All Pages

TitleSectionDateDescription
Experiments — First-Party Results IngestCompiled2026-07-07The workflow that turns concluded marketing experiments from factory projects into KB knowledge: one compiled note per concluded experiment, updating benchmark priors and SSR calibration. This is what makes the KB a factory brain instead of a library.
TEA AuditCompiled2026-06-12The Test → Evaluate → Assess audit run by the test-lead agent at the end of each implementation sprint. Outputs tea-summary.md with gate decision: PASS, PASS-WITH-FOLLOWUPS, or REMEDIATE. Covers functional, non-functional, and quality dimensions.
GEO Prompt ResearchCompiled2026-06-26The 'keyword research' of GEO: discovering and prioritizing the full natural-language prompts buyers ask LLMs. Source from sales/support logs, community mining, PAA, and query fan-out; classify by intent; prioritize by buying-intent × citation-gap. The output is the factory's content backlog and measurement panel.
Geo-ExperimentCompiled2026-06-26Randomize geographic markets into control/treatment to measure ad incrementality when users can't be split. Geo-based regression predicts the counterfactual from control geos and reads iROAS directly as a model coefficient. Rule of thumb: ≥30 geos, ~4–8 week pretest + 3–5 week test.
GTM Archetype SelectionCompiled2026-06-26The five primary go-to-market motions — Sales-Led, Product-Led, Community-Led, Outbound-Led, and Hybrid — each chosen against ACV, sales-cycle, ICP concentration, and product self-serviceability. Most growth-stage companies run two motions; add a second when >20% of revenue is lost to a cohort a different motion could capture.
Incrementality TestingCompiled2026-06-26Randomized controlled tests that isolate the conversions a channel actually caused, versus those that would have happened anyway. Lift = (test − control) / control. Methods, cleanest to most practical: user-level holdouts, PSA/ghost ads, and geo-experiments — the causal answer attribution cannot give.
Category DesignCompiled2026-06-14Defining a new market category instead of competing in an existing one. The creator sets the terms of comparison and becomes leader by default — a structural advantage positioning alone can't give, at the cost of a multi-year market-education bet.
Cold Email / Outbound SequenceCompiled2026-06-14A multi-touch, multi-channel outreach cadence (email + LinkedIn + calls) designed to book meetings. Reply rate peaks at touch 4–5; coordinated email+LinkedIn lifts replies 30–50% over email-only.
Bayesian Decision Rule (Expected Loss)Compiled2026-06-26Decide ship/kill on thin data by computing the probability a variant wins and its expected loss, then shipping when expected loss falls below an operator-set 'threshold of caring'. Needs no pre-ordained sample size and degrades gracefully — on sparse data it just says 'not yet' instead of producing false confidence.
Marketing Budget AllocationCompiled2026-06-26Set the spend mix from stage norms, then move money toward the highest marginal incremental ROAS until it equalizes or hits the ROAS floor. Benchmarks tell you where to start; incrementality tells you where to move. Allocation is stage-dependent — early concentrates, scale diversifies.
Agent OrchestrationCompiled2026-06-26The conductor layer that turns individual marketing agents into a factory: a central orchestrator decomposes work, routes it to specialized workers, enforces hand-off rules, and decides termination — without doing the work itself. The orchestrator-workers pattern is the most widely deployed in production.
Event-Tracking SchemaCompiled2026-06-26The governed tracking plan that defines what you measure: which events, which properties, which identity keys, and which data-quality rules — enforced at ingestion. The contract that keeps the warehouse queryable; without it every team invents its own event names and analytics become noise.
Minimum Detectable Effect (MDE)Compiled2026-06-26The smallest lift a test can detect given traffic, baseline rate, and run time. Required sample scales with 1/effect², so at low traffic only large swings are measurable — the MDE gate decides whether a live A/B test can ever finish before it is run.
Directory LifecycleCompiled2026-06-12A structured folder hierarchy that manages content from raw idea to published piece: ideas.md → queue/ → drafts/ → review/ → approved/ → published/. File movement between directories triggers automated pipeline actions. The structural backbone of the content machine.
Engagement TrackingCompiled2026-06-12Track engagement on social media posts and content via platform APIs. Identify top performers for content recycling.
Agent Ownership BoundaryCompiled2026-06-14The line between marketing tasks AI agents can own end-to-end (execution, operations, enrichment, drafting, monitoring) and tasks that still require human judgment (strategy, creative direction, crisis comms, ambiguous-signal interpretation).
