diff --git a/ARCHITECTURE.md b/ARCHITECTURE.md index 50f86ed..055eaa7 100644 --- a/ARCHITECTURE.md +++ b/ARCHITECTURE.md @@ -306,6 +306,7 @@ gets fixed. - launches/new-product-launch/references/launch-narrative.md (for net-new products) - strategy/messaging-architecture - marketing-science/research + - marketing-science/measurement - marketing-science/brand-equity - positioning/{relevant product} - brand-voice/{relevant product} @@ -336,6 +337,11 @@ gets fixed. marketing-science/research - foundation/marketing-os + marketing-science/measurement + - foundation/marketing-os + - marketing-science/research + (measurement contract, metric hierarchy, instrumentation, and decision rules) + marketing-science/archetyping - foundation/marketing-os diff --git a/README.md b/README.md index 60f645b..2eccd90 100644 --- a/README.md +++ b/README.md @@ -14,7 +14,7 @@ The repository is organized into seven top-level folders: 4. **strategy/** — Higher-order strategy skills, including the gated Compound Brand workflow, messaging architecture, and one-pagers 5. **craft/** — Identity and execution skills, including verbal identity, art direction, copywriting, editing, naming, and channel-specific output 6. **launches/** — Orchestration skills for three launch tiers plus the flagship GTM, calendar, brief, claims, wireframe, and email systems -7. **marketing-science/** — Research, archetyping, and brand-equity theory that inform strategy and identity +7. **marketing-science/** — Research, measurement, archetyping, and brand-equity theory that inform strategy and identity ## Dependency model @@ -28,6 +28,7 @@ Ready for structured use: - Every master positioning and brand voice - Evidence-led research, competitive auditing, and positioning matrices +- Decision-led measurement contracts for campaigns, launches, and experiments - Compound Brand strategy through art direction - Aaker-led brand-equity modeling - Messaging architecture and flagship launch systems diff --git a/foundation/marketing-os/references/method.md b/foundation/marketing-os/references/method.md index 4a37df0..e50adbe 100644 --- a/foundation/marketing-os/references/method.md +++ b/foundation/marketing-os/references/method.md @@ -112,6 +112,36 @@ CAC, MQLs, click-through — those are Austin's territory, the growth team's. Th --- +## Measurement doctrine + +Measurement exists to drive decisions. Dashboards are a reporting surface. Every +strategy, campaign, launch, and experiment names the objective, causal hypothesis, +baseline, primary decision metric, target or threshold, diagnostics, guardrails, data +source, owner, window, attribution limits, and decision rule before results arrive. Load +`marketing-science/measurement` to build the contract. + +**One objective, one primary decision metric.** Supporting numbers can explain the +result or protect against unintended damage. They cannot all be north stars. Volume is +paired with quality or audience composition, and percentages show their denominators. + +**Marketing and Growth share evidence without collapsing mandates.** Marketing measures +demand, memory, associations, cultural position, earned attention, and audience quality. +Growth measures conversion, activation, revenue, retention, and funnel efficiency. A +shared launch needs both views and named owners. Last-click attribution does not become +the final account merely because it is available. + +**What is measurable is not identical to what matters.** Hard-to-attribute effects still +require observable signals, appropriate time horizons, and honest confidence labels. +Quantitative evidence is paired with customer behavior, qualitative evidence, and +cultural context. Judgment stays in the system; unsupported intuition does not get a +free pass. + +**No baseline means no improvement claim.** When the baseline is missing, say so, +establish one, and name which conclusions remain unavailable. Results end in a decision: +continue, change, stop, invest, or collect better evidence. + +--- + ## Taste as strategic asset **Taste defined.