Learning path · Level 10 · Tutorial 3

Set Decision Rules for Measurement and Experiments

Create a measurement and experimentation framework that turns the plan into an accountable decision process. Define review schedules, evidence requirements, consent safeguards, and practical rules for responding to results without overreacting to noisy or incomplete data.

  • Level: Advanced
  • Duration: 25 min
  • 8 steps
Start tutorial Back to the path
Set Decision Rules for Measurement and Experiments

What you will go through

  1. Anchor Measurement to Your Existing Plan Turn your existing capstone plan into a traceable measurement baseline and a compact measurement matrix. 3 min
  2. Define Which Results Count Define a reconciliation rule so business success is reported once, while channel-attributed results and supporting signals remain useful without overstating what they prove. 3 min
  3. Set Evidence and Consent Readiness Gates Define evidence-readiness gates, permissible fallbacks, and conclusion limits for incomplete or compromised measurement in your existing capstone plan. 3 min
  4. Write Decision Rules Before Results Arrive Precommit to evidence-based responses that distinguish a performance trigger from proof, protect non-negotiable guardrails, and assign authorization. 4 min
  5. Assign Review Cadences and Decision Authority Set review timing, action authority, urgent escalation, and decision records for your existing capstone plan. 3 min
  6. Prioritize Experiments That Could Change the Plan Build a small, decision-focused experiment backlog that respects the existing capstone plan, evidence readiness, approved budget, and available capacity. 3 min
  7. Sequence Experiments and Precommit to Evidence Schedule capstone experiments without contaminating their evidence, and precommit to a response when results are inconclusive. 3 min
  8. Apply and Finalize Your Decision Framework Integrate your existing capstone measurement matrix, decision rules, review authority, safeguards, and experiment sequence into a bounded response to uncertain evidence. 3 min

What you will learn

  • Build a measurement matrix linking each objective to an outcome measure, supporting channel signals, data sources, an owner, and a review cadence.
  • Separate business-level results from overlapping channel attribution claims when defining the plan's success measures.
  • Specify decision thresholds, observation windows, evidence checks, and guardrails for maintaining, investigating, pausing, or revising activities.
  • Prioritize and sequence a small experiment backlog that addresses consequential assumptions without conflicting simultaneous changes.
  • Define consent and access prerequisites, fallback evidence, and limits on conclusions when measurement is incomplete.

Before you start

  • Connect Channels in an Integrated Marketing Plan
  • Build a Capacity-Aware Marketing Roadmap
  • Build a Decision-Focused Looker Studio Dashboard
  • Design and Interpret a Simple A/B Test
  • Attribution limitations, measurement consent, data-access requirements, and channel-report interpretation.

Step 1 of 8

Anchor Measurement to Your Existing Plan

Turn your existing capstone plan into a traceable measurement baseline and a compact measurement matrix.

Anchor measurement in the existing plan

Create the Capstone baseline

Measurement must govern the plan you already built—not a replacement plan. Start by creating a Capstone baseline for your capstone business.

Capture the exact objective wording and targets, included and deferred channels, roadmap phase names, budget assumptions, named owners, and staff or fulfillment limits. Preserve the plan’s original currency and measurement units.

Mark every input as:

  • Confirmed — recovered from your prior plan.
  • Provisional — a working assumption reconstructed now.
  • Not recovered — unavailable; keep it visible as a dependency.

Do not silently fill a missing target, constraint, or owner with a plausible guess. A measurement rule cannot be fully accountable when a critical input is unknown.

What the baseline connects

A central capstone-plan record links objectives, channel choices, roadmap phases, budget, owners, and capacity constraints, each with a visible status marker.

The Capstone baseline keeps the existing plan inputs connected to measurement; missing inputs remain visible rather than being replaced.

Recover inputs and build the matrix

Recover without inventing

If your earlier notes are incomplete, make a compact recovery worksheet before building the matrix. For each field, record the value and its status:

  1. Business name
  2. Exact objective and target
  3. Included and deferred channels
  4. Roadmap phase name and timing
  5. Budget assumption and currency
  6. Owner or role
  7. Staff, approval, or fulfillment limit

Use your capstone business if its name is unavailable. A reconstructed working value is Provisional; an absent value is Not recovered. Keep both in the baseline as dependencies for later review.

