Pick what you're deciding — an ADR, policy, standard, strategy, design, role — then tune its character on the seven dimensions. The instrument proposes the best-fitting decision-making method and shows which decision-rights tools (DARE, RACI, RAPID, DACI) fit it, and which quietly break it.
The seven processes on the spectrum are not Harmel-Law's invention — they are long-standing decision rules from separate traditions. The autocratic–consultative–delegative–group ladder is the participative-leadership taxonomy of Vroom and Yetton (1973), refined with Jago (1988). Consent comes from the sociocratic movement — rooted in Quaker practice, formalised by Gerard Endenburg in the 1970s and codified for modern use by Rau and Koch-Gonzalez (2018). Voting and consensus trace back to social-choice theory and cooperative governance (Arrow, 1951). The advice process was developed by Dennis Bakke at AES Corporation in the 1980s–90s and set out in The Decision Maker (2013): anyone can make any decision provided they seek advice from those meaningfully affected and from relevant experts, then decide and log the reasoning. It is distinct from consultative (the initiation right is at the edge, not held by the boss) and from consent (there is no veto — the decider still decides). What Harmel-Law and Wells (2025) added is the comparison itself: plotting these standard processes against decentralisation of decision rights and response speed, and splitting each decision into option-making, decision-taking and decision-sharing.
The engine is a contingency model. Vroom–Yetton set the logic: choose the method by decision-quality needs, acceptance and commitment requirements, time pressure, and where the information sits. The complexity axis draws on Snowden and Boone's Cynefin framework (2007), which separates clear, complicated, complex and chaotic domains and prescribes a different response for each — analyse in the complicated domain, but probe–sense–respond (experiment) in the complex one. The certainty-versus-agreement reading is the Stacey matrix. That is why the tool proposes analytical, expert-led methods for complicated decisions and consent's "safe-to-try" experiments for complex ones.
Two classic arguments underpin the decentralised end. Hayek (1945) held that the knowledge needed for good decisions is dispersed and local, so decisions improve when made close to that knowledge. Simon's bounded rationality explains why no single decider can optimise a complex, interdependent problem: people satisfice under cognitive and time limits, so distributing and integrating judgement — as consent does through reasoned objections — can outperform one mind. Both effects hold only where there is psychological safety and voice (Edmondson, 1999).
RACI, RAPID, DARE and DACI are practitioner tools, not decision rules. RAPID was developed by Bain & Company; DACI at Intuit; the case for naming a single decision owner was made in Rogers and Blenko's "Who Has the D?" (2006). They allocate roles around a decision that already assumes one accountable decider — which is why the tool marks them Avoid the moment the recommendation crosses into consent, democratic or consensus territory.
Because these tools concentrate the final say in one named role, they can be captured — used to formalise a power imbalance rather than to check it. The clearest abuse is naming a Decider who is also a stakeholder in the decision: someone on the very team the decision affects. With no counter-veto and advice that is non-binding by design, consultation becomes cosmetic and the tool lends a pre-existing asymmetry the appearance of due process. In governance terms this collapses the separation of decision management from decision control (Fama & Jensen, 1983): whoever frames and makes a call should not also be the one who ratifies it when they hold a stake in the outcome.
The impact shows up in behaviour, not just fairness. Greer, Jordan and Sytch (2026) document that rights dictated top-down — the "boss's spreadsheet" — produce no buy-in, so people quietly decline to play their roles; that deciders who act on rank rather than the best-informed view degrade decision quality (in their case a senior partner overrode the due-diligence data on instinct and passed on a future unicorn); and that concentrating the call in one contested meeting breeds power struggles and disengagement. The pattern compounds: self- and group-serving bias in the decider, suppressed voice (Edmondson, 1999), and lost local knowledge (Hayek, 1945).
The correction is structural, not moral. Co-create the rights rather than dictate them; give the call to whoever is closest to the work and least conflicted; and for high-stakes or complex decisions move to consent — a distributed, reasoned veto a single stakeholder cannot capture — or the advice process, where the decider must seek and publish advice. Those are the two methods this tool recommends exactly where the rights tools turn "Avoid."
The scoring is a transparent, weighted additive heuristic in the spirit of multi-attribute decision analysis (Keeney & Raiffa, 1976): each dimension contributes, by a stated weight, to a single decentralisation-demand score that maps onto the spectrum. The weights are reasoned defaults, not empirically calibrated coefficients, and the boundaries between methods are deliberately soft. Treat the output as a structured prompt for judgement, not a verdict.
No method wins everywhere. Good decision-making is fit-for-purpose: it matches the character of the decision — its urgency, reversibility, complexity, where the knowledge sits, and how much buy-in it needs — to the right distribution of decision rights, and only then to the right role-allocation tool.
Reading it. The dark marker on the spectrum bar shows the decentralisation demand of the current profile — it moves continuously as you tune, even when the named method hasn't yet flipped. Allocation tools (DARE/RACI/RAPID/DACI) presuppose a single accountable Decider, so as the recommendation crosses into Consent, Democratic or Consensus they are marked Avoid — not because they are bad, but because they re-install the single-Decider veto those methods deliberately remove; there you shift to consented, time-boxed roles and a visible decision log. A reasoning aid, not a substitute for judgement. Method spectrum after Harmel-Law & Wells (2025).