The Delphi method is a structured forecasting and decision-making technique developed in the 1950s by the RAND Corporation for the United States Air Force [1]. The method seeks to achieve expert consensus on complex questions where quantitative data is insufficient or unavailable, using an iterative, anonymous process that systematically refines collective judgement.
In telecommunications, media and technology sectors, the Delphi method proves particularly valuable for:
- Technology forecasting: Adoption timelines for 6G, quantum computing, generative AI
- Risk assessment: Cybersecurity threats, regulatory disruption, competitive dynamics
- Strategic planning: Market scenarios over 5-10 year horizons
- Investment decisions: Emerging market entry, technology bets, M&A targeting
This article examines the methodological foundations, practical implementation, and applications of the Delphi method, drawing on EXXING's experience conducting expert panels across European and African TMT markets.
Methodological Foundations
Core Principles
The Delphi method rests on four foundational principles, each addressing specific limitations of traditional group decision-making [1][2]:
| Principle | Description | Rationale |
|---|---|---|
| Anonymity | Experts do not know each other's identities | Eliminates social pressure, authority bias, and groupthink |
| Iteration | Multiple consultation rounds (typically 3-4) | Enables progressive convergence toward consensus |
| Controlled feedback | Aggregated results shared between rounds | Allows experts to revise positions based on collective input |
| Statistical aggregation | Median, quartiles, standard deviation | Quantifies consensus level and opinion dispersion |
The anonymity principle is particularly important. Research by Janis (1972) on "groupthink" demonstrated how social dynamics in face-to-face groups suppress dissenting views and produce premature consensus [3]. The Delphi method's anonymous structure preserves intellectual independence whilst enabling collective intelligence.
When to Use Delphi
Favourable Conditions:
| Condition | Example |
|---|---|
| Absence of reliable historical data | Forecasting 6G adoption (no precedent) |
| Complex problems requiring multidisciplinary expertise | Assessing AI impact on telecommunications employment |
| Need for stakeholder legitimacy | Regulatory policy development |
| High uncertainty with significant consequences | Strategic investment decisions |
Unfavourable Conditions:
| Condition | Alternative Approach |
|---|---|
| Abundant quantitative data | Econometric modelling |
| Time pressure (Delphi requires 2-4 months) | Rapid expert interviews |
| Experts unavailable or systematically biased | Scenario planning |
| Well-understood phenomena | Standard forecasting methods |
Theoretical Underpinnings
The Delphi method draws on several theoretical traditions:
Wisdom of Crowds: Surowiecki (2004) demonstrated that aggregated independent judgements often outperform individual expert predictions [4]. The Delphi method operationalises this insight through structured aggregation.
Bayesian Updating: Each Delphi round can be understood as experts updating their prior beliefs based on new information (the group's aggregated response). The iterative structure facilitates rational belief revision.
Deliberative Democracy: The controlled feedback mechanism enables a form of deliberation without the distortions of face-to-face interaction, producing decisions that reflect considered collective judgement.
Implementation Methodology
EXXING's Delphi implementation follows a structured six-step process, refined through application across multiple TMT forecasting exercises.
Step One: Problem Definition and Expert Selection
Problem Framing: The forecasting question must be precisely defined. Vague questions produce vague answers. Effective Delphi questions are:
- Specific: "What will be the 5G population coverage in Nigeria in 2030?" rather than "How will 5G develop in Africa?"
- Measurable: Quantitative where possible, with clear units and timeframes
- Relevant: Connected to decisions the commissioning organisation faces
- Bounded: Focused enough that experts can provide informed judgement
Expert Selection Criteria:
| Criterion | Description | Verification Method |
|---|---|---|
| Expertise | 10+ years relevant experience | CV review, publication record |
| Diversity | Varied perspectives (industry, academia, regulation) | Panel composition analysis |
| Independence | No disqualifying conflicts of interest | Declaration of interests |
| Availability | Commitment to complete all rounds | Calendar confirmation |
Panel Size: Research suggests 15-30 experts optimises the trade-off between statistical reliability and administrative feasibility [2]. EXXING typically targets 20-25 experts, anticipating 15-20% attrition across rounds.
