Primary Data Collection Techniques: A Fieldwork Guide to Surveys, Interviews, Groups and Observation

Primary data collection techniques all look simple on a slide: run a survey, hold some interviews, watch people shop. Making any of them work in the field takes more planning. You need the right people, recruited at the right cost, with consent recorded and the data kept in usable shape. This is a practitioner's guide to the main primary data gathering methods, written from the fieldwork side: what each one is good for, what it needs to run, and where it usually goes wrong.

If you want the wider picture first, including how primary and secondary data relate, start with our overview of data collection methods in research.

Before you pick a technique: three questions

Primary research is expensive compared with reusing existing data, so it pays to be clear about why you are collecting it. Three questions settle most debates before they start.

1. What decision will this data support? "Understand the customer" is not a decision. "Choose between two pack designs for launch in March" is.
2. Do you need to measure, or to understand? Measurement points towards structured techniques such as surveys. Understanding points towards conversation and observation.
3. Who exactly has the answer? A general consumer, a hospital pharmacist and an IT decision-maker in a mid-sized firm need very different recruitment routes.

The core techniques, field by field

Each technique below is described in the same way: what it needs before fieldwork can start, and where it earns its place. The point is less to rank them than to show what each one asks of the project team.

Surveys

Surveys collect structured answers from many people, which makes them the default for sizing markets, tracking metrics over time and comparing segments.

What it needs to run:

• A questionnaire tested for flow, length and wording.
• A sample source: an online panel, a client customer list, or interviewers working from a frame.
• Quotas that reflect the population, set up so fieldwork can be steered while it runs.
• Incentive rules and a plan for checking response quality.

Best for, and what to watch out for: Surveys are best when you already know which questions matter and need reliable numbers. They struggle when the topic is new to you, because a closed question cannot capture an answer you did not anticipate. Long grids and repetitive scales invite fatigue; the checks that catch it are covered in our post on online survey data quality.

In-depth interviews

One-to-one interviews, in person, by phone or on video, let a skilled moderator follow the respondent's reasoning and probe on the things that matter.

What it needs to run:

• A discussion guide that sets topics without scripting every word.
• Careful screening, since a handful of interviews cannot absorb a wrongly recruited respondent.
• Scheduling, recording consent and a plan for transcripts and notes.

Best for, and what to watch out for: Interviews are the natural choice for B2B, expert and healthcare professional research, where respondents are scarce, busy and knowledgeable. They also work well for sensitive topics that people will not discuss in a group. The risk is over-reading a few voices; interviews explain patterns well but cannot size them.

Focus groups

Focus groups bring six to eight people together, physically or online, for a moderated discussion. Their value lies in interaction: people react to each other and to stimulus such as concepts, ads or packaging.

What it needs to run:

• Recruitment that balances the group without mixing people who will silence each other.
• Over-recruitment to cover no-shows.
• A venue or online platform, a moderator and, often, client viewers.

Best for, and what to watch out for: Groups are efficient for early concept work and for hearing the language customers use. They are weaker for individual behaviour, and a dominant participant can steer the room. A good moderator manages that; a poor recruit makes it worse.

Observation

Observation records what people do rather than what they say they do. It ranges from accompanied shopping trips and in-home visits to usability sessions and analysis of behaviour logs.

What it needs to run:

• An observation protocol listing what to record and how.
• Clear consent, particularly for video or photography in homes and workplaces.
• Time: observation produces rich material that takes longer to code than survey data.

Best for, and what to watch out for: Observation is strongest where claimed and actual behaviour differ, such as habitual purchases, workarounds in software or how products are really used at home. The main risk is that people change their behaviour when watched, so sessions need enough time for them to settle.

Diaries and longer-term tasks

Diary studies and online communities ask participants to record experiences over days or weeks, often with photos or short videos. They sit between surveys and interviews: structured prompts, open responses, collected over time.

They suit routines, journeys and anything that happens too occasionally to catch in a single session. Drop-off is the main challenge, so incentives are usually staged and engagement is checked daily.

Product and in-home usage tests

In a usage test, participants receive a product, use it in normal conditions for a set period and then report back through a survey, a diary or an interview. It is one of the few techniques that captures reactions after real use rather than first impressions.

What it needs to run: product logistics, including shipping, storage and returns; a screener that confirms people actually use the category; and safety or allergen information where relevant. Respondent addresses become part of the project data, so they need the same care as any other personal information.

Best for, and what to watch out for: formulation changes, new variants and anything where repeated use matters. Logistics are the usual failure point. A courier delay can push the whole timeline back, so build slack into the schedule.

Intercepts and on-site interviews

Intercept interviews catch people at the moment of interest: leaving a store, visiting an event or using a service. They capture fresh memories, but the sample is shaped by where and when interviewers stand. Record locations, times and refusals so the sample can be described honestly later.

Comparing what each technique asks of fieldwork

The table below looks at the practical demands each technique places on a project team, rather than the kind of insight it produces.

Sample sizes here are deliberately described in relative terms. The right number for any study depends on the population, the number of subgroups you need to read and the precision the decision requires.

Recruiting the right respondents

Across all of these primary data gathering methods, recruitment quality decides the quality of the output. A few habits make a large difference.

• Write the screener from the sample design, not from memory
• Check feasibility per country and segment before quoting a timeline
• Agree incentive levels that reflect respondent effort and seniority
• Track every respondent ID back to its source
• Re-screen high-value qualitative recruits before the session
• Record consent in a form you can produce later

Consent and transparency are not optional extras. Professional codes such as the MRS Code of Conduct set out the standards research practitioners are expected to maintain, including honesty with participants about the purpose of the research.

For specialist audiences, recruitment is often the longest part of the project. Our guide to B2B and healthcare research sample covers the extra steps those audiences need.

Combining techniques

Few projects rely on a single technique. A common pattern is to start with a small number of interviews to learn the language and the issues, then field a survey to measure how widespread they are, then return to a few respondents to explain a surprising result. Sequencing techniques like this is the core of mixed-methods research, which we cover in more detail in quantitative and qualitative data collection techniques.

1. ExploreA few interviews or groups to frame the problem in customers' own words.
2. MeasureA survey with quotas to size what the exploration found.
3. ExplainFollow-up conversations with selected respondents where the numbers surprise you.

Where Sampledge fits

For online surveys in particular, the slowest part is often getting from "we need 500 completes from this audience" to a live, priced and trackable fieldwork job. SampAudience is Sampledge's DIY sample marketplace for that step: you define the audience, configure the study, check feasibility and cost, purchase, launch and then track progress, across B2B, B2C and healthcare audiences. When the project grows to several suppliers and quota cells, SampTrek keeps suppliers, live fieldwork and reconciliation in one project record.

To try it with your own study, create a Sampledge account, or talk to our team about a specialist audience.