Different assessments need different answers.
Article four in The Proctoring Question series. The last article asked how secure an assessment really needs to be. This one asks what follows from that answer: once you know the level of assurance you need, how do you choose the model that delivers it?
A question that I am often asked is: “What’s the best way to proctor an exam?”
It sounds like a sensible question. But there is no single solution that solves every security concern, satisfies every regulator and works perfectly for every candidate in every assessment.
Every qualification is different. Its purpose, stakes, candidate population, geography and regulatory environment combine to create a unique set of risks and constraints. No single proctoring model can be perfect for them all.
We now have more proctoring options available than ever before, all offering different levels of assurance, flexibility and complexity. So, although there is no single “best” model, there is most definitely a model, or combination of models, that provides the right level of assurance for a particular assessment.
From Best Practice to Better Judgement
Several years ago, I was invited to sit on a panel tasked with defining “best practice” for organisations who had adopted remote proctoring solutions. I only lasted three meetings. Everyone else in the room was trying to find a single, one size fits all solution. They were trying to find one answer to a question that required many.
It is frustrating that similar conversations still take place today, the idea that there is a single, perfect, model that can be adopted by all is fundamentally flawed. Assurance is not all or nothing, it exists on a spectrum.
At one end, there are low-stakes assessments where the primary goal is learning, feedback or practice. At the other, there are high-stakes decisions: licensing, professional qualifications or employment registration where a compromised result carries significant consequences for candidates, employers and the public.
The same security model cannot be applied across that entire spectrum.
Lighter-touch approaches may be perfectly appropriate for entry-level or formative assessments, where the risk of occasional misconduct is outweighed by the value of wider access and lower friction. That same model would be inadequate for high-stakes exams where public safety, professional standards or regulatory compliance are on the line.
Rather than searching for one universal model, we need to understand where each assessment sits on the spectrum and what level of assurance it genuinely requires.
Different Models, Different Strengths
Choosing the right level of assurance means understanding how each proctoring model operates in practice.
In-centre invigilation: Provides a familiar, controlled environment with visible oversight, standardised equipment and a clear separation between the assessment setting and the candidate’s home life. It can provide high assurance, but it brings travel, fixed schedules and less flexibility for global or remote candidates.
Live remote proctoring: Brings many of the benefits of in-centre oversight into the candidate’s own environment. It can provide strong assurance without requiring travel, but it depends on reliable connectivity, suitable home conditions and the ability of proctors to interpret behaviour through a limited view on a screen.
Record-and-review / AI-assisted models: Capture sessions for later review and use AI to surface behaviours for human follow-up. They can scale more easily and offer consistent flagging, but they rely heavily on good configuration, clear thresholds and robust governance of false positives and false negatives.
Low-touch or unproctored assessments: Rely more heavily on assessment design, honour codes and analytics than on direct oversight. They can be appropriate where the stakes are lower, the tasks are less vulnerable or the primary purpose is learning and feedback.
Each of these models can be applied effectively, or poorly. None is inherently superior, and none is suitable for every assessment. Their value lies in giving us different ways to match the level of oversight to the assessment, the risk and the candidates involved.
The Same Exam Can Need Different Decisions
Two organisations can be running similar exams and still arrive at different proctoring decisions, because their risk appetites and contexts differ.
A regulator-driven body may face legal or reputational constraints that push it toward more controlled models, even where remote options are technically possible. A global education provider working across time zones and infrastructures may need to lean harder on remote and hybrid models simply to reach its candidates. A professional body with a mixed portfolio may adopt a centre-based model for capstone exams, while using lighter remote approaches for earlier stages.
The appropriate model depends on the consequences of an unreliable result, the likelihood and type of misconduct, the expectations of regulators and stakeholders, and the realities of candidate access, geography and scale.
Once we accept that the model must fit the circumstances, we can move beyond blunt positions such as “remote is too risky” or “exam centres are the only safe option”, and instead reach the conclusion: For this assessment, with these candidates and at these stakes, this is the level of assurance we need.
The Hybrid Approach
Hybrid models are often misunderstood as a compromise between remote and centre-based delivery. In reality, they are frequently a deliberate way of applying different levels of assurance where different levels of risk exist.
We might use lighter-touch remote invigilation for low-stakes or entry-level assessments, while reserving more controlled arrangements for final or high-stakes components. We might combine automated flagging with targeted human review, focusing human effort where it adds the most value. We might also offer both centre-based and remote sittings, with clear criteria governing which route is appropriate in which circumstances.
Hybrid approaches recognise that assurance is not all or nothing. Different parts of the same qualification may require different levels of control, and those requirements may change as threats, technologies and expectations evolve.
Making Better Decisions
One of the biggest risks in proctoring is allowing the technology to drive the strategy.
It is easy to be captivated by feature lists and vendor stories, and just as easy for us to be influenced by isolated incidents that push us towards more extreme positions than the evidence alone supports.
Better decisions require a governance framework, and a written record of the answers to four questions.
Four questions to answer before you choose a model:
- What is the purpose and objective of the assessment?
- What are the main risks and vulnerabilities?
- What are the strengths and limitations of each proctoring model
- Why did select a particular approach and how will we review that choice
This process will not lead us to one universally “best” proctoring model. It will lead us to a model that is appropriate for the assessment in front of us. More importantly, it forces us to acknowledge the trade-offs rather than defaulting to whatever feels safest, cheapest or most familiar.
A Final Thought
When I tell people that there is no “best proctoring model, I know that the answer can be disappointing. Everyone wants the best model.
The truth is there isn’t one.
For me, that is good news. Variety isn’t a weakness. Different assessments deserve different levels of assurance. Selecting the most suitable model is not about whether we always choose remote, or we always choose exam centres. It is about whether we can confidently explain, justify and defend why we chose the model that we did. The quality of the decision matters more than the model itself.
More in this series
The Proctoring Question began with Designing Trust: The Next Chapter of Remote Proctoring, which set out why trust, not surveillance, is the measure of a proctoring decision.
Previously: How Secure Is Secure Enough? on deciding how much assurance a given assessment actually requires.
Next: Candidate experience and wellbeing
If you are weighing up which model fits a particular qualification, we are happy to talk it through. Cirrus works with awarding bodies, regulators and education providers across all four of the models above. We would rather help you reach a decision you can explain and defend than sell you the one with the longest feature list.