Step Two: Questionnaire Design
The initial questionnaire combines quantitative and qualitative elements:
Quantitative Questions: Request specific numerical estimates with confidence intervals.
Example: What percentage of African mobile subscribers will use 5G services by 2030?
- Point estimate: _____ %
- 90% confidence interval: _____ % to _____ %
- Confidence in estimate: ☐ Low ☐ Medium ☐ High
Qualitative Questions: Explore reasoning and identify factors experts consider important.
Example: What are the three most significant barriers to 5G adoption in sub-Saharan Africa? Please rank in order of importance.
Scenario Questions: Test sensitivity to key assumptions.
Example: If spectrum costs decreased by 50%, how would your 5G penetration estimate change?
Step Three: Round One - Initial Estimates
Round One establishes the baseline distribution of expert opinion. Results are analysed for:
- Central tendency: Median (preferred to mean for robustness to outliers)
- Dispersion: Interquartile range (IQR), standard deviation
- Outliers: Responses outside 1.5 × IQR flagged for follow-up
Case Study: 5G Adoption Forecast (West Africa)
EXXING conducted a Delphi study for a private equity fund evaluating telecommunications investments. Round One results for "5G subscriber penetration in West Africa by 2030":
| Statistic | Value |
|---|---|
| Number of respondents | 22 |
| Median estimate | 18% |
| Interquartile range | 12% - 28% |
| Minimum | 5% |
| Maximum | 45% |
| Standard deviation | 11.2% |
The wide dispersion (IQR of 16 percentage points) indicated significant expert disagreement, warranting further rounds.
Step Four: Rounds Two and Three - Convergence
Subsequent rounds provide experts with:
- Aggregated statistics from the previous round
- Anonymous justifications from experts with outlying views
- Opportunity to revise estimates in light of this information
Round Two Feedback Example:
The median estimate for 5G penetration in West Africa by 2030 was 18%, with an interquartile range of 12-28%.
Experts estimating above 30% cited: Rapid smartphone adoption, declining device costs, leapfrogging fixed infrastructure.
Experts estimating below 10% cited: Spectrum availability constraints, limited 5G use cases for mass market, affordability challenges.
Please revise your estimate if you wish, or confirm your original response.
Convergence Metrics: EXXING tracks convergence using:
- IQR reduction: Successful Delphi typically reduces IQR by 30-50% across rounds
- Stability: Median change of less than 5% between rounds indicates convergence
- Outlier persistence: Experts maintaining outlying views despite feedback warrant attention
Step Five: Final Analysis and Reporting
The final Delphi report presents:
Consensus Findings: Central estimates with confidence intervals, representing the collective expert judgement.
Dissenting Views: Reasoned arguments from experts who maintained positions outside the consensus, which may identify risks or opportunities the majority overlooked.
Sensitivity Analysis: How estimates vary under different assumptions about key drivers.
Confidence Assessment: Overall confidence in the forecast, considering expert agreement levels and the quality of reasoning provided.
Case Study Results: West Africa 5G
| Round | Median | IQR | Standard Deviation |
|---|---|---|---|
| 1 | 18% | 12-28% | 11.2% |
| 2 | 16% | 13-22% | 7.8% |
| 3 | 15% | 14-19% | 5.1% |
The panel converged on a 15% penetration estimate (IQR: 14-19%), significantly lower than industry marketing projections but consistent with historical technology adoption patterns in the region. This finding influenced the fund's investment thesis, leading to reduced exposure to 5G-dependent business models.
Step Six: Validation and Learning
Post-hoc validation compares Delphi forecasts against actual outcomes when data becomes available. EXXING maintains a forecast tracking database to assess methodology performance and identify systematic biases.
