Original research for digital PR

Original research can give a digital PR story a finding readers cannot get from a company announcement. Start with a question people outside the business need answered, choose a method that can answer it, and keep a record of how the evidence was produced. Publish only a claim the results support.

A data-led story can position a brand as a primary source, but its value rests on a credible, interesting finding that supports a wider story. Earned media and reputation matter; a backlink is not the sole objective.

From question to published claim

Decide what the research needs to establish

First decide whether the aim is to describe what a defined group reported, count events in a defined set of records, or examine a change over time. These questions call for different evidence. A survey about reported experience cannot, by itself, establish what people actually did; a comparison across periods cannot establish why a change happened without further work.

Define the population, unit and period in the brief. “Australian customers who used this service during the study period” is a different group from “Australians”. If the business can reach only its own customers or volunteers, plan language that stays within that group; more responses do not, on their own, remove a selection problem.

Choose the method to fit the question. Interviews can explore how participants understand a process, but do not yield a population rate. A survey records answers to specified questions; whether those answers support a population estimate depends on its design.

Analysis of existing records can reveal a pattern in recorded activity if the records and definitions are suitable. Check whether existing work already answers the question before collecting more information. For a migration topic, the Productivity Commission’s Migrant Intake into Australia inquiry assessed benefits and costs of temporary and permanent migration and considered alternative ways of determining migrant intake.

For digital PR, ask whether the finding gives journalists or content creators credible data for a story people beyond the business might care about. The angle should come from the result rather than simply announcing that a study was commissioned.

Choose between a census and a sample

A census counts every unit in the population; a sample uses part of it to estimate characteristics of the whole. The ABS’s “Census and sample” guidance notes that a census avoids sampling error and can provide detailed information about small subgroups, but it may take longer and cost more in staff and money. It can also be difficult to enumerate every unit within the available time.

A sample will generally cost less and may produce results sooner. Its design should be robust and large enough for reliable representation, with the required accuracy, cost and timing considered together. A sample can be representative when good sampling techniques are used, but a small sample may not represent the population and gives less detail about subgroups.

Census or sample: how the two approaches compare

Plan the claim before collecting data

Draft the kind of result sentence the study might support. Identify the group, measure, period and method it would need to name. If the planned sample could never support the intended headline, revise the question or the design before collection.

For example, a service provider might want to tell a story about how its customers describe their experience during a defined period. A survey could support a claim about those reported answers if the design supports it, but not a claim about what customers actually did or what all Australians experienced. The story angle must stay with the finding the evidence can support.

Decide which comparisons matter, how groups will be defined, how missing answers will be handled and what would make the study unsuitable for the proposed claim. Keep a dated record of changes after collection begins. Exploring an unexpected pattern is legitimate; describe it as exploratory rather than implying it was the original test.

Detailed research-question design, including survey wording, answer options, testing and coding, is covered in a companion article. The ABS’s archived “Questionnaire Design” guidance notes that poor design can affect response rates, response quality and conclusions.

Keep the route to each result visible

Keep a clear record of how the sample was assembled and how each result was produced, so the route from collection to claim can be explained. Describe what the method observed and identify groups it may have missed.

During collection, check whether coverage, missing answers or response patterns undermine the intended measure. Record departures from the planned method rather than silently changing definitions; if a problem changes what the result means, narrow or stop the claim.

The ABS Data Quality Framework, May 2009 (archived), describes accuracy in terms of coverage, sample, non-response and response error. It notes that coverage can be assessed by comparing the population included in a collection with the target population.

Detailed sample and exclusion records are covered in a companion article. The ABS notes that coverage of statistical measures could be assessed by comparing the population included for the data collection with the target population.

What the ABS Data Quality Framework treats as accuracy

Decide what can be released

Compare the proposed headline with the population, measure, period and calculation actually used. Would a reader infer a wider group, a cause or greater certainty than the work supports? Narrow the claim, gather better evidence or hold the finding.

Put a qualification that changes the main result where readers first meet it. Make the sponsor, method, collection dates, sample source, relevant counts, important exclusions and any weighting accessible. Explain what the result cannot establish.

The American Association for Public Opinion Research has a Transparency Initiative, and CPRA lists a Code of Ethics among its resources. These are named references for transparency and professional conduct, not substitutes for a sound design.

Make supporting calculations and method information available in a form that permits scrutiny without exposing private records. Keep someone available to explain the work and correct an error. The outcome may be a narrow finding, a better question for a later study, or a decision not to make a public claim. None guarantees editorial coverage.

Disclose these when you release a research claim