Why workshops go wrong
A QRA is only as good as its inputs, and its inputs come largely from workshops. After leading and independently reviewing QRA workshops on UK infrastructure, water, nuclear, defence and industrial capital programmes — including QRA leadership across the National Highways £1bn+ portfolio, independent validation on Mitsubishi Chemical UK (£400m), and a risk transformation across 400+ Northumbrian Water projects — I have a settled view on this: the model, however sophisticated, cannot compensate for poorly elicited three-point estimates, incomplete risk identification, or the social dynamics that cause risk workshops to produce comfortable consensus rather than honest uncertainty ranges. The facilitator's job is to create the conditions in which people say what they actually think, rather than what is politically convenient.
Most risk workshop failures fall into one of four categories: the dominant voice problem (one or two confident people set the tone for the whole room and everyone else adjusts to them); the optimism anchoring problem (people anchor their pessimistic estimates too close to their most likely, because moving far enough into pessimistic territory feels like admitting the project is failing); the omission problem (the workshop captures the risks everyone already knew about and misses the less obvious ones that actually cause the most damage); and the social pressure problem (people will not raise risks they perceive as personal criticism of colleagues, or that they fear will make them look negative).
None of these problems are solved by a better spreadsheet or a more sophisticated Monte Carlo tool. They are solved by facilitation — by the design of the workshop process and the skill of the person running it. This guide covers the practical techniques that work on real projects, including the less comfortable ones that most facilitation guides skip over.
Pre-workshop preparation
The quality of a risk workshop is largely determined before anyone enters the room. Preparation is the most important investment in workshop quality, and it is the one most often skimped on. Three things are essential: participant selection, pre-reading, and a structured agenda with enough slack to allow genuine discussion.
Participant selection matters more than most facilitators acknowledge. You need people with direct knowledge of the work — the engineers, the package managers, the commercial lead. You also need people with institutional perspective — someone who has done a similar project before, who knows what went wrong last time, and who is not so close to this project that they have lost the ability to see it objectively. You do not want the room dominated by the project leadership team, who have the most to gain from optimistic estimates. Senior leaders should attend if at all, at the end, to hear the outputs — not at the start, where their presence chills honest contributions from more junior participants.
Pre-reading should include: the current risk register (so participants have thought about the existing risks before arriving), the programme summary (so participants can think about schedule risks in context), the cost plan structure (so cost risks can be identified at the right level of granularity), and any lessons-learned outputs from similar previous projects. Sending pre-reading the day before and expecting it to be read is optimistic — send it a week before, and follow up two days before to confirm attendance and encourage preparation. A brief covering note explaining what the workshop will ask of participants helps them arrive with the right frame of mind.
The agenda should allow at least half a day for a meaningful risk identification and quantification exercise on any project above about £10m. A two-hour risk workshop on a £50m programme is not enough time to produce credible inputs. Build in a proper lunch break — tired, hungry people give worse estimates than rested ones — and do not schedule the calibration exercise (the hardest cognitive work) for the end of the day when energy is lowest.
Opening with opportunities, not just threats
Most risk workshops open by asking participants to list things that could go wrong. This immediately frames the exercise as a list of bad news, which sets a defensive tone and can make the room feel like an implicit criticism of the project team's planning. A better opening reframes the exercise: we are here to understand uncertainty — both upside and downside — so that the project team can make better decisions. That means starting with opportunities as well as threats.
Asking the room "what could go better than we've planned?" before asking what could go worse has several practical benefits. It counteracts optimism anchoring by establishing that the exercise is genuinely bidirectional — the pessimistic end of a three-point estimate is not a prediction of failure but a symmetric acknowledgement of uncertainty. It surfaces potential schedule pull-ins, cost savings, and commercial opportunities that should be included in the risk model. And it creates a more constructive psychological contract with the participants: they are contributing to a planning tool, not cataloguing the team's mistakes.
Opportunities can be explicitly included in a cost risk model as negatively valued risks — risks with an upside impact that reduces expected cost. Not all QRA tools handle this elegantly, but the principle is straightforward: if there is a 20% chance that ground conditions prove better than expected and save £200k, that is a risk with a negative cost impact and it should be in the model. Excluding opportunities systematically biases the QRA output toward overestimating expected cost, which leads to over-provisioned contingency that is genuinely wasteful.