Measurement matrix template

Create one row for every selected objective. Keep the metric definitions and units already chosen in your capstone plan.

Replace bracketed prompts only with information from your Capstone baseline.
ObjectiveBusiness outcomeSupporting signalSourceOwnerReview cadence
[exact capstone objective][outcome measure and unit][channel signal tied to a decision][authorized source][named owner or role][review label and timing]
[next exact objective][outcome measure and unit][supporting signal][authorized source][named owner or role][review label and timing]

Tip

Two checks before you proceed

Coverage check: every selected objective needs a business outcome, authorized source, accountable owner, and review cadence.

Traceability check: every supporting signal must inform a specific plan decision. A channel signal can help diagnose or monitor activity, but it is not automatically the business outcome.

Check the role of each measure

Match measures to roles

Match each measurement-matrix field to its decision role.

Tap an item and then its matching pair.

Apply the baseline to one objective

Draft an accountable matrix row

Draft one row for your own measurement matrix using an existing Capstone baseline objective. Use the exact labels below and preserve your capstone’s metric definitions and units.

Objective:
Business outcome:
Supporting signal:
Source:
Owner:
Review cadence:

Then add:

  • Status: Confirmed, Provisional, or Not recovered
  • Dependency: name any unrecovered target, constraint, source access, or owner that prevents this row from supporting an accountable decision.

Do not create a new objective or channel. If nothing is missing, write: “No unrecovered input; baseline confirmed.”

Write at least 180 characters (0/180).

Step 2 of 8

Define Which Results Count

Define a reconciliation rule so business success is reported once, while channel-attributed results and supporting signals remain useful without overstating what they prove.

Separate one total from contributing signals

Separate the result layers

Use the measurement matrix from your Capstone baseline to report results in three separate layers:

  1. Business outcome — the consolidated result for the exact objective and metric definition already chosen. This is the plan’s success measure.
  2. Channel-attributed result — a channel or platform’s claim about its contribution. It can help explain performance, but it is not an additional total.
  3. Supporting signal — a diagnostic signal, such as engagement or qualified traffic, that helps interpret delivery and customer behavior.

Write one reconciliation rule for your plan: Report the business outcome once from its designated consolidated source. Report channel-attributed results and supporting signals separately; never add overlapping channel claims to create a business total.

Keep the Capstone baseline’s exact objectives, metric definitions, units, selected channels, and named sources. Do not replace them with newly invented measures.

One business outcome; separate explanatory reports

A consolidated ledger is distinct from three overlapping channel report panels and a separate row of diagnostic signals.

The consolidated source reports the business outcome once. Overlapping channel-attributed results explain possible contribution; supporting signals diagnose what happened.

Reconcile before interpreting

Apply source precedence and comparison checks

Choose source precedence from the metric definition and authorized coverage—not from which platform reports the largest number.

For the business outcome, use the authorized consolidated record that covers the defined outcome. For a channel-attributed result, use that channel’s defined report and label it as attributed. A supporting signal comes from the source that directly measures that signal.

Before reconciling a period, compare: the reporting period, the conversion definition, treatment of refunds or cancellations where applicable, and duplicate transactions. If these do not align, state the limitation instead of forcing a combined result.

Reporting layers and source roles

Use the source selected in your Capstone baseline when it meets the metric definition and required coverage.

A source can be useful without being authoritative for every result.
Reporting layerReport it asSource precedenceUse it to answer
Business outcomeOne consolidated resultAuthorized consolidated business record matching the definitionDid the objective progress?
Channel-attributed resultSeparate channel claimThat channel’s defined reporting sourceWhat does this channel report?
Supporting signalDiagnostic observationSource that directly measures the signalWhat may explain delivery or behavior?

Example

Use channel claims as explanation, not arithmetic

Practice evidence

Your Capstone baseline names a consolidated business record for its business outcome. A search platform and an email platform both report credit for some of the same completed outcomes during the same period.

A defensible report can say: “The consolidated business outcome is reported from the designated record. Search- and email-attributed results are shown separately as explanatory evidence.”

It cannot say that the two platform claims are unique outcomes when their overlap is unknown. Nor can the consolidated outcome alone prove that either channel created incremental impact. It describes business performance, not causal lift.

