The viewing room is one of the oldest fixtures in market research. A focus group runs on one side of the glass. On the other, in a darkened room, sit the people who commissioned the work: brand managers, agency staff, policy officers, program managers. They watch, they take notes, they form impressions that will shape decisions long before the final report lands.

The practice has migrated online without much fuss. Client observers now join video focus groups as muted, camera-off attendees, or watch a recording afterwards.

For the most part this is benign. Commissioners who watch data collection tend to understand their own evidence better, and there is real methodological value in a program manager hearing directly from the people their program serves. But from an ethics review perspective, the viewing room raises three questions that applications frequently leave unanswered:

1. Do participants know the observers are there? 
2. What are observers permitted to do with what they see, and who is holding them to that? 
3. Where does the data go, particularly when the observation is virtual?

The default is disclosure to participants

The National Statement on Ethical Conduct in Human Research 2025 (the National Statement) does not use the phrase "one-way mirror," but the relevant provisions are not hard to locate.

Paragraph 2.2.2 requires that consent be based on sufficient information and adequate understanding of both the research and the implications of participation. Paragraph 2.2.6(f) requires that participants be told how privacy and confidentiality will be protected. Paragraph 3.1.30 requires that information on data management be included in what participants receive during the consent process. Paragraph 3.1.58 asks evaluators and researchers to distinguish between disclosure to specific third parties, sharing with other researchers, and disclosure to the public, and to clarify whether that sharing is subject to participant consent.

An observer sitting behind glass, or logged into a video call, is a specific third party receiving identifiable data in real time. That is a disclosure. Whether or not the observer takes a single note, they are seeing faces, hearing voices, and forming views about identifiable people.

Industry practice points the same way. ISO 20252, the international standard for market, opinion and social research, includes dedicated annexes for physical observation (Annex C) and digital observation (Annex D), and a certified provider will have documented procedures covering client attendance at fieldwork. The Research Society's Code of Professional Behaviour is built on transparency and participant protection, and requires Company and Client Partner organisations to ensure that everyone they engage complies with the Code, whether or not those people are members. International codes are more explicit still: the UK Market Research Society's qualitative guidelines require members to clarify with participants the capacity in which observers are present, and to present clients as clients.

So the starting position for most projects is straightforward. Tell participants that observers will be present, tell them who those observers are, and tell them what the observers can and cannot do with what they see.

Disclosure creates its own problems, and they need managing rather than ignoring

Telling participants that the department funding their support service is watching from the next room is not a neutral act.

Paragraph 2.2.9 of the National Statement is clear: no person should be subject to coercion or pressure in deciding whether to participate. Chapter 4.4 deals with dependent or unequal relationships, and its list of examples explicitly includes government or private service providers and the individuals or communities to whom the service is provided. Paragraph 4.4.4 asks evaluators, researchers and reviewers to consider whether potential participants may be compromised in their ability to decline, and whether they might be adversely affected by a decision to participate or not.

Disclosure also has methodological consequences. A participant who knows the commissioner is watching may soften criticism, inflate satisfaction, or simply say less. Social desirability bias is not itself an ethics issue, but it becomes one when it undermines the value of the research, because Paragraph 1.6 requires that the likely benefit justifies the potential risks. Research that cannot deliver its intended benefit has a weaker justification for imposing any burden at all.

None of the above factors argue for concealment. Disclose observation at recruitment rather than at the door, so that declining costs the participant nothing. Describe observers by role rather than by name where naming would add pressure. Give participants a genuine and easy mechanism to ask that observers not attend their session. Have the moderator confirm at the start of the session, out loud, who is watching.

When observation is not disclosed, Chapter 2.3 applies directly

There are projects where telling participants would defeat the purpose. Behavioural observation, mystery shopping, and some service-experience work fall into this category. The National Statement accommodates this, but with specific terms.

