Principles of Product Management: Decision Intelligence and Decision Record (DACI, Decision Logs)

Decision Intelligence and Decision Record (DACI, Decision Logs)

Decision Intelligence & Decision Records (DACI, Decision Logs)

What This Is

Decision Intelligence (DI) is the discipline of making high-quality, repeatable product decisions by structuring who decides what, when, and how—and documenting the rationale. It matters because unclear ownership, endless debates, or forgotten context lead to slow execution, misaligned teams, and failed launches. A Decision Record (like DACI or a Decision Log) is the artifact that captures this process, ensuring accountability and traceability.

Real-world example: At a fintech startup, the team debated whether to add a "Buy Now, Pay Later" (BNPL) option at checkout. Without a DACI framework, the debate dragged on for weeks—engineering wanted to build, marketing wanted to test, and legal was unsure. After assigning a Driver (PM), Approver (CPO), Contributors (Eng, Legal, Marketing), and Informed (Customer Support), the team aligned in 48 hours, documented the decision in a log, and shipped the feature in 6 weeks—boosting conversion by 12%.

Key Terms & Frameworks

• DACI (Driver, Approver, Contributors, Informed):
  A RACI variant for product decisions. Driver (owns the process, e.g., PM), Approver (has final say, e.g., CPO), Contributors (input providers, e.g., Eng, Design), Informed (kept in the loop, e.g., Support).

• Decision Log:
  A living document (e.g., Notion, Confluence) tracking decision, owner, rationale, date, and status. Example columns: Decision, DACI Roles, Alternatives Considered, Data/Insights, Outcome.

• OODA Loop (Observe-Orient-Decide-Act):
  A military-derived framework for rapid decision-making. Observe (gather data), Orient (analyze context), Decide (choose action), Act (execute). Used in agile PM to avoid analysis paralysis.

• Pre-Mortem:
  A technique where the team imagines the decision failed and brainstorms reasons before committing. Example: "Our BNPL feature flopped—why? Poor UX? Regulatory issues? Low merchant adoption?"

• Weighted Scoring Model:
  A formula to compare options objectively: Score = Σ (Weight × Rating)
  (Weight = importance of criteria, e.g., "User Impact" = 40%; Rating = 1–5 for each option).

• Disagree and Commit:
  A principle where team members align on a decision even if they initially disagreed, to avoid blocking progress. Example: "I still think we should A/B test first, but I’ll commit to the launch plan."

• Decision Quality (DQ) Framework:
  Six criteria for high-quality decisions: Frame (problem definition), Alternatives, Information, Values (goals), Logic (reasoning), Commitment.

• Two-Way Door vs. One-Way Door:
  Two-way door: Reversible decision (e.g., UI tweak). One-way door: Irreversible (e.g., sunsetting a feature). Prioritize speed for two-way doors; rigor for one-way doors.

• Decision Velocity:
  Speed × Quality of decisions. High velocity = fast and good decisions. Example: Amazon’s "Disagree and Commit" culture enables high velocity.

• Stakeholder Mapping (Power/Interest Grid):
  Plot stakeholders by influence (power) and interest to tailor communication. Example: Legal = high power, low interest → keep informed; Eng = high power, high interest → involve early.

Step-by-Step / Process Flow

1. Frame the Decision
2. Write a one-sentence problem statement (e.g., "Should we add BNPL to checkout?").
3. Define success criteria (e.g., "Increase conversion by 10% without hurting NPS").

4. Identify one-way vs. two-way door (e.g., BNPL is two-way—we can roll back).

5. Assign DACI Roles

6. Driver: PM (owns the process).
7. Approver: CPO or GM (final say).
8. Contributors: Eng, Design, Legal, Marketing (input providers).

9. Informed: Support, Sales (kept in the loop).

10. Gather Inputs & Alternatives

11. Run a pre-mortem (e.g., "Why might BNPL fail?").
12. Use weighted scoring to compare options (e.g., BNPL vs. installment plans vs. status quo).

13. Collect data (user research, competitive analysis, tech feasibility).

14. Decide & Document

15. Approver makes the call (e.g., "Launch BNPL with a 3-month pilot").
16. Log the decision in a Decision Record (include: decision, rationale, alternatives, data, owner, date).

17. Example log entry:
    | Decision | Add BNPL to checkout (pilot) |
    |-------------------|-----------------------------|
    | Driver | PM (Alex) |
    | Approver | CPO (Jamie) |
    | Rationale | 20% of users abandon cart at payment; competitors offer BNPL |
    | Alternatives | Installment plans, status quo |
    | Data | User interviews, competitor analysis |
    | Date | 2024-05-15 |

18. Communicate & Align

19. Share the Decision Record with stakeholders.
20. Use Disagree and Commit to unblock dissenters (e.g., "I still think we should test first, but I’ll commit to the pilot").

