Choosing the Right Research Method: Surveys, Secondary Data, Observations

Every agribusiness decision – from expanding into a new market to testing a new fertilizer – begins with a question. But the quality of the answer depends entirely on how you collect and analyze information. Once a research design is in place, the next critical step is selecting the right research method. Four methods form the backbone of agribusiness research: surveys, secondary data analysis, observations, and experiments. Each serves a distinct purpose, and choosing the wrong one can produce results that are misleading, incomplete, or simply not useful. This post breaks down each method – what it is, when to use it, and what trade-offs to expect.

What are research methods and why does selection matter? 🔗

Research methods are the specific techniques used to gather and analyze data once a research design has been established. The selection of a method is not arbitrary – it is driven by the research objectives, the availability of existing data, the time frame of the study, and budget constraints. According to a data science curriculum by Excelsior University, understanding when and how to apply each approach is essential for strategic data acquisition that balances scientific rigor with practical feasibility. In agribusiness, where production cycles are long and decisions are costly to reverse, choosing the wrong method can have downstream consequences that extend far beyond the research itself.

Surveys: collecting primary data directly from stakeholders 🔗

Surveys are one of the most widely used methods for collecting primary data – information gathered directly for a specific research purpose. They are particularly effective when the goal is to understand the opinions, behaviors, preferences, or intentions of farmers, consumers, suppliers, or other stakeholders.

For example, if a cooperative wants to understand why farmers in a region are slow to adopt drip irrigation, a survey can uncover motivations, perceived barriers, and cost concerns that no secondary source would capture. Research published in the journal Society & Natural Resources confirms that surveys remain a primary method for studying agricultural producers’ natural resource management attitudes and behaviors, with government program participant lists and private vendor samples being among the most commonly used sampling frames.

Types of survey methods 🔗

Survey formats vary significantly and are chosen based on the research context. Personal interviews – whether face-to-face or computer-assisted – offer depth and allow the researcher to clarify questions. Telephone and mobile surveys provide wider geographic reach at a lower cost. Self-administered surveys, including mail and hand-delivered questionnaires, are cost-effective for large samples, though mail surveys often suffer from low response rates while hand-delivered formats tend to perform better.

When designing surveys, question clarity and unbiased wording are critical. As Excelsior University’s data science resource explains, structured questionnaire design requires careful attention to question wording, response format, and survey organization to ensure valid and reliable data collection. A poorly worded survey in agribusiness can generate responses that seem informative but lead to entirely wrong conclusions.

When to use surveys 🔗

Surveys work best when the information you need exists only in people’s minds – their preferences, attitudes, and decision-making processes. They are especially well-suited for market research, needs assessments, and adoption studies. However, they have limitations: respondents may give socially desirable answers rather than honest ones, and self-reported behaviors do not always match actual behavior.

Secondary data: leveraging existing information 🔗

Secondary data refers to information already collected by another party for a different purpose. In agribusiness research, this includes government statistics, industry reports, academic publications, market databases, and an organization’s own internal records such as sales history or financial statements.

The advantages are significant. Secondary data is generally faster and cheaper to access than primary data. It can cover large geographic areas and long time periods that no individual researcher could replicate alone. For instance, the FAO’s FAOSTAT database provides production, trade, food security, and price data for over 245 countries going back to 1961 – a resource no single survey could match. Similarly, the USDA’s Economic Research Service publishes detailed datasets covering farm financial indicators, commodity projections, agricultural productivity, and rural economic conditions across the United States.

Internal vs. external secondary data 🔗

Secondary data falls into two broad categories. Internal sources include data held within an organization – financial records, operational logs, past sales reports, and customer transaction histories. These are particularly useful for trend analysis within a business. External sources include government publications, peer-reviewed journals, trade association reports, commercial market research databases, and international institution datasets. The University of California Berkeley’s agricultural data guide lists key external resources including FAOSTAT, the USDA National Agricultural Statistics Service (NASS), and the World Bank’s World Development Indicators – all freely accessible to researchers.

Limitations to keep in mind 🔗

Secondary data is not a perfect substitute for primary data. It may not align precisely with the research question, measurement methods may have changed over time, and the researcher has no control over how the data was originally collected. Evaluating source credibility and understanding the original collection methodology are essential steps before using any secondary dataset.

Observation methods: recording what actually happens 🔗

Observational research involves systematically recording behaviors, processes, or phenomena without directly questioning or interacting with the subjects. It is especially valuable when there is a gap between what people say they do and what they actually do.

Consider a study on how farmers apply pesticides. A survey might capture farmers’ self-reported practices, but direct observation may reveal shortcuts taken during busy seasons that would never appear in a questionnaire response. As noted in an analysis by the World Bank’s Living Standards Measurement Study, even when assessing crop variety adoption, enumerator observation often produced more reliable results than basic survey questions alone – highlighting how observational methods can complement or outperform self-reported data in agricultural contexts.

Forms of observation in agribusiness 🔗

Personal observation involves a researcher directly watching and recording behavior in real time – useful for studying planting techniques, market interactions, or harvesting practices. Mechanical observation uses technology such as cameras, GPS trackers, and sensors to collect data continuously and without human bias. Time-lapse cameras can document plant growth, GPS units can monitor livestock movement, and environmental sensors can record soil conditions around the clock.