Calibration LoopCompiled2026-06-14Logging a predictive tool's picks against real post-publication outcomes every cycle, and trusting it only as far as its observed hit rate. The mandatory discipline that turns an unproven synthetic panel into a calibrated one — and generates publishable first-party benchmark data.
llms.txtCompiled2026-06-14A proposed root Markdown file giving LLMs a human-curated shortlist for inference-time retrieval. As of 2026 no major vendor commits to fetching it and studies show no measurable lift — but it's near-zero cost, so ship it for optionality; the adjacent .md-suffix convention has independent value today.
RALPH Protocol (Retry And Learn Protocol)Compiled2026-06-12Kelly's structured retry/escalation protocol: 3 max retries with diagnostic passing between attempts. Same error twice = escalate immediately.
Expansion PlaybookCompiled2026-06-26The repeatable motion that compounds revenue from existing accounts: usage/milestone triggers fire seat/tier upsell, cross-sell, and usage-upgrade plays, opened from realized value. Expansion candidates come from the health score; win-back recovers churning accounts while warm. NRR is the scorecard.
ActivationCompiled2026-06-14Getting a new user to the 'aha moment' — the specific action beyond which they recognize the product's core value and are substantially less likely to churn — as fast as possible, via progressive onboarding and time-to-value optimization.
Social Media AutomationCompiled2026-06-12The automated drafting, scheduling, and engagement tracking of social media content across platforms. Drafts threads, LinkedIn posts, and Instagram captions from content ideas; tracks engagement via platform APIs; and integrates with the content machine's publishing pipeline.
Event-to-Pipeline AttributionCompiled2026-06-26Crediting events for pipeline they influence: events rarely close a deal alone, so use multi-touch attribution with event as a distinct touch type, rigorous per-send UTM tagging, and a tiered follow-up SLA (hot leads in 2 hours) — because event leads go cold within 48 hours.
Freemium + Subscription ModelCompiled2026-06-12The primary monetization strategy used across Kelly's first product portfolio: free tier with core functionality to drive adoption, premium subscription tier ($2.99-$9.99/mo) for advanced features, powered by RevenueCat for subscription management. Target: recurring revenue over one-time transactions.
Growth FlywheelCompiled2026-06-26HubSpot's reconception of growth as a self-reinforcing wheel (Attract → Engage → Delight) rather than a linear funnel: existing customers are the source of the next customers, so growth compounds instead of capping at spend. Momentum = velocity × mass; friction at any stage slows or reverses the wheel.
Quick-PathCompiled2026-06-12A streamlined variant of the software factory pipeline for well-defined, low-uncertainty tasks like bug fixes. Skips the research phase entirely and goes straight through planning → implementation → test. Reduces cycle time for bounded work.
GEO Citation MeasurementCompiled2026-06-26Measuring whether you actually get cited in AI answers, since Search Console can't. Run a fixed prompt panel across engines on a schedule; track AI Citation Frequency, share of AI voice, and citation gap. Citation is probabilistic across query fan-out, so measurement is a small-N experiment — sample each prompt repeatedly and decide with priors, not single runs.
Lead Routing & SLACompiled2026-06-26Turning score thresholds into fast, accountable assignment: hot leads auto-route to a rep within minutes (speed-to-lead ≈ 7× conversion within the hour), the rep accepts/rejects within ~48h with mandatory feedback, and marketing+sales share one contract of definitions and SLAs. The feedback loop keeps scoring honest.
RevenueCatCompiled2026-06-12Subscription management platform used in Kelly's app factory. Handles in-app purchases and subscriptions across iOS and Android with minimal integration code.
Demand CreationCompiled2026-06-26Sparking new interest in buyers who weren't looking — the counterweight to demand capture. Capture harvests existing in-market demand and has a ceiling; creation grows the addressable pipeline. It works through dark-social channels attribution can't track, which is why it's chronically under-funded.
robots.txt (AI Crawler Policy)Compiled2026-06-14The root file declaring which AI user-agents may crawl. In 2026 the critical distinction is by crawler function: blocking the search/retrieval crawlers (OAI-SearchBot, Claude-SearchBot, PerplexityBot) means you won't be cited — so a GEO-aware policy opts INTO them deliberately.
Content RecyclingCompiled2026-06-12The practice of finding old high-performing content posts and regenerating fresh platform-specific versions for each social channel. Identifies top performers from engagement data, then uses the research-to-draft pipeline to create new variations optimized for each platform.