Validation Findings: Across 15 Delphi studies conducted between 2015-2020 with outcomes now observable:
| Metric | Result |
|---|---|
| Forecasts within IQR | 73% |
| Mean absolute error | 18% of forecast value |
| Systematic bias | Slight optimism (+3% on average) |
These results compare favourably with alternative forecasting methods for similar high-uncertainty questions.
Advanced Techniques
Real-Time Delphi
Traditional Delphi requires weeks between rounds for analysis and questionnaire revision. Real-Time Delphi uses online platforms enabling continuous updating, with experts seeing aggregated results immediately and revising estimates at will [5].
Advantages: Faster completion (days rather than months), higher engagement, richer interaction.
Disadvantages: Reduced deliberation time, potential for herding behaviour, requires sophisticated platform.
Policy Delphi
Policy Delphi adapts the method for policy analysis, focusing on identifying options and understanding trade-offs rather than achieving consensus [6]. Disagreement is valued as revealing important considerations.
Application: EXXING has used Policy Delphi to support regulatory consultations, identifying stakeholder concerns and policy options that might otherwise be overlooked.
Cross-Impact Analysis
Cross-Impact Analysis extends Delphi by exploring interdependencies between forecasted events. Experts estimate not only individual event probabilities but also conditional probabilities (how one event affects others).
Application: Assessing how 5G deployment affects fixed broadband investment, or how AI adoption influences telecommunications employment.
Limitations and Mitigations
| Limitation | Description | Mitigation |
|---|---|---|
| Expert selection bias | Panel composition affects results | Transparent selection criteria, diversity requirements |
| Anchoring | Initial estimates influence subsequent rounds | Careful questionnaire design, multiple question framings |
| False consensus | Convergence may reflect social pressure rather than genuine agreement | Anonymity, explicit solicitation of dissenting views |
| Resource intensity | Delphi requires significant time and expert commitment | Clear scope definition, efficient administration |
Conclusion
The Delphi method provides a rigorous framework for harnessing expert judgement on questions where quantitative data is insufficient. Its structured approach—anonymity, iteration, controlled feedback, statistical aggregation—addresses the known limitations of group decision-making whilst preserving the benefits of collective intelligence.
For TMT strategy, Delphi proves particularly valuable for:
- Technology adoption forecasting where historical precedents are limited
- Market evolution scenarios requiring synthesis of diverse perspectives
- Risk assessment for emerging threats and opportunities
- Strategic planning under high uncertainty
EXXING combines methodological rigour with deep TMT expertise, conducting Delphi studies that inform investment decisions, regulatory strategy, and corporate planning.
Facing strategic uncertainty?
EXXING's forecasting practice designs and conducts Delphi studies tailored to your strategic questions, drawing on our network of TMT experts across Europe and Africa.
Schedule a consultation | Explore our methodologies
References
[1] Dalkey, N., & Helmer, O. (1963). "An Experimental Application of the Delphi Method to the Use of Experts." Management Science, 9(3), 458-467.
[2] Linstone, H.A., & Turoff, M. (1975). The Delphi Method: Techniques and Applications. Addison-Wesley.
[3] Janis, I.L. (1972). Victims of Groupthink: A Psychological Study of Foreign-Policy Decisions and Fiascoes. Houghton Mifflin.
[4] Surowiecki, J. (2004). The Wisdom of Crowds. Doubleday.
[5] Gordon, T., & Pease, A. (2006). "RT Delphi: An Efficient, 'Round-less' Almost Real Time Delphi Method." Technological Forecasting and Social Change, 73(4), 321-333.
[6] Turoff, M. (1970). "The Design of a Policy Delphi." Technological Forecasting and Social Change, 2(2), 149-171.
[7] Rowe, G., & Wright, G. (2001). "Expert Opinions in Forecasting: The Role of the Delphi Technique." In Principles of Forecasting (pp. 125-144). Springer.