The calibration problem
The most intellectually demanding part of a risk workshop is calibration: helping people produce three-point estimates that genuinely represent 90% confidence intervals rather than anchored, too-narrow ranges. Research on human probability estimation is unambiguous — we are bad at it, and we get worse under time pressure and social pressure. The facilitator's role is to slow down the estimation process and introduce specific techniques that counteract the known biases.
The most effective calibration technique is the direct challenge. When an estimator gives a minimum of £400k, a most likely of £500k, and a maximum of £600k, ask them directly: "Is there genuinely only a 5-10% chance that the actual cost of this item will be above £600k?" Most people, when asked that question plainly, will acknowledge that a 5-10% tail is too narrow and revise their maximum upward. The same challenge applies to the minimum: "Is there genuinely only a 5-10% chance that this comes in under £400k?" Applying the challenge to both tails often results in ranges that are 30-50% wider than the original estimate — and more realistic.
A complementary technique is historical anchoring. Before the estimator gives their three-point range for a specific risk or activity, show them data from comparable risks or activities on previous projects. If the cost overrun for ground investigation packages on similar projects has ranged from −10% to +150%, that is the relevant empirical distribution — not the estimator's intuition. Reference class data is the most powerful single intervention against optimism bias in risk workshops, and facilitators who have assembled relevant historical data before the workshop will consistently get more credible outputs than those who rely entirely on expert elicitation.
Finally, pre-commit before anchoring. Ask estimators to write down their minimum and maximum values on paper before they hear the most likely value from a colleague. Once a most likely value is spoken aloud, the minimum and maximum estimates anchor to it immediately. By committing independently first, each participant is forced to reason from their own knowledge rather than adjusting someone else's anchor.
Drawing out quiet voices and handling dominant ones
Group dynamics in risk workshops are predictable. The most senior or most confident people speak first and most often. Their estimates anchor the group. More junior people with direct operational knowledge — who often have the most realistic view of what can go wrong — adjust their views to match the room rather than challenging upward. The facilitator's job is to deliberately invert this dynamic.
The most reliable technique for drawing out quieter voices is structured individual responses before group discussion. Before any discussion of a risk or estimate, ask each participant to write their minimum, most likely, and maximum on a sticky note or an index card. Then go around the room collecting the values before any are spoken aloud. Display all the values simultaneously — on a whiteboard or shared screen. The spread of responses, rather than the first voice, becomes the starting point for discussion. People who would not have volunteered a challenging outlier in open discussion are now contributing to a visible range that the whole room can see.
For dominant voices, the most effective intervention is time-limited speaking with explicit invitation to others. If one person is giving detailed responses to every risk item, the facilitator can say: "That's really helpful — before we go further, I want to make sure we've heard from everyone in the room. [Name], what's your view on the lower end of this range?" Naming specific people to contribute is more effective than a general invitation, which dominant voices will fill again. This is not about silencing expertise — a confident, knowledgeable voice is an asset in a risk workshop. It is about ensuring that their estimates are tested against the perspectives of others in the room who may have different knowledge.
The pre-mortem technique is particularly useful for surfacing risks from quieter participants. Asking people to write down (individually and silently) all the reasons the project has failed generates more honest responses than open brainstorming, because the anonymous framing removes the social pressure to be positive. The facilitator collects the responses, reads them without attribution, and uses them as the basis for structured discussion. Risks that emerge from this exercise — particularly those that appear on multiple participants' lists but were not on the pre-existing register — are usually the most important findings of the workshop.
Five red flags your QRA workshop is theatre, not analysis
Most of the QRA workshops I am asked to independently review have at least one of the following five problems. Two or more of them, and the model running off those inputs is not delivering the assurance it appears to — even if the S-curve looks immaculate. These are the patterns to look for whether you are commissioning a workshop, sitting in one, or inheriting outputs from one that has already been run.
One: the estimates have not moved. If the three-point ranges produced by the workshop are within 10–15% of the most likely value on a complex package with multi-month durations or seven-figure costs, the facilitator either did not challenge the inputs or accepted the first numbers offered. A credible range on a major-package risk is usually 30–50% wider than the estimator's opening figure. Narrow ranges feel professional and produce a tight S-curve — but they are systematically wrong, because real projects experience wider variance than committed teams instinctively admit to.