Practice a defensible reporting conclusion

Interpret an overlapping-results snapshot

Practice evidence — For the same reporting period, the authorized consolidated record shows your capstone’s exact business outcome at 96% of its Capstone baseline target. Its definition includes the required refund or cancellation treatment and duplicate transactions have been removed.

Three selected channels report attributed credit for that same outcome: Search 62% of target, Email 51% of target, and Paid Social 34% of target. Their attribution claims may overlap.

Which reporting conclusion is defensible?

Step 3 of 8

Set Evidence and Consent Readiness Gates

Define evidence-readiness gates, permissible fallbacks, and conclusion limits for incomplete or compromised measurement in your existing capstone plan.

Establish an evidence-readiness gate

Gate evidence before interpreting it

A readiness gate sits between your measurement matrix and any conclusion. For each planned measure in your Capstone baseline, record whether the evidence is usable before it is interpreted.

Check five conditions:

  • Authorized access: the named owner may use the source for this purpose.
  • Applicable consent: required permissions and safeguards are in place.
  • Collection status: collection is active and functioning as intended.
  • Coverage: the evidence represents the relevant period, population, and channel scope.
  • Accountable owner: a named roadmap owner, or Measurement owner when a name was not recovered, can verify the record.

Keep an unmet condition visible as a dependency. Do not replace it with a convenient proxy or treat missing observations as zero outcomes.

A five-checkpoint readiness gate showing access, consent, collection, coverage, and ownership before evidence can be used.

A measure is decision-ready only when its access, consent, collection, coverage, and ownership checks are satisfied.

Tip

Keep the matrix tied to the plan

For each Objective in your existing matrix, keep the same Business outcome, Supporting signal, Source, Owner, and Review cadence already selected. Add a short readiness note beside the source—for example: confirmed, provisional, or an explicit gap.

A source can be technically available yet fail this gate. Access might not be authorized; consent coverage might have changed; or the source may omit part of the period. Availability is not permission, and permission is not sufficient coverage.

Select the narrowest defensible conclusion

Use fallbacks with explicit limits

Choose a fallback based on the decision—not on the easiest metric to obtain.

If you need a consolidated total and an authorized aggregate business record has appropriate coverage, use it for the business outcome. It cannot recreate a missing channel-attributed result.

If you need a diagnostic clue, appropriately collected feedback may help explain what people report. It does not establish causal lift or prove that a channel caused the outcome.

If no permissible, decision-relevant evidence remains, record an evidence gap. The correct conclusion may be Inconclusive, not a substitute number.

Fallback hierarchy

Practice evidence: use this routing logic relative to your capstone’s own targets and limits.

A fallback changes what can be concluded; it does not restore the planned evidence.
Decision needPermissible evidencePermitted conclusion
Business totalAuthorized aggregate business record with adequate coverageBusiness outcome for the covered period
Diagnostic questionAppropriately collected feedbackA clue worth investigating
Channel contributionMissing or compromised attributionEvidence gap; no channel-attributed result
Causal impactFeedback or overlapping platform claims onlyInconclusive; no causal-lift claim

Handle compromised or incomplete measurement

Warning

Protect people and protect the conclusion

A missing observation is a loss of visibility, not proof that the business outcome fell. Before comparing periods, check whether consent coverage, collection status, definitions, or source coverage changed. A changed collection basis can make an apparent change misleading.

Never bypass a consent or privacy safeguard with another identifier, tracking method, or seemingly anonymous dataset. Aggregate data is not automatically permissible; it still needs authorized access and applicable consent safeguards.

If you suspect a consent failure, promptly contain the affected collection and escalate through your organization’s process. Assess unaffected compliant activity separately; do not assume all evidence is invalid.

Make the limitation auditable

Write a compact note beside the affected measure:

Evidence: what remains authorized and covered.
Limitation: what is missing, changed, or compromised.
Action: contain affected collection when needed; otherwise preserve the evidence gap.
Authorization: the responsible organizational process or owner.
Next review: when readiness can be checked again.

This is not a performance decision rule yet. It is a record of what the available evidence can safely support.

Apply the fallback and safeguard rules

Readiness-gate practice

Match each Practice evidence situation to the most defensible response. Base your choice on evidence permission and what the evidence can actually support.