Paragraph 2.3.1 permits limited disclosure that does not involve active concealment or planned deception, provided the applicant can demonstrate all of the following: 

• that no suitable alternative involving fuller disclosure exists; 
• that the benefits justify both the limited disclosure and any risk to community trust in research; 
• that the project involves no more than low risk and the limited disclosure is unlikely to affect participants adversely; 
• that the precise extent of the limited disclosure is defined; and,
• that wherever possible and appropriate, participants will afterwards be told what was omitted and why, and offered the opportunity to withdraw their data.

Three points deserve emphasis, because they are where applications tend to come apart.

1. The low risk threshold in 2.3.1(c) is a real constraint. Undisclosed observation is not available for projects that are more than low risk, which rules it out for a good deal of sensitive social research and evaluation.
2. The debrief and withdrawal provision in 2.3.1(e) is not a formality. In a commercial context it is genuinely awkward, because the observer has already formed their impression and cannot un-see the session. Applicants need to say what withdrawal will actually mean in practice, and if it will mean deleting recordings, transcripts and notes, they need to be able to deliver that.
3. Paragraphs 2.3.2 and 2.3.4 matter enormously for determining the review pathway. For lower risk applications, not mentioning that observers are present is passive non-disclosure and can (all else being equal) be reviewed under lower risk processes. Telling participants that no one else is watching, or building a protocol designed to leave that impression if asked, is active concealment. Under paragraph 2.3.4, only a full HREC can review and approve research involving active concealment or planned deception. 

Observers are not members of the evaluation/research team

This is the point most often missed in applications. The evaluation/research team is bound by the approved protocol, by their organisation's policies, and by the conditions of ethical approval. A client observer is bound by none of those things unless someone makes it so.

Paragraph 3.1.57 of the National Statement requires that sharing of data between collaborators and sites be secure and proportional to the sensitivity of the information. Paragraph 3.1.26 requires that participants understand whether third parties will know who has participated. An observer who recognises a participant, or who repeats something they heard at a team meeting the following week, creates precisely the harm these provisions exist to prevent. In a small policy or service sector, this is not hypothetical.

The mechanism is a written observer undertaking, signed before the observer enters the room or joins the call, and provided to the review body as part of the application. An example of this is below:

Observer undertaking

I understand that I am attending this session as an observer and not as a member of the research team. I agree that I will:

• keep confidential everything I see and hear, including the identities of participants;
• not record, photograph, screenshot or otherwise capture any part of the session;
• not take notes that identify any individual participant, and refer to participants only by number or pseudonym;
• not disclose the identity of any participant to anyone, including colleagues within my organisation;
• not contact any participant during or after the session;
• not attempt to interact with participants or influence the conduct of the session, except by raising matters privately with the research team;
• destroy any notes I make once the project is complete, or hand them to the research team; and
• report to the research team promptly if I recognise a participant.

I understand that these obligations continue after the project ends.

Virtual observation adds data management obligations

When observation happens through a video platform, the observer is not just watching. They are sitting inside the data collection environment, on a device the evaluation/research team does not control.

Three obligations follow, and applications should address each of them explicitly.

1. No surreptitious recording. Platform recording rights should be restricted to the moderator or host. Local recording should be disabled at the account level, not left to good intentions. The observer undertaking should say so as well, because platform controls are not absolute and screen capture software exists.
2. No data egress. Observers should not be sent recordings, transcripts or raw notes as a matter of course. If the client is to receive video, that should be a separate, described, consented arrangement with defined retention, not an informal file share after the session. Paragraph 3.1.58 of the National Statement tasks applicants to be explicit about this, and paragraph 3.1.31 requires participants to be told about any limits on withdrawing their data. Once a recording sits on a client's server, withdrawal becomes very difficult to honour.
3. Controlled presence. Observers should join through a separate link or waiting room, with cameras and microphones off and display names that identify them as observers. Chat functions available to observers should be restricted to the research team. The moderator should confirm the observer list before the session starts, because uncontrolled meeting links have a way of being forwarded.

