How to do sampling in market research, with best practices for B2B

Sampling can make or break any market research project. Put simply – if you include the right people, well-designed research can get you incredible insights to inform your strategies.

However, if you include the wrong people, it can ruin your results. There is plenty of bad research out there and it’s often because the wrong people have been interviewed.

Having the right sampling process in market research means:

1. Identifying the most relevant audience to include
2. Ensuring the right ratio or representation of sub-groups
3. Speaking to enough people to get robust, reliable data
4. Verifying that the respondents are genuine

Often, point 1 is straightforward. For example, if you’re conducting B2B customer satisfaction research, your customers are the relevant audience. But for other projects, such as market assessment research, the target population is less clear-cut.

In this article, we’ll run through points 3 and 4 in our best practices section. These are crucial factors to get right, but industry viewpoints are varied – we’ll share our recommendations.

Usually, when researchers talk about sampling, they’re thinking about point 2 – weighing up which sampling technique will ensure fair, unbiased representation. We’ll explore some of the key considerations and methods in the next two sections.

Sampling considerations in B2B market research

Getting the right mix of different segments in a bespoke research project is a very careful balancing act. It takes a lot of experience to estimate which combination will deliver the best results.

Let’s take an example. Among other criteria, in B2B research, it’s common to sample and analyze the final data based on the company profile (e.g. industry, size, and so on).

Imagine you’re conducting some exploratory market assessment research to get feedback from potential buyers about your new product idea. In this case, it’s a SaaS solution.

When sourcing B2B respondents to take part, variables can include country, industry, company size, department, and role seniority. That’s a lot already.

And within each variable, there’s plenty to consider. Let’s take company size as an example. 

There are two main ways to define company size – either by annual revenue or by the number of employees – so already, there’s another call to make there. You could aim for a sample split by country size that’s broadly representative of the market – taking the Gartner glossary definitions of business sizes:

• Small businesses have under 100 staff, or below $50m in annual revenue
• Midsize ones have 100-999 staff, with between $50m and $1bn in annual revenue
• Enterprises have 1000 or more staff and $1bn+ in annual revenue

According to NAICS data, 89% of businesses in the US have under 100 staff. Using a more approximate calculation, 89% also have under $50m in annual revenue. The vast majority of other businesses are midsize, with enterprises under 1% by headcount or annual revenue.

So, in your research sample, you could aim for 89-90% of respondents to work at small firms, about 10% at midsize ones, and just a few at enterprises. However, the downside is you won’t get many – or any – insights about enterprises, due to low sample size.

Moreover, what if you’ve previously identified midsize businesses as the primary customer base for your new SaaS solution? In that case, perhaps you’ll get more value out of the research if you aim for a roughly 50:50 split of small and midsize businesses, with a few enterprises too.

And what if secondary research tells you that enterprises provide 40% of all market spend on similar products? Arguably, it may make more sense to get 40% of your responses from larger businesses.

There are implications here too though – for example, you might need to set expectations with stakeholders. Some may be expecting a 90:10 representation of small to midsize businesses in the research, to more accurately reflect the general market.

This is just one example of the many different considerations around sampling in market research. And in B2B research, there’s a further complication regarding sample sizes.

As we’ll discuss in more detail later – in any country, genuine B2B decision-makers are not easy to recruit. This can make it challenging to get enough sample to filter by multiple criteria such as industry, company size, department, and role seniority.

Different sampling techniques in market research

Broadly speaking, there are two main approaches to creating a sampling frame. These are probability sampling methods and non-probability ones.

Probability sampling methods involve the random selection of respondents from an audience, to closely resemble the target market.

In contrast, non-probability sampling is not based on random choice. It’s a more selective way of choosing research respondents based on a set of predetermined criteria.

In B2C market research, probability sampling is popular – in particular, if the target audience is the general population, or most of it – e.g. smartphone users.