Match each evidence gap or condition to its authorized fallback, restricted conclusion, or required response.

Tap an item and then its matching pair.

Step 4 of 8

Write Decision Rules Before Results Arrive

Precommit to evidence-based responses that distinguish a performance trigger from proof, protect non-negotiable guardrails, and assign authorization.

Rule anatomy

Turn a measure into a governed decision

Use your Capstone baseline as the source of truth: the same objective wording, business outcome definition, selected channels, roadmap phases, budget assumptions, and capacity limits govern the rule.

A decision rule is a precommitment: it states what will be checked, how long evidence must accumulate, what can override the normal wait, and who may authorize the response. It prevents a single dashboard movement from becoming an unreviewed plan change.

Use a business outcome as the primary performance measure whenever possible. A supporting signal or a channel-attributed result may help explain what happened, but overlapping channel claims are not unique business results and must not be added together.

The parts of a decision rule

A workflow links a metric, trigger, observation window, evidence checks, guardrails, action, owner, approver, and recheck; a guardrail can bypass the normal observation window.

A threshold is only one part of the rule. Evidence checks and guardrails determine whether—and how—the plan responds.

Decision-rule fields

Complete these fields before results arrive.

Use these exact labels in your capstone decision rules.
FieldWhat you precommit
MetricDefined business outcome or approved supporting signal
TriggerAcceptable range or condition that warrants a response
Observation windowEnough comparable time for conversion lag and opportunity
Evidence checksFreshness, coverage, collection, attribution, and context checks
GuardrailsNon-negotiable budget, fulfillment, customer-harm, or consent limits
ActionMaintain, Investigate, Pause, or Revise
OwnerPerson or role that monitors and carries out the action
ApproverPerson or role authorized to approve the action
RecheckWhen the same evidence will be assessed again

Separate signal, evidence, and protection

Set a trigger without overclaiming

Derive a trigger from the exact target or operating limit in the Capstone baseline. A recovered limit is Confirmed. A temporary working limit is Provisional and needs explicit approval. If a target, constraint, or owner cannot be recovered, mark it Not recovered as a visible dependency; do not invent a number or name.

Write the core condition in this form:

If [trigger], after [window and checks], then [action], authorized by [approver].

Crossing a trigger does not prove a channel caused the result or that an experimental difference is statistically reliable. It tells the team which bounded response is justified after the promised evidence review.

Make the evidence window do its job

Before interpreting a trigger, check the authorized source is sufficiently current; collection and consent coverage are comparable; channel-attributed claims are not being summed as business results; and a relevant operational or external explanation has been considered.

Choose an observation window that fits the capstone's conversion lag and decision timing. Reporting can be frequent, but a result becomes decision-ready only after the defined window and checks. When evidence is incomplete, the defensible conclusion is Inconclusive—not an invented cause.

Warning

Guardrails override normal waiting

A guardrail can override the normal observation window. A hard budget, fulfillment, customer-harm, or suspected consent problem may require Pause of the affected activity or collection and prompt escalation under your organization’s process. Assess unaffected compliant activity separately.

Do not bypass consent or access requirements with alternate identifiers, tracking methods, or supposedly aggregate data. If the trigger is met but checks are inadequate, use a bounded Investigate action or a protective Pause; do not claim success or failure as fact.

Practice

Practice the rule sequence

Build the core conditional sentence. Put the clauses in the order that prevents a metric threshold alone from dictating a response.

Your sentence

Available words

Apply it to your capstone

Write your capstone decision rule

For one objective in your Capstone baseline, write a usable decision rule. Include these labels: Metric, Trigger, Observation window, Evidence checks, Guardrails, Action, Owner, Approver, and Recheck.

Then write its core condition exactly in this form: If [trigger], after [window and checks], then [action], authorized by [approver]. Use exactly one of these actions: Maintain, Investigate, Pause, or Revise.

Add a guardrail override that states what would require a protective Pause before the normal window closes. Base every value on the existing capstone objective and constraints. Mark unrecovered inputs as Confirmed, Provisional, or Not recovered rather than fabricating them.

Write at least 400 characters (0/400).

Step 5 of 8

Assign Review Cadences and Decision Authority

Set review timing, action authority, urgent escalation, and decision records for your existing capstone plan.

