Trusted research starts with the process.

If you want members to give you honest answers, they need to trust the research process.

And if you want a board, government department or other stakeholder to act on the findings, they need to trust the evidence that comes out of it.

Those are slightly different things, but good research needs both. Below are what we think are the 6 steps to ensure your research produces outputs that can be trusted

1. Independence matters. Give people confidence to answer honestly

Confidentiality is one of the foundations of good member research, particularly when you are asking about sensitive issues such as satisfaction with the association, workplace experiences, business performance or views on industry policy.

Members need to know that giving an honest answer will not come back to them personally.

Using an independent research provider can strengthen that confidence because there is genuine separation between the organisation commissioning the research and the individual responses. At Survey Matters, for example, we will never tell a client who has or has not completed their survey, and individual responses are not disclosed unless a respondent has explicitly agreed to be identified.

That independence matters. People are generally more willing to be candid when they know their individual answers will not be passed back to the organisation they are commenting on.

2. Make sure the numbers can be trusted too

Confidentiality helps people trust the process. Research methodology helps people trust the findings.

A survey can generate hundreds or thousands of responses and still give a misleading picture if the people who respond are systematically different from the population you are trying to understand.

That is why sample composition matters. Who responded? Who didn't? Are some parts of the membership over- or under-represented? And does the data need to be weighted to better reflect the population?

The same applies when comparing groups. A difference between younger and older members, for example, may look interesting in a chart but simply reflect random variation in the sample. Significance testing helps distinguish differences that are supported by the data from those we should be more cautious about interpreting.

These things are largely invisible in the final report. But they are what allow an organisation to stand behind the numbers when they are scrutinised.

3. Check the data, not just the outputs

There is another part of research quality that has become more important as AI has become ubiquitous: making sure the responses themselves are genuine.

Survey data has always needed cleaning. Researchers look for duplicate responses, unusually fast completion times, inconsistent answers, straight-lining and other patterns that suggest a response may not be reliable.

Now there is another question to consider: was the response actually written by the person who submitted it?

AI-generated answers can be particularly relevant in open-ended questions. A response may be grammatically perfect and detailed, but that does not necessarily mean it reflects a respondent's genuine experience or opinion.

There is no single automated test that can reliably answer that question. It requires looking at the data itself - reviewing patterns across responses, checking open-ended comments in context and investigating anything that does not look consistent with the rest of the survey.

That human review is still an important part of good research. Before analysing what the data means, you need confidence that the data is worth analysing in the first place.

4. Think carefully about AI and respondent data

AI can be useful in research, including helping researchers work with large volumes of qualitative information. But using it responsibly requires clear boundaries around the data that goes into those tools.

Personally identifiable respondent information should never be copied into publicly available generative AI tools. The same caution should apply to verbatim comments or datasets that could allow an individual to be identified when combined with other information. 

This is not simply a matter of being cautious about new technology. Once personal information enters an external AI system, it can become much harder to know or control how that information is stored, accessed or subsequently used.

The starting point should therefore be the same as it is for any research system: understand what data is being processed, where it is going, who can access it and what protections are in place. 

For this reason, we have very strict policies, internal procedures and rules around how AI is used with survey data.  We also publish an AI transparency policy on every survey we undertake and would suggest this to associations as good practice to increase trust in the survey process.

AI does not remove those obligations. It makes them more important.

5. Think about security and where the data goes

The same principle applies to the broader research technology stack.

Associations should know where survey data is stored, who can access it and what security protections their survey and research platforms have in place.

That is particularly important when surveys collect personal, commercially sensitive or potentially identifiable information. Convenience should not be the only consideration when choosing research software.

Members are giving you their information on the assumption that it will be handled appropriately. The systems and processes behind the research need to justify that confidence.

This the reason Survey Matters only uses research tools where there is a guarantee that the data is stored in Australia.  Again, this is something we would strongly recommend associations look into with their own research tools, and only using companies who  commit to data hosting in Australia.

6. Close the loop

There is one final part of a trusted research process that is easy to overlook: what happens afterwards.

Members who repeatedly complete surveys but never hear about the findings - or see anything change as a result - have less reason to participate the next time.

You do not need to publish every result or act on every suggestion. But sharing what you learned, acknowledging where members raised concerns and explaining what the organisation intends to do with the findings shows that their contribution was worthwhile.

Ultimately, trusted research is not created by one methodological decision.

It comes from the whole process: protecting confidentiality, ensuring the sample is sound, checking the quality of the responses, handling data and AI responsibly, applying appropriate analysis, and being transparent about what was learned and what happens next.

Get those things right and you are not only more likely to produce evidence people can trust. You are also more likely to have members willing to contribute to it again.