** Taste is point of view — the judgment that something belongs. Which idea is worth writing about, which AI tool is worth covering, which guest is worth interviewing, which sentence earns its place, which campaign is worth shipping. Taste is the quality of attention paid to a decision most companies make on autopilot. A high taste level means, by definition, that some people won't "get it." Good. diff --git a/launches/canonical-process.md b/launches/canonical-process.md index e8b9b72..126f050 100644 --- a/launches/canonical-process.md +++ b/launches/canonical-process.md @@ -96,6 +96,8 @@ note in the standard format (below). - Dream tweet and dream headline — verbatim and specific enough to score at the retro; the headline should validate the intended category belief or customer outcome - Positioning and story: what we made, why we made it, who it's for +- Name the measurement owner, inventory available data and instrumentation, and capture + the pre-launch baseline or mark it `NO BASELINE` - Deep competitive audit — messaging, positioning, colors, pricing (`competitive-audit` skill) - Name the pricing/offer owner — Brandon and the GM lock pricing, targeted for wireframe approval - Naming phase if needed, per Rule 5 @@ -135,7 +137,9 @@ note in the standard format (below). - Design the full website; ship to the builder within two weeks of green light - Builder builds, one–two weeks; QA owner named, definition of done agreed at handoff - Product designer and GM apply the brand in the app, from the locked brand book -- Measurement and KPIs locked in Growth's GTM plan, metrics included — Growth owns measurement +- Growth completes the measurement contract with `marketing-science/measurement` inside + the GTM plan: objective, hypothesis, baseline, primary metric, formulas, targets, + diagnostics, guardrails, sources, owner, window, attribution limits, and decision rules - Austin locks the GTM plan one week before launch: customer, campaign hierarchy, proof/seeding system, starter behavior, editorial role, video suite, measures, and owners - Douglas builds the Marketing Launch Calendar — including retro and sustain — into the Notion Calendar @@ -159,7 +163,9 @@ note in the standard format (below). ### Post-launch -- Retro — always. Scored against the Week 1 success definition: dream tweet, dream headline, metrics +- Retro — always. Scored against the Week 1 success definition and pre-committed + measurement rules: dream tweet, dream headline, primary outcomes, diagnostics, and + guardrails. End with a decision, not a metric recap. - Sustain per the GTM plan and Marketing Launch Calendar — flexes by launch - Customer receipts, seeded-user evidence, and case studies feed the next acquisition cycle; proof targets are scored at the retro alongside business and brand outcomes @@ -205,8 +211,11 @@ Established July 2026 on the All Access, Every Agent, and Cora 2.0 launches. The 3. **Naming.** Per Rule 5. 4. **Red light.** One more round; solve the reasons within a week if possible; budget one–two weeks before re-presenting. Show the slack on the calendar. -5. **Measurement.** Growth owns measurement and KPIs, locked inside the GTM plan. - Marketing's retro scores against the Week 1 success definition. +5. **Measurement.** Growth owns measurement and KPIs, locked as a measurement contract + inside the GTM plan with `marketing-science/measurement`. Marketing owns the brand, + demand, cultural, and qualitative outcomes it expects the contract to include. + Marketing's retro scores against the Week 1 success definition and the contract's + decision rules. No baseline means no improvement claim. 6. **In-product brand.** The GM builds the app. Marketing connects with the GM and engineering twice during the brand build and once at brand lock; the product designer and GM incorporate the identity. Not marketing's workstream. @@ -255,7 +264,7 @@ A new launch's roster is collected, never inherited. These are examples to offer beats a complete-looking false one. - Check every item the same way; name what wasn't covered instead of implying it passed. -## Skill registry (status as of July 18, 2026) +## Skill registry (status as of July 23, 2026) | Skill | Status | Role | |---|---|---| @@ -265,6 +274,7 @@ A new launch's roster is collected, never inherited. These are examples to offer | codex-gut-check | **New in this bundle** | The Week 5 bridge | | launch-brief | **New in this bundle** | Week 1 foundation generator | | gtm-plan | **New in v1.4** | Market belief, campaign, proof, activation, and measurement strategy | +| marketing-science/measurement | **New in v1.5** | Measurement contract, metric hierarchy, instrumentation, and decision rules | | claims-clearance | **New in this bundle** | Pre-ship verification gate | | wireframe-from-copy | **New in this bundle** | Copy doc → annotated lo-fi wireframe | | launch-email-flows | **New in this bundle** | Four-flow launch email pattern | @@ -273,6 +283,14 @@ A new launch's roster is collected, never inherited. These are examples to offer ## Changelog +**v1.5 — July 23, 2026 (Douglas × Codex)** +- Added `marketing-science/measurement`, a reusable