Two worked examples

Example 1: Departmental staff observing evaluation stakeholder interviews

A state department has commissioned an evaluation of a family support program. The evaluator proposes interviews with service users, and the department has asked whether two program staff can observe a sample of interviews to build their understanding of client experience.

Chapter 4.4 of the National Statement applies before anything else. The participants are service users, the observers work for the funder of the service they receive, and the sector is small enough that recognition is plausible. Non-disclosure is not available here, because the risk profile is unlikely to sit at low risk and paragraph 2.3.1(c) is therefore not satisfied.

The workable design is disclosed observation with strong safeguards: 

• observation is described at recruitment; 
• participants can decline observation while still taking part; 
• observers sign an undertaking and leave if they recognise a participant; 
• observers do not attend interviews with participants who have current casework with them; and 
• the evaluator triangulates observed and unobserved interviews to check for response effects. 

If the department will not accept those constraints, the honest answer is that observation should not proceed.

Example 2: A virtual focus group with client observers

A market research firm is running online focus groups on consumer response to a proposed product redesign. Four client-side staff want to observe live.

Here the risk profile is genuinely low and the arrangement is routine. The application should nonetheless state: 

• that participants are told at recruitment that client observers will be present and that observers will not see their surnames or contact details; 
• that observers join with cameras and microphones off under an "Observer" display name; 
• that recording is restricted to the host and local recording is disabled; that observers sign an undertaking; 
• that no recording or transcript is provided to the client, only a de-identified report; and 
• that the moderator confirms the observer list before admitting participants.

The responses to the above examples are not onerous. They are each roughly a paragraph in an application, and it converts an arrangement that a committee would otherwise have to ask about into one it can simply approve.

Sample participant information wording

For disclosed observation, the following language provides a starting point for you to adapt the detail for your own projects.

Who else will be there?

As well as the researcher running the session, some people from [organisation], who have asked for this research to be done, will be watching the session. They will be [in a separate room behind a viewing window / joining the online session with their cameras and microphones switched off]. They are there to listen and learn, not to take part.

These observers have signed an agreement that they will keep everything they see and hear confidential, will not record the session, will not write down anything that identifies you, and will not contact you afterwards. They will not be given a copy of the recording or the transcript.

If you would prefer that no observers are present for your session, please tell us when you register and we will arrange that. It will not affect your ability to take part, and it will not affect [any service you receive].

Two things to note. The wording names the observers' organisation, because "some other people" invites exactly the anxiety it is trying to avoid. And it offers a real mechanism for participants to choose not to be observed.

Summary

Observation of data collection by commissioners is normal, useful, and mostly low risk. It is also a disclosure of identifiable data to third parties who are not bound by the approved protocol, and it should be described in the ethics application as deliberately as any other data handling arrangement.

Disclosure to participants is the default, and the coercion and social desirability risks that disclosure creates are design problems to be managed, not reasons to conceal. Where observation genuinely cannot be disclosed: 

• Paragraphs 2.3.1 to 2.3.4 of the National Statement apply directly; 
• the low risk threshold is a real limit; 
• the debrief and withdrawal provision has to be deliverable; and 
• anything amounting to active concealment goes to full HREC review.

In all cases, observers need written obligations, and virtual observation needs explicit controls on recording and on where the data ends up.

If you are designing a project involving client or commissioner observation and are unsure how to describe it in an application, get in touch. This is a topic where a short conversation before fieldwork is booked saves considerably more time than a round of committee feedback afterwards.

This article draws on the National Statement on Ethical Conduct in Human Research (2025). This article is intended as practical guidance and does not constitute legal advice. If you have questions about whether your project requires ethics review, get in touch.

AI Disclosure: Initial drafts of the content for this article were prepared using Large Language Models with input from Iris Ethics staff who guided the scope and design. Subsequent revisions and final versions were developed and approved by Iris Ethics staff.

​