Wideband Delphi: The Root of Scrum Poker

Wideband Delphi is a structured group-estimation method - independent expert estimates first, group discussion second, repeated until the estimates converge - and it's the direct historical ancestor of Scrum Poker. Every technique in the agile estimation family that separates "estimate privately" from "discuss the spread" traces back to this method. This page covers where it came from, how it actually works, and when it's still the right choice today - usually for higher-stakes estimates than a sprint planning session needs.

Where it comes from

The Delphi method itself originated at the RAND Corporation in the 1950s, developed by Olaf Helmer, Norman Dalkey, and colleagues as a way to gather reliable expert forecasts on questions with no objective answer available yet - originally for technological and military forecasting, not software. Its core insight: a panel of experts responding independently, then revising their views after seeing an anonymized summary of the group's spread, produces better-calibrated group judgment than either an open group debate (where the loudest or most senior voice dominates) or simple averaging of independent guesses (which discards the information in why estimates differ).

Barry Boehm and colleagues adapted Delphi specifically for software cost estimation, describing the "Wideband Delphi" variant in his 1981 book Software Engineering Economics. The "wideband" in the name refers to a genuine methodological difference from the original ("narrowband") Delphi process: classical Delphi kept participants isolated, communicating only through a coordinator and written, anonymized summaries between rounds. Wideband Delphi adds structured group discussion meetings between rounds - a wider communication bandwidth among participants - while still preserving independent, anonymous initial estimates to avoid anchoring.

How a Wideband Delphi round actually works

1. A coordinator presents the item to be estimated to a panel of estimators, along with any specification or context available.
2. Each estimator independently produces an estimate and a written rationale - not just a number, but the reasoning behind it. This step happens without discussion, preserving independent judgment.
3. The coordinator collects and anonymizes the estimates, then shares the full spread (and often the written rationales) with the group.
4. The group discusses the spread in a meeting - this is the "wideband" step that distinguishes it from classical Delphi. Estimators with outlying values explain their reasoning; the group asks questions and surfaces assumptions.
5. Each estimator re-estimates, informed by the discussion but still recording their own independent number.
6. Repeat rounds 3-5 until the estimates converge to an acceptable range, or the coordinator determines further rounds won't meaningfully change the outcome.

The structural principle - independent judgment first, anonymized comparison second, discussion third, re-estimate - is exactly what Scrum Poker compresses into a few minutes per story: private votes, simultaneous reveal, discussion of outliers, one re-vote. Scrum Poker is Wideband Delphi's fast, game-ified descendant, trading formal rounds and written rationales for speed.

Why it's slower, and when that's worth it

Wideband Delphi's overhead - written rationales, a coordinator role, multiple structured rounds - makes it considerably slower per estimate than Scrum Poker. That overhead buys something Scrum Poker doesn't produce: a documented, auditable trail of who estimated what and why, at each round. For a fifteen-minute sprint-planning story, that documentation is pure cost. For an estimate that needs to survive external scrutiny - a contractual commitment, a regulated environment, a high-stakes go/no-go decision - the paper trail is the point, not overhead.

Use Wideband Delphi when:

• The estimate needs a defensible, documented rationale - not just a number, but a record of why.
• Stakeholders outside the estimating team will scrutinize the process later.
• The item being estimated is large and consequential enough that the extra rigor pays for itself (a major contract bid, a regulated deliverable).

Use Scrum Poker instead when:

• Estimating sprint-ready stories at normal team cadence - the overhead isn't justified per item.
• Speed matters more than a documented audit trail.
• The team already has a shared calibration and doesn't need written rationales to interpret a number.

A worked comparison

A government contractor needs to estimate a fixed-price bid for a multi-month software deliverable, where the estimate itself will be scrutinized by a contracting officer if costs run over. Three senior engineers independently estimate effort and write a paragraph justifying their number. The coordinator shares the anonymized spread - two estimates cluster near 400 person-days, one comes in at 650. In the discussion meeting, the high estimator explains they're accounting for an integration risk the others hadn't considered (a legacy system with poor documentation). The group agrees the risk is real; all three re-estimate, converging near 480 person-days in round two. The written rationale from every round becomes part of the bid's supporting documentation - exactly the kind of paper trail a fifteen-minute Scrum Poker round wouldn't produce, and exactly what this bid needs.

Wideband Delphi's legacy in modern agile estimation

Understanding Wideband Delphi clarifies why Scrum Poker's hidden-vote mechanism works, rather than just accepting it as a rule: it's not an arbitrary game convention, it's a direct simplification of a rigorously tested group-estimation method built specifically to avoid anchoring and groupthink. When someone asks "why can't we just discuss the story and agree on a number," the honest answer is that open discussion-then-estimate was tried, studied, and found less reliable than independent-estimate-then-discussion - which is the entire reason Delphi, wideband or otherwise, exists.

Run the fast version

For the estimation your team actually does every sprint, Scrum Poker gets Wideband Delphi's core benefit - independent judgment before discussion - without the overhead. Create a free room and run a round the way this method's descendants have worked for over 40 years: estimate privately, compare, discuss, converge.