measurement contract for strategies, + campaigns, launches, experiments, and ongoing programs. +- Required a baseline, primary decision metric, formula, target, diagnostics, guardrails, + instrumentation, owner, window, attribution limits, and pre-committed decision rules. +- Kept Marketing and Growth measures connected but distinct, and required launch retros + to end in a decision rather than a dashboard recap. + **v1.4 — July 18, 2026 (Douglas × Codex)** - Added `launches/gtm-plan`, the missing strategy layer between the approved launch brief and channel execution. diff --git a/launches/gtm-plan/SKILL.md b/launches/gtm-plan/SKILL.md index 09931fc..ce13b45 100644 --- a/launches/gtm-plan/SKILL.md +++ b/launches/gtm-plan/SKILL.md @@ -150,6 +150,12 @@ Every clip should prove one claim. Do not turn the campaign into a feature tour. ## Measurement +Load `marketing-science/measurement` and complete its measurement contract before launch. +The contract must define the objective, hypothesis, baseline, primary decision metric, +formula, target or threshold, diagnostics, guardrails, data source, owner, window, +attribution limits, and pre-committed decision rule. A list of plausible metrics is not +a measurement plan. + Lock four classes of measures with dates and owners: - **Business:** qualified demand, paid conversion, revenue, capacity, retention, and referrals @@ -161,6 +167,10 @@ Label each measure as offer, campaign, activation, or operations. Activation com Use a funnel model before launch when capacity, revenue, or sales targets matter. Show the assumptions and track quality or composition beside volume. A full event with the wrong room or a large waitlist of poor-fit applicants is not success. +If no baseline exists, write `NO BASELINE`, establish the earliest defensible observation +window, and do not claim improvement. Name the action the team will take for success, +mixed evidence, failure, and an inconclusive result. + ## Required output Produce two layers. @@ -175,7 +185,8 @@ Use seven sections and keep them brief: 4. **The campaign** — platform, enemy, proposition, optional campaign demand, conversion action, payoff, articulation, and proof system 5. **The journey** — conversion and lifecycle path, including decision points and recovery 6. **The route to market** — three to six phases, each with one job and only its required proof or action -7. **The scorecard and gates** — funnel assumptions, a small set of outcomes, launch gates, and owners for open decisions +7. **The scorecard and gates** — the completed measurement contract, funnel assumptions, + a small set of outcomes, launch gates, and owners for open decisions ### Chosen appendices @@ -200,6 +211,7 @@ Dream tweets and headlines are optional stress tests for positioning. Use them o | Planning a curated or application-based event | `references/curated-event-gtm.md` | | Auditing whether an activation has replaced the offer | `references/concept-inflation-case.md` | | Verifying Every Agent market claims | `marketing-science/messaging/references/evidence-bank-every-agent.md` | +| Building the scorecard, targets, instrumentation, and decision rules | `marketing-science/measurement` | | Building the upstream message hierarchy | `strategy/messaging-architecture` | | Producing the launch narrative | `launches/new-product-launch/references/launch-narrative.md` | @@ -210,7 +222,7 @@ Dream tweets and headlines are optional stress tests for positioning. Use them o - `launches/launch-brief` (approved input) - `launches/new-product-launch/references/launch-narrative.md` (for net-new products) - `strategy/messaging-architecture` -- `marketing-science/research` · `marketing-science/brand-equity` +- `marketing-science/research` · `marketing-science/measurement` · `marketing-science/brand-equity` - `positioning/{relevant product}` · `brand-voice/{relevant product}` - `launches/claims-clearance` @@ -224,6 +236,9 @@ Dream tweets and headlines are optional stress tests for positioning. Use them o - [ ] Campaign platform, enemy, proposition, optional demand, conversion action, and payoff each have a clear job - [ ] No optional activation has silently become an admissions rule, onboarding task, PR premise, or success metric - [ ] Offer, campaign, activation, and operations measures remain separate +- [ ] The measurement contract has a baseline, one primary metric per objective, formulas, + targets, diagnostics, guardrails, sources, owners, windows, attribution limits, and + decision rules - [ ] Proof exists before channel volume begins - [ ] The journey names decisions, handoffs, service