The Delphi Method is designed to help experts reach consensus on an issue that concerns them all, by using questionnaires, a moderator and anonymous feedback. The process is structured in a way that encourages experts to reconsider their initial opinions and move towards a workable agreement.
Streamlined versions of the Delphi Method are being adopted by digital platforms to develop ontologies for AI services where the precise meanings of technical words (controlled vocabulary) need to be agreed on by domain experts.
How does the Delphi Method work?
In the first round, a panel of experts responds to some open-ended questions relating to the issue at hand. A moderator analyses the responses, which are anonymous, and identifies common themes. These themes provide the basis for a second-round questionnaire.
In each round, the experts are asked to consider the group feedback and re-evaluate their initial answers accordingly. This iterative process encourages opinions to converge, and continues until the minimum bar for consensus is reached, or significant progress can no longer be made. Typically there are between two and four rounds.
A 2025 study published in Systematic Reviews recommended standardised reporting guidelines to improve transparency and rigour when using Delphi methods.
Where does the Delphi method work best?
The Delphi Method is well-suited to domains where the issues are complex and the stakes are high. Issues relating to the design of medical ontology standards, for example, require thoughtful consideration and attention to detail, rather than a quick brainstorming session.
When experts are multinational and spread across different timezones, the Delphi Method is the most practical solution because it doesn’t require face-to-face meetings or live discussions. It is also suitable for consensus building digital platforms and knowledge engineering.
Even when unsuccessful, the Delphi method helps to clarify specific areas of disagreement that need further discussion to resolve.
Why is the Delphi Method anonymous?
Anonymity tends to encourage more honest and uninhibited responses in domains where status and authority hold sway, such as academic or research circles. Without the protection of anonymity, peer pressure, politics and money can influence results.
With the Delphi Method, experts can provide their input independently, avoiding direct interaction with other participants.
What are alternatives to the Delphi Method?
The Nominal Group Technique (NGT) can be a good fit for simpler jobs involving smaller groups. It is a face-to-face approach where the issue at hand is presented, and each team member writes down their own answer. A selection of these answers is then presented, discussed, and ranked by the rest of the team.
While the Delphi Method supports deeper, iterative reflection, NGT tends to work better for generating new ideas and making team-based decisions at pace. With a skilled moderator, NGT can ensure that all team members have a voice. The ranking system is also transparent and easy to grasp.
Other consensus building methods include focus groups and workshopping. Unlike Delphi, these approaches are built on open dialogue and face-to-face group collaboration, which effectively excludes remote teams and those where language barriers exist.
Can voting create a consensus?
Voting methods are perhaps the most efficient tool for reaching decisions. However the consensus results they produce can be fragile. This is because majority votes tend to minimise useful debate and discount minority views. When the rationale behind a decision is not properly scrutinised or broadly accepted, any consensus obtained is likely to be superficial and therefore vulnerable to collapse over time.
Who invented the Delphi Method?
The Delphi Method was originally conceived in the 1950s by the Rand Corporation to help obtain a collective industry view on the future direction of technology in the military sphere. It is still used for forecasting purposes today, especially in circumstances where empirical data is limited or uncertain.