Audits represent a structured form of observation focused on checking compliance with specific standards at set intervals. Food safety audits and organic certification inspections are common examples in agribusiness. Content analysis applies observational logic to documents, social media, or media content, and is useful for studying how agribusinesses are represented publicly or how market narratives evolve.

When observation works best 🔗

Observation is particularly appropriate when behavioral patterns are the focus and when asking questions would either be impractical or would alter behavior. However, observer bias remains a concern – researchers bring their own perspectives that can influence interpretation. Using structured recording tools, training multiple observers, and using mechanical systems where appropriate helps minimize this risk.

Experiments: establishing cause-and-effect relationships 🔗

Experimental methods are used when researchers need to go beyond describing what happens and determine why it happens – specifically, to establish causal relationships between variables. An experiment involves deliberately manipulating one variable (the independent variable) while controlling others, and measuring the effect on an outcome (the dependent variable).

A review published in Psychological Methods notes that randomized controlled trials (RCTs) are considered the gold standard for causal inference precisely because random assignment to treatment and control groups minimizes bias. In agriculture, this method has a long history. Randomized field experiments for agricultural research date back to at least the 1920s at institutions like Rothamsted Experimental Station in the UK, where statistical approaches to field trial design were first formalized.

On-farm and field experiments 🔗

In agribusiness, experiments are often conducted directly on farms or in controlled plots. Oregon State University Extension provides practical guidance on on-farm experimental design, recommending that researchers use replicated, randomized treatments to ensure that observed yield differences are caused by the treatment – not by pre-existing soil variability or other confounding factors. Common designs include paired comparisons (for testing two treatments), randomized complete block designs (for three or more treatments), and split-plot designs (for testing interacting treatment factors).

SARE (Sustainable Agriculture Research & Education) notes that four to six replications per treatment block are generally sufficient for most on-farm research projects. This replication is essential: without it, a researcher cannot distinguish a genuine treatment effect from natural field variability.

Limitations of experimental methods 🔗

Experiments are resource-intensive. They require careful planning, control of variables, sufficient time to observe outcomes, and often significant investment in land and labor. Research combining randomized experiments with satellite-based observational data points out that while field trials are the gold standard for causal inference, they often suffer from small sample sizes and may not generalize well to farms with different soil types, climates, or management histories. This is why experiments are often combined with observational or secondary data to improve the broader applicability of findings.

How to choose the right method 🔗

No single method is universally superior. The right choice depends on what you need to know, the resources available, and the trade-offs you are willing to accept.

If your goal is to understand farmer or consumer attitudes at scale, a survey is the most direct route. If you need historical context, market benchmarks, or broad trend data without the cost of primary collection, secondary data is the logical starting point – and resources like FAO and the USDA ERS make this highly accessible. When the goal is to understand actual behavior in context – rather than reported behavior – observation delivers more authentic data. And when a business or policy decision hinges on proving that a specific input, practice, or intervention causes a particular outcome, only an experiment can provide that level of certainty.

In practice, many rigorous agribusiness research projects use a combination of methods. A researcher might begin with secondary data to understand the existing landscape, conduct a survey to explore attitudes, observe practices in the field to validate survey findings, and then run a controlled experiment to test a specific intervention. Excelsior University’s analysis describes this integration well: survey research identifies needs and preferences, experiments test specific interventions, and observational analysis reveals how changes affect behavior at scale. Together, these methods create a far more complete picture than any single approach could.

Time and cost constraints are also real factors. Surveys can deliver results quickly; secondary data analysis can often begin immediately if suitable datasets exist. Observations and experiments, especially longitudinal ones, require more time and planning. Budget-limited projects often start with secondary data and surveys before committing resources to the more expensive observational or experimental approaches.

What do you think? If you were designing a study to understand why smallholder farmers in your region are not adopting a proven soil conservation technique, which research method – or combination of methods – would give you the most reliable and actionable results? And how would the time frame and budget of your project influence that choice?

References

1. https://express.excelsior.edu/datascience/chapter/chapter-2-6-data-collection-methods-surveys-experiments-and-observations/
2. https://www.tandfonline.com/doi/full/10.1080/08941920.2022.2081392
3. https://www.fao.org/statistics/data-collection/agriculture/en
4. https://www.ers.usda.gov/data-products
5. https://guides.lib.berkeley.edu/ARE/finddata
6. https://www.worldbank.org/en/programs/lsms/priority-themes/survey-methods/agriculture
7. https://pmc.ncbi.nlm.nih.gov/articles/PMC6086368/
8. https://extension.oregonstate.edu/catalog/pub/em-9381-experimenting-farm-introduction-experimental-design
9. https://www.sare.org/publications/how-to-conduct-research-on-your-farm-or-ranch/basics-of-experimental-design/common-research-designs-for-farmers/
10. https://arxiv.org/pdf/2112.13700
11. https://www.fao.org/countryprofiles/data-sources/en/