levels, failure states, and follow-up - [ ] Application data and private logistics are separated for gated experiences diff --git a/marketing-science/measurement/SKILL.md b/marketing-science/measurement/SKILL.md new file mode 100644 index 0000000..177a976 --- /dev/null +++ b/marketing-science/measurement/SKILL.md @@ -0,0 +1,175 @@ +# Measurement + +Build a measurement contract that turns a marketing objective into an observable result +and a decision. Use this skill before a campaign, launch, experiment, or sustained +marketing program begins. The contract tells the team what it is trying to learn, how +the evidence will be interpreted, and what happens next. Dashboard specifications come +later. + +Measurement supports judgment. It does not replace it. What is easy to count is not +automatically what matters. Hard-to-attribute effects still need evidence. + +## When to invoke + +- When a strategy, campaign, launch, experiment, or ongoing program needs success criteria +- When a brief contains goals such as awareness, demand, cultural impact, or engagement + without defined evidence +- When a dashboard has many metrics but no primary decision metric +- When Marketing and Growth need a shared scorecard without collapsing their mandates +- When a result must be interpreted, compared with a baseline, or turned into a decision +- Before launch, when instrumentation, ownership, and measurement windows can still be fixed + +## Start with the decision + +Write the decision this evidence must inform before choosing a metric. Examples: + +- Continue, change, or stop the campaign +- Increase or reduce investment +- Keep or revise the message, offer, audience, channel, or first-value path +- Promote an experiment into a repeatable program +- Decide whether the launch created demand, converted it, or merely generated activity + +If the result would not change a decision, the metric is context rather than a key +performance indicator. Keep it only when it helps explain the primary result. + +## The measurement contract + +Every contract contains these fields. Mark an unknown `OPEN`; do not fill it with an +invented benchmark. + +| Field | Required decision | +|---|---| +| Objective | The business, customer, brand, or learning outcome this work should create | +| Hypothesis | The causal claim: if we do X for Y, we expect Z because of Q | +| Unit of analysis | Person, account, subscriber, customer, session, cohort, market, or other defined unit | +| Baseline | Current value, source, and comparison period; write `NO BASELINE` when none exists | +| Primary decision metric | The single measure that best determines whether the objective was met | +| Formula | Numerator, denominator, exclusions, and segmentation rules | +| Target or threshold | The value and date that separate success, mixed evidence, and failure | +| Diagnostic metrics | A small set that explains why the primary metric moved or did not | +| Guardrails | Measures that catch damage, poor-fit growth, degraded quality, or unintended effects | +| Qualitative and cultural signals | Named evidence that captures meaning the quantitative measures miss | +| Data source and instrumentation | System of record, events or fields required, access, and known gaps | +| Owner | The person accountable for collection, analysis, and reporting | +| Window and cadence | Start, stop, maturation period, reporting rhythm, and comparison period | +| Attribution limits | What the method can and cannot claim caused the result | +| Decision rule | The action for success, mixed evidence, failure, and an inconclusive result | + +Write decision rules before results arrive. A useful rule is explicit: + +> If the primary metric reaches [target] by [date] without breaching [guardrail], take +> [action]. If it lands in [mixed range], inspect [diagnostics] and take [bounded action]. +> If it falls below [failure threshold], take [action]. If data quality is insufficient, +> fix [instrumentation] before drawing a conclusion. + +## Metric hierarchy + +Use four layers. Do not give every available number equal status. + +1. **Primary decision metric:** one measure tied directly to the objective +2. **Diagnostics:** usually two to five measures that explain the primary result +3. **Guardrails:** usually one to three measures that prevent a local win from becoming + an overall loss +4. **Context:** trends and observations worth monitoring but not used to declare success + +Use a rate when opportunity or exposure varies. Pair volume with quality or composition. +Report absolute numbers beside percentages when