1. Align reviews with evidence

Set review clocks from the roadmap

Use the Capstone baseline as the calendar anchor: its Confirmed roadmap phases, channel lead times, decision deadlines, budget assumptions, capacity constraints, and named owners determine when a review is useful. Keep Provisional inputs visible, and treat Not recovered inputs as dependencies rather than silently replacing them.

An Operational review checks whether delivery, data health, spending, and guardrails remain within agreed bounds. A Strategic review assesses objective progress and assumptions only after the relevant observation window, conversion lag, and evidence checks have been met. Immediate escalation bypasses the calendar for serious harm, a suspected consent failure, or a hard-limit breach.

A frequent report is not automatically decision-ready evidence. You may monitor a source daily while reserving a strategic decision until its planned evidence window closes.

Reporting frequency is not evidence readiness

A roadmap timeline separates frequent reporting from longer evidence windows and shows an urgent escalation path outside the normal calendar.

Match review timing to phase dependencies and evidence readiness, not merely to how often a dashboard refreshes.

Review purposes

Use the review label to define the purpose before interpreting a result.

Adapt the timing to the Capstone baseline rather than copying an illustrative schedule.
ReviewPrimary focusWhen it occurs
Operational reviewDelivery, source freshness, spending, guardrailsAt the operating cadence that fits the active roadmap phase
Strategic reviewObjective progress and assumptionsAfter the defined observation window and evidence checks
Immediate escalationSerious harm, consent failure, hard-limit breachImmediately; do not wait for a scheduled review

2. Map actions to authority

Give every action an authority

Use only these four operational actions:

  • Maintain: continue an activity within agreed bounds.
  • Investigate: resolve a defined evidence or performance question.
  • Pause: temporarily stop the affected activity or collection.
  • Revise: change an assumption, tactic, scope, or allocation.

Separate roles even when one person holds more than one role. The Measurement owner monitors evidence and documents limits. The Activity owner investigates delivery and can take bounded protective action. The Plan approver authorizes material change. Retain the named roadmap owners where they exist; otherwise use these role labels.

Write each conditional rule in this form: “If [trigger], after [window and checks], then [action], authorized by [approver].” A trigger starts disciplined action; it does not prove causality.

Authorization map

This map clarifies accountability without changing the responsibilities already assigned in the Capstone baseline.

A material Revise action must still fit the approved budget and capacity, or receive renewed approval.
NeedAccountable rolePermitted responseAuthorization boundary
Monitor evidence and data healthMeasurement ownerMaintain or open InvestigateWithin agreed rules and review scope
Resolve a delivery or evidence questionActivity ownerInvestigateWithin approved scope and guardrails
Protect people, consent, or a hard limitActivity ownerPause affected activityImmediate containment; escalate promptly
Change tactic, scope, assumption, or allocationPlan approverReviseMaterial change requires approval and feasibility check

3. Record exceptions without overclaiming

Escalate exceptions and preserve the decision

Immediate escalation is for a serious exception, not a faster version of routine reporting. If a consent failure is suspected, Pause the affected collection promptly and follow your organization’s escalation process. Assess unaffected, compliant activity separately; do not assume that all evidence is invalid.

For a proposed Revise action, check the current Capstone baseline before approval: Does it stay within the approved budget and capacity? Does it disrupt a phase dependency or named owner? If not, the Plan approver must grant renewed approval before the material change proceeds.

Every review leaves a short decision record. It makes the evidence boundary visible and gives the next owner a clear recheck point.

Example

Example decision record

Evidence: Practice evidence — the planned source was unavailable for part of the current observation window.

Limitation: The incomplete window cannot support a conclusion about objective progress.

Action: Investigate

Authorization: Measurement owner opens the investigation within the agreed review scope.

Next review: Operational review after source restoration and the required evidence checks.

4. Practice: route a signal

Route the decision

Match each Practice evidence signal to its correct review, operational action, and accountable role.

Tap an item and then its matching pair.

Step 6 of 8

Prioritize Experiments That Could Change the Plan

Build a small, decision-focused experiment backlog that respects the existing capstone plan, evidence readiness, approved budget, and available capacity.

Choose experiments that could change the plan

Start with the decision at stake

Priority begins with a decision, not a test idea. Start with the Capstone baseline: retain its exact objective wording, selected channels, roadmap phase names, approved budget assumptions, and capacity constraints.