Probability sampling is possible in B2B research, with a large enough database. For niche audiences though, often non-probability sampling is the only viable option.

The following is a list of common sampling techniques, arranged in terms of relevance or suitability to B2B research, ending with the most appropriate:

• Convenience sampling
• River sampling
• Nationally representative (nat rep) sampling
• Judgment sampling
• Simple random sampling
• Systematic sampling
• Stratified or cluster sampling
• Quota sampling

Let’s run through the differences:

#1 Convenience sampling

Recruiting based on choosing those actively volunteering to take part is called convenience sampling.

For example, you might advertise the research and accept everyone who replies, with little to no screening, or minimal eligibility criteria. There tends to be little to no financial incentive for participation – respondents are volunteers, after all.

This is arguably the lowest-effort form of sampling and naturally, it has the most limitations as a result. You’re very unlikely to get a representative target audience, across your key segments, with convenience sampling.

This recruitment method is also subject to volunteer bias – those replying are more likely to do so because they’re interested in or even passionate about the subject matter. This could skew the results, as you’re less likely to get the more middle-of-the-road viewpoints.

#2 River sampling

River sampling involves creating a sample by inviting respondents to research while they are already online, engaged in other activities.

For example, you can use online banners, promotions, ads, or other offers to provide invitations. 

You place these CTAs in relevant places for your target audience – e.g. a branded website, industry trade publications, and so on. 

The idea is that interested respondents should fit the profile of who you’re looking to speak to, because they were already looking at relevant material. However, you still need to screen them to make sure.

#3 Nationally representative (nat rep) sampling

This is a recruitment method based on country-specific sampling. It’s common in consumer research projects.

The specific variables will vary by country – a nat rep sample in the US is different from one in the UK, for example. Therefore, you need to be very precise to avoid making a sampling error.

Nat rep sampling broadly reflects a country’s population based on basic demographics, such as people’s age, gender, region, occupation, and household income.

B2B research does not typically use nat rep sampling though, because consumers aren’t the target population or audience.

#4 Judgment sampling

Recruiters primarily use their judgment to select respondents via this sampling method.

For example, if a respondent has a useful example or perspective to share, differentiating them from others, a recruiter could make a judgment call to prioritize this individual.

However, this is a very subjective and selective approach to sampling. A more objective approach is to base recruitment on impartial screening criteria, first and foremost.

Nevertheless, good screening often uses some judgment. If respondents meet all the eligibility criteria and fall into the same segment, but A can articulate their viewpoints more clearly than B, recruiters should prioritize A for a better chance of richer insights.

#5 Simple random sampling

A simple random sample removes the bias of judgment sampling by using a random number generator, or an equivalent randomized selection method, to select respondents.

In B2B research, this is often only feasible using a large CRM database of customers, who have opted in to be contacted about market research projects. 

Otherwise, it’s rare to have a big enough sample of genuine respondents meeting all the eligibility criteria to have the luxury of selecting them at random.

#6 Systematic sampling

Systematic sampling takes simple random sampling a step further and is the more common approach. 

If you want to survey 1% of your database, you select every 100th customer until you reach the overall target.

However, neither simple random nor systematic sampling lets you ensure a representative sample of a B2B customer base. There’s a risk that the randomness will skew towards one particular segment over another.

#7 Stratified or cluster sampling

Stratified sampling corrects the limitations of simple random and systematic sampling, by first separating eligible research participants by segment.

Then, you recruit respondents systematically and proportionally, until you meet all the set quotas.

For example, you could first segment the audience by industry, company size, department, and role seniority. Another way to segment in B2B is by separating different participants in the B2B purchase process – the buyer, purchase influencers, and product end-users.

Cluster sampling uses a similar principle. Technically, it involves first dividing the audience into smaller clusters, each one representative of the makeup of the overall target population. Then you take a random sample.

These are fair, logical sampling techniques – but again, often only possible in B2B research by using a large CRM database.