Two: nobody in the room referenced a comparable project. If a risk like ground conditions, commissioning duration, or design-development cost was quantified without anyone presenting data from how similar work has actually performed, the estimates are pure intuition. Reference class data is the single most powerful antidote to optimism bias, and a facilitator who walked into the room without it has not done the preparation. The question to ask after any workshop is: "What comparable project data did we anchor against?" If the answer is silence, the workshop did not have an external check on the team's internal anchor.
Three: the same three voices spoke for three hours. If the facilitator's notes or the workshop recording show that the majority of the contributions came from two or three people in a room of ten or twelve, the workshop captured those people's views — not the room's. The estimates that emerged are anchored on whoever spoke first, and the operational knowledge held by the package engineers, junior planners, and specialist subcontractor representatives never made it into the model. Diverse contribution is not a soft facilitation virtue; it is what produces wider, more honest ranges.
Four: the risk register barely grew. A well-run workshop on a complex programme typically surfaces material new risks that were not on the existing register — usually because they had been quietly avoided, or because nobody had looked at the project from the right angle before. If a half-day workshop produces a register that is functionally identical to the one the team walked in with, either the existing register was already exceptional (rare on inherited projects) or the social pressure to avoid raising new exposure points won. Workshops that confirm what was already known are usually workshops that missed what was not.
Five: the outputs were not in front of the project team within 48 hours. If the draft risk register, action list, and summary of key findings did not reach participants while the discussion was still fresh in their minds, memories drifted, agreements weakened, and entries got softened in post-workshop review. The workshops that change projects are the ones whose outputs are treated as live decisions within two days. The workshops that go into a folder and stay there — and there are many — were compliance exercises, not analysis.
If two or more of these are true on a workshop you have inherited or commissioned, the right response is not to re-run the simulation. It is to re-run the inputs. The model behaves exactly as it should given what it was fed; the assurance failure is upstream of the tool.
Wrap-up and turning outputs into usable risk register entries
The end of a risk workshop is where most of the value is either captured or lost. A workshop that generates rich discussion but ends without clear outputs — specific risk entries with agreed probability and impact ranges, an action list with owners and due dates, and a summary of the key findings — has wasted most of its investment.
The facilitator should have been capturing risk entries throughout the workshop in a structured format, not just taking notes. Each entry needs at minimum: a unique identifier, a cause-risk-effect statement (not just a risk title), a probability of occurrence, a three-point cost impact range, a three-point schedule impact range, a risk owner, and the agreed mitigation actions. If the quantification exercise is done well during the workshop, these can be entered directly into the risk register tool on screen, shared with the room, and confirmed before the session ends. Sending a draft risk register for review after the event is a weaker process — memories fade, people disagree about what was agreed, and entries get softened in post-workshop review.
The summary of key findings should be presented back to the group before the workshop closes. "The three risks driving the most uncertainty are X, Y, and Z. The areas where we have the least confidence in our estimates are A, B, and C. We identified two risks today that were not on the existing register — these are D and E, and they are now flagged as high-priority for immediate mitigation planning." This summary serves two purposes: it confirms that the facilitator has correctly understood the workshop outputs, and it sends participants home with a clear message about what matters most. Without a verbal summary, the workshop's findings are distributed across sticky notes and spreadsheet rows, and nobody is clear on the priority.
Follow-up within 48 hours is critical. Send the draft risk register, the action list, and a brief summary of the key findings to all participants while the workshop is still fresh in their minds. Ask for corrections and additions by a specific date — two weeks is usually enough. After that window, the register should be treated as confirmed. Workshops that do not produce a final register within a month of the session tend not to produce one at all, and the investment in facilitation is wasted.
A QRA workshop is the highest-leverage activity in the risk process. The simulation tool, the risk register format, and the report template are downstream — the inputs decide whether the model produces a number you can defend at a gateway or one that everyone privately doubts. SOMA facilitates and independently reviews QRA workshops on UK infrastructure, water, nuclear, defence and industrial capital programmes — most often when a project is approaching a funding gate (IPA Gate 1 or 2 on government major programmes, or an equivalent Concept-to-Assessment transition on the MoD CADMID lifecycle), when a previous workshop's outputs are not landing, or when an in-house team wants an independent calibration check before the simulation is run. If that describes where you are, it is the conversation we are best at having.