Choose an experiment only when it can resolve a consequential assumption and change a real plan decision—for example, whether an activity can be Maintained, needs to be Revised, or should remain deferred. A novel test that cannot change a decision does not earn scarce capacity.

Keep baseline inputs visibly labeled Confirmed, Provisional, or Not recovered. Do not invent a target, budget limit, owner, or channel simply to make an experiment look ready.

A diagram of experiment candidates passing through decision-value and readiness checks, with one selected and others deferred.

Use a readiness gate after identifying the decision an experiment could change. Importance alone does not make a candidate ready.

Tip

Avoid the test-count trap

A useful experiment may still end with an Inconclusive evidence conclusion. Its value is that the question, measure, evidence requirement, and possible decision are explicit before capacity is committed.

Record readiness before ranking

Use one compact backlog record

Create one comparable record for every candidate. The primary measure must preserve the definition and unit already used in your capstone. The evidence requirement states what would be sufficient for the affected decision; it is not a promise that the experiment will prove causality.

Use the existing named roadmap owner. If no name is available, use the role label Activity owner. Record a missing prerequisite as a Dependency rather than silently treating it as available.

Experiment backlog record template

markdown
Assumption: [Provisional: exact assumption from Capstone baseline]
Decision affected: [decision that could change]
Experiment: [bounded test or diagnostic activity]
Primary measure: [existing metric definition and unit]
Evidence requirement: [precommitted decision-relevant evidence]
Guardrail: [existing budget, consent, fulfillment, or customer-harm limit]
Dependency: [access, consent, asset, traffic opportunity, or delivery capacity]
Conflict: [known interference or none known]
Capacity: [available staff time and approved budget]
Owner: [named roadmap owner or Activity owner]
Sequence: [priority position or Deferred]

Prioritize consequence, uncertainty, and feasibility

Compare candidates qualitatively. Do not convert these judgments into a formula that pretends to create certainty. A readiness dependency can override apparent importance: no authorized access, applicable consent, viable asset, traffic opportunity, or delivery capacity means the experiment must wait.

If a feasible activity cannot generate evidence relevant to its decision, narrow the question, use it only as a clearly limited diagnostic activity, or defer it. Do not claim that it can give a definitive answer.

Qualitative priority checks

Use these checks to make a reasoned priority judgment.

A candidate must be both decision-relevant and feasible within the existing plan.
CheckQuestionPriority implication
Decision consequenceWould the result change an approved plan decision?Higher when the decision is material
Current uncertaintyIs the assumption still unresolved?Higher when uncertainty blocks action
Feasible evidenceCan the primary measure meet its evidence requirement?Defer or narrow if not
ReversibilityCan the activity remain bounded by its guardrail?Prefer safer, reversible learning
CapacityDoes it fit approved budget and available staff time?Defer lower-value work when it does not

Check a priority decision

Select the first experiment

Your current roadmap phase has capacity for one experiment. Which candidate should be ranked first?

Apply the backlog to your capstone

Create your prioritized experiment backlog

Build a small ranked backlog from your Capstone baseline. Use up to three candidate experiments that fit the current roadmap phase.

For each candidate, record: Assumption, Decision affected, Experiment, Primary measure, Evidence requirement, Guardrail, Dependency, Conflict, Capacity, Owner, and Sequence.

Then identify your first-ranked candidate and justify it using both decision consequence and feasible evidence. Mark at least one lower-priority candidate as Deferred, and state which existing capacity limit means it must wait. Keep unrecovered inputs visible as Not recovered or as a dependency; do not replace them with invented details.

Write at least 300 characters (0/300).

Step 7 of 8

Sequence Experiments and Precommit to Evidence

Schedule capstone experiments without contaminating their evidence, and precommit to a response when results are inconclusive.

Find conflicts in the existing plan

Check for interference before scheduling

Use the prioritized experiment backlog and roadmap from your Capstone baseline. Do not add a new channel, budget line, or team member to make a test fit.

Before scheduling, run a conflict check across every proposed experiment and nearby operational change. A conflict exists when changes share an audience, landing page, offer, budget, tracking configuration, or constrained staff capacity. Different channels can still conflict when they alter the same customer path or draw on the same limited resource.