#8 Quota sampling

A very common type of recruiting for B2B research, quota sampling uses similar principles to stratified sampling, yet it’s not a probability-based sampling technique.

This recruitment method focuses on selecting respondents by segment, to conduct fieldwork with as representative a sample as possible of your target audience.

If a large CRM database isn’t available for market research purposes, quota sampling is the next best way to recruit as per your segmentation. It just lacks the systemic aspect – i.e. there is no large database to work with, letting you reach out to every 100th customer.

Instead, the sample tends to be built from scratch, using sophisticated recruitment techniques to find high-quality and reliable B2B research participants. 

An experienced researcher maintains high overall standards, keeping to the target quotas without making any sampling errors.

Best practices for the recruitment stage in B2B research

#1 Try snowball sampling to leverage network relationships

For qualitative research projects, sampling a hard-to-reach or niche audience, snowball recruitment can be an efficient way to find more senior B2B respondents.

It’s a more informal sampling technique and usually, not one you plan to use at the start of a project. It’s usually a backup method for finding sample or an extra, bonus way of recruiting.

You ask insightful respondents if they can suggest others to interview. If they’re particularly helpful, they’ll make an introduction and recommend to those people that they speak to you.

In turn, these new respondents may suggest others you can reach out to, and so on. In this way, you leverage senior executives’ experience and network, to get a level of access you wouldn’t achieve otherwise.

While there is a risk of judgment bias in snowball recruitment, it’s usually one worth taking. Customers are well-placed to give you relevant feedback for your research topics, so if you trust their input, you can also trust them to make good recruitment recommendations. 

#2 Use strict screening to only include genuine B2B respondents

An added complication of sampling in B2B research is the hard-to-recruit target market. Industry-specific products and services are more specialized than mass-market goods.

There are fewer B2B buyers than there are consumers and only a small proportion will take part in market research projects. In an attempt to facilitate sampling, there are some B2B research panels – but we wouldn’t recommend using them.

The senior decision-makers you want to speak to are unlikely to be sitting on these panels. Many of them are still selling consumer sample and offering low incentives.

Wherever you source a B2B sample, it’s essential to use several techniques designed to screen out any fake respondents. These include balanced scales, logic traps, ‘red herring’ answers, and open-ended questions requiring industry-specific knowledge.

#3 Build your own database for sampling on multi-stage projects

If you conduct a lot of research projects, you should consider setting up your own database of participants to sample from time and time again.

It’s a great way to ensure respondent quality. These respondents could be buyers in your customer base, or other contacts you can verify as genuine decision-makers (for example, by checking their profile on LinkedIn).

A good way to start is by asking respondents on the next project if they’d consider taking part again in the future – naturally, some say no, but many tend to agree.

This is one of the best ways to do sampling on multi-stage projects – for example, when you want to do both qualitative and quantitative research or iterate something several times.

While some third-party recruiters offer multi-stage sampling, not all do. They tend to prefer finding respondents one stage at a time (while others propose high fees to re-recruit them). 

#4 Be comfortable with lower research sample sizes in B2B

Most consumer surveys are based on large numbers of responses – often, thousands. They are relatively easy to get because most of the general population qualifies as a consumer.

B2B markets are smaller, therefore the typical sample size in B2B research is smaller too – often, hundreds.

This means that in many cases, you can take several hundred completes in B2B quantitative surveys as a robust number.

Just bear in mind that due to lower sample sizes compared to consumer research, without a large database of customers, it’s challenging to get robust data for interlocking quotas.

For example, you may have enough sample to sort the results by industry. But you may not have enough to filter the results by both industry and company size simultaneously.

Some statistical trade-off techniques may not be feasible at the analysis stage because they require larger sample sizes.

It means that the more complicated your sampling method or quota breakdown, the harder it will be. Recruiting at scale in B2B research is challenging, but possible if you have the right smart tools and techniques.