every possible subgroup needs a robust sample. It means deciding upfront which comparisons must be robust and which can reasonably be directional.

Incidence matters, too

The expected incidence of a critical audience should also be considered before committing to the sample structure.

If the buyers who matter most represent only 10% of the broader population, simply increasing the general market sample may be an inefficient way to reach them. Targeted recruitment or an intentional oversample may provide much greater analytical value.

Understanding likely incidence early helps determine whether the sample should prioritise representative market coverage, targeted audience depth, or a combination of both.

Representation and analytical power are different things

Disproportionate sampling can be valuable when a strategically important group represents a relatively small proportion of the market.

Researchers may deliberately oversample a specialist industry, application, buyer role, or customer type so there is enough sample to analyse that group meaningfully.

Where appropriate, weighting can restore those groups to their correct proportions when producing overall market estimates. But weighting does not create information that was never collected. A very small subgroup remains a very small subgroup.

Sometimes the research needs both: a representative view of the market and enough additional sample among priority audiences to understand them properly.

Every sample design involves trade-offs

With a finite research budget, increasing analytical depth in one area usually means reducing it somewhere else, or increasing the overall investment.

If country-level results will determine local investment, those country bases may deserve priority. If the decision is which industry to pursue globally, industry depth may matter more. If the business needs to understand a specialist buyer group that exists across markets, a strong cross-market sample of those buyers may be more valuable than maximising every individual country base.

Good sample design is not about making every possible cut equally robust. It is about making deliberate choices about where precision matters most.

When geography should drive the sample

None of this means substantial or equal country samples are inherently wrong.

If research will inform market entry, local pricing, country-level brand performance, sales resource allocation, regulatory strategy, or market-specific product decisions, robust country-level evidence may be exactly what is required.

Often, the best solution is a hybrid: sufficient sample to understand important geographic differences, combined with deliberate oversamples of the industries, applications, customer segments, or buyer groups most important to the broader business decision.

Four questions to ask before approving a global sample plan

Before fieldwork begins, research teams should be able to answer four questions:

1. What decisions does this research ultimately need to support?
2. At what level will those decisions be made - country, industry, application, customer segment, buyer type, or some combination?
3. Which comparisons require robust evidence, and which can be directional?
4. After accounting for incidence and planned subgroup analysis, will the audiences that matter most have sufficient bases?

Put the precision where the decision is

There is no single correct architecture for a global research study.

A beautifully balanced sample can still be misaligned with the business question. Conversely, a deliberately unbalanced sample can be exactly the right design when it puts greater analytical strength behind the audiences and opportunities that matter most.

The goal should be to understand where the business needs confidence, design the sample accordingly, and be transparent about what the resulting evidence can, and cannot support.

The sample should follow the decision, not the map.