For each Capstone baseline input, retain its status: Confirmed, Provisional, or Not recovered. A Not recovered dependency is a reason to defer or narrow a test—not a reason to invent a schedule or capacity allowance.

A conflict is about shared conditions

A timeline diagram shows two experiments colliding through shared customer-path and resource connections, while a third is scheduled after them.

Treat shared audiences, pages, offers, budgets, tracking, and people as potential interference—not merely shared channel names.

Commit before results arrive

Sequence, isolate, or defer

Repair a conflict with one of three choices:

  • Sequence: complete the first experiment’s committed window and review before changing the shared condition.
  • Isolate: run work in parallel only when the audience, customer path, budget, tracking, and constrained capacity are genuinely separated.
  • Defer: move a lower-value change outside the available roadmap capacity or until its dependency is resolved.

Parallel activity is not proof of a valid comparison. Preserve the conditions needed to answer one decision question at a time.

Pre-launch decision commitment

Write this commitment before launch for every scheduled experiment. It prevents a favorable early signal from being reinterpreted as a winner.

Complete these fields using the metric definitions, targets, limits, owners, and phases already selected for your capstone.
Commitment fieldWhat to specify
Decision questionWhich plan decision could change?
Primary measureOne defined measure from the capstone
ComparisonWhat the result will be compared with
Observation windowComparable period allowing for conversion lag
Evidence requirementPredefined basis for adoption, rejection, or Inconclusive
GuardrailBudget, fulfillment, customer-harm, consent, or data-quality limit
Owner and review pointNamed owner and scheduled decision review
Inconclusive pathAuthorized Maintain, Investigate, Pause, or Revise response; any extension or redesign needs approval and remaining capacity

Practice: repair a schedule conflict

Choose the defensible repair

Practice evidence: Two proposed items in your capstone backlog affect the same selected audience. One changes the offer on a selected landing page; the other changes the offer in an email that sends that audience to the same page. Both also require the same Activity owner during the same roadmap phase.

Which schedule best protects a decision-relevant comparison within the existing plan?

Build your defensible experiment record

Apply the sequence and commitment

Create one concise pre-launch record for your capstone.

  1. Select two backlog entries—or one entry and a planned operational change—that share an audience, page, offer, budget, tracking configuration, or limited staff capacity. Identify the conflict and state whether you will sequence, isolate, or defer it. If your current entries do not overlap, use the next planned change to a shared condition as the proposed scheduling clash; do not introduce a new channel.
  1. For one experiment, write its Decision question, Primary measure, Comparison, Observation window, Evidence requirement, Guardrail, Owner, and review point.
  1. State the response if the evidence is Inconclusive. Name one authorized operational action—Maintain, Investigate, Pause, or Revise—and specify what approval and remaining roadmap capacity would be required before any extension or redesign.

Write at least 450 characters (0/450).

Step 8 of 8

Apply and Finalize Your Decision Framework

Integrate your existing capstone measurement matrix, decision rules, review authority, safeguards, and experiment sequence into a bounded response to uncertain evidence.

1. Assemble and check the framework

Assemble one governed framework

Your framework is complete only when it governs the same plan you already approved. Open the Capstone baseline and assemble one working record for your capstone business:

  • Keep the exact objective wording, metric definitions, units, selected and deferred channels, roadmap phase names, budget assumptions, and capacity constraints.
  • Keep each measurement-matrix row: Objective, Business outcome, Supporting signal, Source, Owner, and Review cadence.
  • Keep the linked decision-rule fields, review authority, and experiment-backlog sequence.
  • Mark each input Confirmed, Provisional, or Not recovered. A Not recovered item stays visible as a dependency; it is not silently replaced.

This is an integration check, not a new planning method. The final framework should let a reviewer trace every action back to the existing plan and its evidence limits.

One framework, connected decisions

A capstone plan connected to measurement, decision, review, experiment, consent, budget, capacity, and approval components.

Each component must point back to the same Capstone baseline.

Consistency inspection

Use this final inspection before treating the framework as ready.

