River
Y CombinatorBacked by Y Combinator
FREE TEMPLATE

Decision Rights Policy Template

Four documents and four sheets that test every escalation against what it actually changed last quarter, not what the org chart assumes.

Free download  ·  No account needed

Decision Rights Policy

Verrian Digital, one quarter of decisions, 120 logged

Escalated44 of 120, above the natural decider
Overturned11 of 44 — the other 33 changed nothing
RuleKeep escalation only above a 30% overturn rate

Four of the eight rules in force

DecisionDecidesOverturn rateVerdict
Discount up to 15%Account Managern/a, 0 escalatedFull authority
Discount 15–25%Account Manager16.7%Full authority, notify only
Discount above 25%CRO approves80.0%Approval required
Refund above $500Finance Director approves60.0%Approval required

The other four decision types in the policy — scope changes, vendor engagements, contractor hiring, deadline exceptions — all cleared full authority this quarter, at overturn rates from 0% to 25%.

Every decision rights template on page one is the same blank grid: list your decisions down the side, list roles across the top, and write a letter in each cell. RACI, DACI and RAPID all do this, and all three ask a person to assert who should decide, from title and instinct, before a single real decision has been checked against the assertion. The letters look precise. Nothing behind them has ever been measured.

This pack starts from the other end. It reads what actually got decided recently, who it went to, and whether the person above the natural decider ever changed the outcome. A RACI can still tell you who is nominally accountable for a step. What this tells you is different: whether the person one level up is exercising real judgment, or just adding a day of latency to a call the natural decider had already made correctly.

Verrian Digital, a 38-person marketing agency, logged 120 decisions across one quarter. 44 were escalated, and only 11 of those 44 actually changed. The other 33, kept below the 30% line, moved straight to full authority with a notify-only step upward, and only two of six decision types kept a real approval requirement. This is the same evidence-first approach the handoff agreement pack applies to a bounced task, the process map applies to a queue, and the operating cadence pack applies to a meeting: measure what happened, then write the rule.

Three sheets, from raw entries to the final rule

Decision Log, Decision Type Register, and Escalation Trigger.

Decision Log

6 of 120 decisions logged at Verrian Digital across one quarter, drawn from meeting notes and approval threads.

DateTypeDecided ByEscalated ToOverturnedOutcome
Apr 7DiscountAccount MgrSales DirectorNoApproved at 18% as proposed
Apr 14DiscountAccount MgrCROYesCut from 30% to 22%
Apr 29RefundSupport LeadFinance Dir.YesCut from $650 to $400
May 8VendorOps ManagerFinance Dir.NoApproved as proposed
May 15ContractorDept LeadCOOYesApproved, end date added
Jun 21DiscountAccount MgrSales DirectorNoApproved at 16% as proposed

Overturned means the substance changed. A question asked or a comment added does not count — only a changed number or an added condition does, which is why four of these six read "as proposed" despite every one being escalated.

Decision Type Register

All 120 decisions, rolled up by type, with the verdict the overturn rate produced.

TypeLoggedEscalatedOverturnedRateVerdict
Discount approval4617635.3%Split by size
Refund or credit145360.0%Kept above $500
Contractor hiring114125.0%Full authority
Scope change249111.1%Full authority
Vendor engagement18600.0%Full authority
Deadline exception7300.0%Full authority
Total120441125.0%33 of 44 changed nothing

Four of six types cleared full authority. Only discount and refund kept any mandatory approval, and only above a size threshold the tier-level test finds next.

Escalation Trigger

Discount and refund split into the tiers where the verdict actually changes.

TypeTierLoggedOverturn RateEscalation
DiscountUp to 15%29n/aNone
Discount15–25%1216.7%Notify only
DiscountAbove 25%580.0%CRO must approve
RefundUp to $5009n/aNone
RefundAbove $500560.0%Finance must approve

The same decision type splits two ways. A 12% discount and a 30% discount are not the same decision wearing different numbers — one has never been overturned, the other is overturned four times out of five.

What's in the pack

01

Decision Log

Every decision logged for the period with who decided it, who it was escalated to, and whether that escalation actually changed the outcome or just approved what was already proposed.

02

Decision Type Register

All decisions rolled up by type, each with its escalation count, overturn count, overturn rate, and the resulting verdict: full authority or a kept approval requirement.

03

Escalation Trigger

The types that split by size or stakes broken into tiers, so a 12% discount and a 30% discount get the separate rules their separate overturn rates justify.

04

Approver Map

The usable grid: who decides alone, who gets a notification with no veto, and who must approve before the decision ships, for every decision type and tier.

05

Decision Rights Policy

The signed document naming every rule in force, with the overturn rate written next to each one so the policy is checkable against the evidence behind it.

06

Decision Log Standard

The seven fields every future decision gets logged with, so next quarter's re-derivation has real data instead of reconstructed memory.

07

How Decision Rights Are Derived

The 30% threshold explained: why an escalation below it is latency with nothing to show, and why the number is a judgment call rather than a formula.

08

What This Does Not Fix

The honest limits: a decision type with no volume yet, a decision nobody logged, genuine disputes over who owns a decision type, and one-off irreversible calls.

How to use it

  1. 1

    Open in River, or download it

    Install the pack and hand River your recent decision history, or download all four documents and four sheets blank and run the test yourself.

  2. 2

    Send the decision history

    Meeting notes, an approval thread, a CRM trail, or just the last several instances of a recurring decision and who they went to.

  3. 3

    Let the overturn rate decide

    Every decision type is tested: escalated count, overturned count, and whether the resulting rate clears the 30% line for keeping a mandatory approval.

  4. 4

    Sign the policy, log what's next

    The surviving rules become the Decision Rights Policy, and every future decision gets logged so next quarter's derivation has real evidence.

Frequently asked questions

Is this free, and what do I get?

Free, no signup needed for the download: four documents and four spreadsheets. The AI half is optional. Send River your recent decision history and it runs the overturn-rate test and drafts the policy for you.

Why not just assign decision rights by org chart position?

Because title says who is authorized, not who is actually catching mistakes. An org chart cannot show that a Director has approved twelve discounts in a row exactly as proposed, which is the fact that should retire that approval step.

What does a 30% overturn rate actually mean?

Of every time a decision type was escalated, the share where the person above actually changed the outcome rather than approving it as proposed. Below 30%, the pack argues the escalation is delay without a matching safety benefit.

What if a decision type doesn't have enough history yet?

It stays on the conservative default, approval required, until it accumulates enough logged instances to test, typically eight to ten. A thin sample gets a cautious rule, never a false verdict borrowed from other types.

How is this different from a RACI matrix?

A RACI names who is accountable for a step in a process. This tests whether escalating above that person, once assigned, is actually changing anything. The two are complementary: run a RACI first if ownership itself is contested.

What does Edit with AI actually do?

Creates a free account, installs this exact pack as a private space, and primes the agent to read whatever decision history you send. Nothing gets written until you send something, and the blank pack is always there to download instead.

Stop escalating decisions that never change

Take the four documents and four sheets blank, or install this pack in River and send your recent decision history.

Edit with AI