21. Update the log with outcomes (e.g., "Pilot increased conversion by 8%—full rollout approved").

22. Review & Iterate

23. After 3 months, retrospect (e.g., "Did BNPL meet success criteria?").
24. Update the Decision Log with learnings (e.g., "BNPL worked, but fraud risk increased—added stricter KYC").

Common Mistakes

• Mistake: Treating all decisions as one-way doors.
  Correction: Classify decisions as two-way (reversible) or one-way (irreversible). Two-way doors (e.g., UI tweaks) can be made faster with less rigor.

• Mistake: Letting the loudest voice decide (e.g., HiPPO—Highest Paid Person’s Opinion).
  Correction: Use DACI to clarify roles. The Approver (not the loudest person) has the final say.

• Mistake: Not documenting decisions, leading to "rewinding" debates.
  Correction: Maintain a Decision Log to avoid revisiting old discussions. Example: "We already decided on BNPL—here’s the log."

• Mistake: Over-indexing on data without considering values (e.g., "Data says BNPL increases conversion, but it misaligns with our brand as a premium service").
  Correction: Use the Decision Quality Framework to balance data with values (e.g., "Does this align with our mission?").

• Mistake: Assigning too many Approvers (e.g., "Let’s get CPO, CTO, and CEO to approve").
  Correction: Limit Approvers to 1–2 people to avoid bottlenecks. Example: CPO approves product decisions; CTO approves tech decisions.

PM Interview / Practical Insights

• Interviewer Probe: "How do you handle a situation where your team disagrees on a decision?" Answer: Use DACI to clarify roles, then Disagree and Commit. Example: "I’d assign a Driver (me), get input from Contributors, and have the Approver (CPO) make the call. Even if I disagree, I’ll commit to the decision."

• Tricky Distinction: Driver vs. Approver

• Driver (PM) owns the process (gathering inputs, documenting).

• Approver (CPO/GM) owns the outcome (final say).
  ⚠️ Trap: PMs often overstep as Approvers—don’t!

• Stakeholder Trap: "Legal says no, but users want this feature. How do you decide?" Answer: Use weighted scoring to balance user impact vs. risk. Example: "Legal’s concern is a 3/5 risk, but user demand is a 5/5 impact. We’ll mitigate risk with a pilot and stricter KYC."

• Decision Velocity Trap: "We need to move fast, but we don’t have all the data." Answer: Use the OODA Loop—Observe (gather enough data), Orient (analyze), Decide (choose), Act (execute). Example: "We don’t need 100% data to launch a BNPL pilot—we’ll learn by doing."

Quick Check Questions

1. Scenario: Your team wants to add a dark mode to your app. Eng says it’s easy, but Design says it’ll take 2 sprints. Marketing wants it for a campaign in 3 weeks. How do you decide?
   Answer: Use DACI to assign roles (PM = Driver, CPO = Approver), then weighted scoring to compare effort vs. impact. If it’s a two-way door, launch a minimal version for the campaign.
   Why: Dark mode is reversible (two-way door), so prioritize speed over perfection.

2. Scenario: Your CEO insists on a feature that data shows will hurt retention. How do you respond?
   Answer: Use the Decision Quality Framework—Frame the problem ("Is this aligned with our goal of long-term retention?"), present alternatives, and push for a pilot to test impact.
   Why: Data trumps opinions, but pilots reduce risk.

3. Scenario: Your team keeps revisiting the same debate about a feature. How do you unblock them?
   Answer: Point to the Decision Log ("We already decided—here’s the rationale") and enforce Disagree and Commit.
   Why: Decision logs prevent "rewinding" debates.

Last-Minute Cram Sheet

1. DACI: Driver (process), Approver (final say), Contributors (input), Informed (loop).
2. Decision Log: Document decision, rationale, alternatives, data, owner, date.
3. Two-way door: Reversible (decide fast). One-way door: Irreversible (decide slow).
4. Disagree and Commit: Align even if you disagree.
5. Pre-mortem: Imagine failure before deciding.
6. Weighted Scoring: Σ (Weight × Rating) to compare options.
7. OODA Loop: Observe → Orient → Decide → Act.
8. Decision Velocity: Speed × Quality of decisions.
9. ⚠️ Trap: PMs often act as Approvers—don’t!
10. ⚠️ Trap: Over-indexing on data without considering values (e.g., brand alignment).