Record unresolved items as dependencies, not as evidence of readiness.
Framework elementCheck against Capstone baselineRecord result
Objectives and measuresExact wording, definitions, units, and targets or limits matchConfirmed, Provisional, or Not recovered
Channels and phasesOnly selected channels and existing roadmap phases appearConfirmed, Provisional, or Not recovered
Rules and authorityOwners, approvers, review labels, and guardrails match the roadmapConfirmed, Provisional, or Not recovered
ExperimentsSequence fits approved budget, capacity, dependencies, and conflictsConfirmed, Provisional, or Not recovered
Evidence readinessSource access and applicable consent status are documentedConfirmed, Provisional, or Not recovered

2. Interpret uncertain evidence

Integrated stress test

Run this Practice evidence stress test against your own framework. It is relative to your capstone’s targets and limits; it is not a claim about your business.

At an Operational review, the consolidated business outcome is still inside its defined observation window. Two included channels each report a channel-attributed result for conversions that may overlap. An authorized source for one supporting signal has an access interruption, and the current consent coverage for that collection cannot be verified.

Meanwhile, an Activity owner proposes a change that would alter a shared audience, landing page, offer, budget, tracking configuration, or another dependency used by a scheduled experiment.

Do not turn early outcome data into a verdict. Do not add channel claims together as unique business results. Do not work around a consent or access problem with alternate identifiers or tracking. Check whether the proposed change conflicts with the experiment sequence before authorizing it.

Evidence boundaries

Read the evidence by what it can support—not by what you want it to prove.

The same evidence can justify monitoring or investigation while remaining insufficient for a causal conclusion.
Practice evidenceWhat is knownLimitation
Business outcome is still in its observation windowThe outcome can be monitored against its defined measureIt is not decision-ready performance evidence yet
Two channel-attributed results overlapEach claim can be diagnostic channel evidenceThey cannot be summed as unique business results
Supporting-signal access or consent coverage is uncertainThe collection condition needs prompt verificationThe affected comparison may be misleading; the conclusion is Inconclusive
Proposed shared change overlaps an experiment dependencyA sequencing conflict may existDo not assume parallel changes can answer the experiment question

3. Write the authorized response

Final application: record a bounded response

Create a concise decision record for the Practice evidence using your completed framework. Apply only the objectives, channels, phases, budget assumptions, capacity constraints, owners, and rules in your Capstone baseline.

Use these exact headings:

Evidence — distinguish the immature business outcome, channel-attributed claims, and affected supporting signal.

Limitation — state what is Inconclusive, including the observation-window, overlap, and access or consent limits.

Action — state the scoped operational action or actions: Maintain, Investigate, Pause, or Revise. Do not create another action label.

Authorization — identify who may contain an urgent consent risk, who investigates, and who authorizes a material change.

Next review — name the appropriate review label and the evidence needed before a conclusion or change.

Then add a short Framework self-check confirming that your response preserves: the exact metric and unit; selected channels and roadmap phase; approved budget and capacity; applicable consent and source-access dependency; and the experiment’s dependency, conflict, and sequence.

Write at least 600 characters (0/600).

4. Close with accountable readiness

Summary

Framework final review

A usable decision framework does not promise certainty. It makes uncertainty visible, protects people and operating limits, and specifies what evidence must arrive before the plan changes.

  • A business outcome is consolidated performance evidence; overlapping channel-attributed results are not additive unique results.
  • An immature window, changed collection condition, missing access, or uncertain consent coverage can make a conclusion Inconclusive.
  • Use only Maintain, Investigate, Pause, and Revise as operational actions, with the appropriate authorization.
  • Keep actions within the Capstone baseline’s budget, capacity, selected channels, phases, and experiment sequence.
  • Leave Not recovered inputs and unresolved access or consent issues visible as dependencies.

Decision framework complete

Congratulations! You completed: Set Decision Rules for Measurement and Experiments

You finalized a measurement and experimentation framework that can respond to uncertain results without overstating evidence or bypassing safeguards.

Download the App now to have access to + 5000 free courses, exercises, certificates and lots of content without paying anything!

  • 100% free online courses from start to finish

    Thousands of online courses in video, ebooks and audiobooks.

  • More than 60 thousand free exercises

    To test your knowledge during online courses

  • Valid free Digital Certificate with QR Code

    Generated directly from your cell phone's photo gallery and sent to your email

Cursa app on the ebook screen, the video course screen and the course exercises screen, plus the course completion certificate