a small denominator could mislead. + +For launches, retain the GTM plan's four outcome classes: + +- **Business:** qualified demand, paid conversion, revenue, capacity, retention, referrals +- **Conversion:** decision time, first value, completion, activation, repeat behavior, + attendance, or participation +- **Proof:** publishable receipts, advocates, outcomes, and case-study candidates +- **Brand:** awareness, branded search, direct and organic traffic, earned coverage, + distinctive associations, and reuse of campaign language + +These classes organize the scorecard. They do not eliminate the need to choose one +primary decision metric for each objective. + +## Marketing and Growth + +Marketing creates demand and cultural gravity. Growth converts demand through the offer, +funnel, onboarding, lifecycle, and retention. Measurement should show both jobs without +pretending they are the same. + +- Marketing owns the hypothesis and evidence for awareness, memory, associations, + cultural position, earned attention, audience quality, and demand creation. +- Growth owns instrumentation and evidence for conversion, activation, revenue, + retention, and funnel efficiency. +- Shared work requires a shared contract with named owners. Attribution software does not + settle the mandate split. +- Last-click and platform-reported attribution are diagnostic views, not complete accounts + of marketing effectiveness. + +When brand effects will mature outside the campaign window, name the expected time +horizon and use leading signals without presenting them as final proof. + +## Data quality and interpretation + +Before interpreting a result, check: + +- Numerator, denominator, exclusions, and duplicate handling +- Sample size and whether the sample represents the intended audience +- Missing data, instrumentation changes, and inconsistent event definitions +- Cohort, channel, customer quality, and audience-composition differences +- Seasonality, launch novelty, promotions, pricing changes, and other confounders +- Correlation versus plausible causation +- Selection, survivorship, response, and confirmation bias +- Whether the comparison period or benchmark is genuinely comparable +- Whether a statistically visible change is large enough to matter operationally + +Do not manufacture precision. Small samples, noisy brand signals, and directional +qualitative evidence should be labeled honestly. Triangulate when no single measure can +carry the claim. + +## Baselines and instrumentation + +Capture the baseline before work begins whenever possible. Use the same definition, +segment, and time basis for the comparison. + +If no baseline exists: + +1. Write `NO BASELINE`. +2. Choose the earliest defensible observation window. +3. Separate the act of establishing a baseline from the claim of improvement. +4. Add the missing event, field, survey, or manual collection step before launch. +5. State which conclusions remain impossible until enough data accrues. + +The minimum viable instrumentation is the smallest reliable setup that can execute the +decision rule. Do not delay useful work to build an ornamental dashboard. + +## Required output + +Return: + +1. The completed measurement-contract table +2. A scorecard showing the primary metric, diagnostics, guardrails, owner, source, + baseline, target, window, and current status +3. The pre-launch instrumentation checklist +4. The result interpretation, when data exists +5. The resulting decision and next measurement date + +Lead with the decision, not the metric dump. Explain quantitative findings in plain +language and show calculations when the result depends on them. + +## Dependencies + +- `foundation/marketing-os` +- `marketing-science/research` + +## Quick checklist + +- [ ] The decision this evidence must inform is explicit +- [ ] Objective and causal hypothesis are written separately +- [ ] The unit of analysis and metric formulas are defined +- [ ] A real baseline is cited, or `NO BASELINE` is visible +- [ ] Each objective has one primary decision metric +- [ ] Diagnostics, guardrails, and context are visibly distinct +- [ ] Targets, thresholds, windows, and decision rules were set before results arrived +- [ ] Data source, instrumentation gaps, and owner are named +- [ ] Quality or audience composition is tracked beside volume +- [ ] Attribution limits and confounders are stated +- [ ] Marketing and Growth outcomes remain distinct but connected +- [ ] Qualitative and cultural signals cover important effects the numbers miss +- [ ] The result ends in a decision, not a dashboard recap