Original Research & Newsworthy Campaigns / Find a Research Question Worth Covering / Lesson 02

Module 01 / Lesson 02 · Read, apply, check

Choose Between Client Data, Surveys, Benchmarks and Experiments

Choose a research method that fits the question and distinguish observations, perceptions, benchmarks and causal tests.

By Adriaan BritsCEO of Sitetrail · MSc Marketing
Former LinkedIn Learning instructor in SEO, Advanced Google Analytics and Google Ads
Course edition: October 2026

Use operational records for observable behavior

Client data can reveal events that already occurred, such as request times, stated deadlines and recorded replies. First inspect whether the fields mean what the research question assumes and whether the records cover the relevant activity. A platform may capture only enquiries handled through its own workflow, excluding telephone or personal email responses. That omission affects interpretation. Operational records can support a valuable account of the observed system without representing every customer or supplier. Define the unit clearly: a request, organization and person are different things. Several requests from the same buyer do not become several independent buyers simply because each occupies a row.

Use surveys for reported experiences and views

A survey can ask people about experiences, priorities or perceptions that a transaction log does not capture. Write questions that respondents can understand and answer from their actual knowledge, with a suitable recall period and response options. Check whether recruitment reaches the audience you intend to describe. An opt-in survey of newsletter readers may be useful for exploring their experiences, but it is not automatically a population estimate. Keep reported dissatisfaction separate from an operational measure of lateness. Those questions can complement each other without being interchangeable or combined into one headline percentage merely because both concern the same general subject.

Use benchmarks and experiments for different purposes

A benchmark compares a defined measure across comparable groups or periods. It needs consistent definitions, coverage and collection conditions; a change in the recording system can resemble a performance trend. An experiment tests the effect of an intervention using a design capable of addressing alternative explanations. For example, a properly designed comparison of follow-up messages might investigate response behavior, but simple before-and-after observation cannot isolate the message from seasonal workload or changed suppliers. Involve an appropriately qualified researcher when causal claims, weighting or complex uncertainty are central. Choose the method for the decision required, rather than attaching scientific language to a convenient campaign asset.

Combine methods without merging their meanings

A small project may use records to describe observed events and interviews or a survey to explore possible explanations. Give each source a distinct job, its own population, period and limitations. Do not treat agreement as automatic causal proof, and investigate disagreement rather than hiding it. Estimate the cost, access, review time and analysis skills each method requires. A modest well-documented study can be a better starting point than an expensive survey that does not address the real question. The method choice should end with a clear statement of what the team expects to learn and what it will still be unable to claim.