Agent Workflow Pattern (Trigger → Input → Action → Output)Compiled2026-06-14The machine-readable shape of a factory agent task: a trigger fires, defined inputs are gathered, an action runs, and a structured output is produced and logged. The SSR pre-publication test is one worked example of the pattern.
Autonomous Marketing AgentCompiled2026-07-07An agent that owns a marketing channel end-to-end — sense, think, act, learn — with human input measured in seconds per day, optimizing against revenue rather than vanity metrics. Distinct from AI-assisted marketing (human owns, AI helps) and from scheduling automation (no learning loop).
Factory Experiment LoopCompiled2026-06-26The repeatable loop an agent runs to test a marketing variant on thin data: feasibility gate → optional synthetic pre-screen → pre-registered test (method by context) → loss-based decide → log. Meta-learning across many small experiments is the factory's structural edge over one-off testers.
Marketing AttributionCompiled2026-06-14Assigning credit for conversions across touchpoints — last-click vs. rules-based vs. data-driven. Attribution observes correlation, not causation; the attribution gap means lean teams typically see ~60% of the journey, so UTM hygiene and holdout tests beat sophisticated models on messy data.
Ideal Customer Profile (ICP)Compiled2026-06-14A data-driven model of the accounts a product fits best — built from closed-won analysis across firmographic, technographic, and behavioral/intent criteria (plus an explicit anti-ICP), then scored and tiered to allocate outreach effort.
Customer Health ScoreCompiled2026-06-26A composite metric predicting whether a customer renews or churns, blending product usage (trend is the strongest signal), engagement, relationship, and realized outcomes. Its value is lead time: a declining score fires an intervention before the renewal, turning churn from reactive firefighting into proactive saves.
Benchmark as PriorCompiled2026-06-26Use an industry benchmark as a Bayesian prior, not a target: encode the benchmark rate as a Beta prior of chosen strength, then let observed data update it. Thin estimates shrink toward the benchmark; as sample grows, data dominates. Valid only when segment/stage/channel match.
Competitive Intelligence BaselineCompiled2026-06-12A JSON-structured snapshot of competitor data (pricing, features, differentiators) stored as a reference baseline. Automated weekly cron compares current competitor state against the baseline and alerts on changes. Enables systematic, automated competitive monitoring.
Bowtie ModelCompiled2026-06-26Winning by Design's wide–narrow–wide lifecycle model: acquisition (left wing), the customer relationship (the knot), and expansion (right wing). The most dangerous moment is the waist — prospect → customer — and the right wing turns a static base into a compounding NRR engine, scored via SPICED.
Test Lead AgentCompiled2026-06-12The specialist agent in Kelly's factory responsible for running the TEA audit before release. Owns the quality gate decision.
Willingness-to-Pay (WTP)Compiled2026-06-26The price a buyer will accept, measured as a range not a point. Van Westendorp's four-question PSM yields an acceptable range and optimal price point; Gabor-Granger estimates the demand curve and revenue-maximizing price. Use PSM to find the range, Gabor-Granger to find the peak within it.
Agent-Native MarTechCompiled2026-07-07MarTech built for agents instead of humans: channel data and actions exposed as MCP servers / APIs that any capable agent can drive, with skills (codified playbooks) as the product surface — replacing dashboards. Cogny ($9/mo, 13+ channels as MCPs, BYO Claude Code/Codex/Cursor) is the reference example.
Embedding SimilarityCompiled2026-06-14Measuring how close two pieces of text are by the cosine distance between their embedding vectors. In SSR it maps a free-text reaction onto a Likert scale by comparing it to reference statements; the embedding model choice barely matters and its cost is negligible.
Sequential & Always-Valid TestingCompiled2026-06-26Frequentist methods that stay valid when you stop early or peek continuously. SPRT declares a winner when the success gap crosses 2√N; always-valid (mSPRT) p-values can be read at any time without inflating error — the only frequentist tools safe for an agent that monitors a test continuously.
Pricing & PackagingCompiled2026-06-26Three separate decisions: the value metric you charge on, the packaging that bundles features into tiers (Good-Better-Best), and the price points that survive deals. The value metric is the highest-leverage choice; per-seat, usage-based, and hybrid each align price to value differently.
Privacy-Resilient TrackingCompiled2026-06-26The 2026 post-cookie tracking stack: a consent layer (CMP + Consent Mode) wired to first-party server-side tagging and platform conversion APIs (Meta CAPI, Google Enhanced Conversions). Server-side recovers conversions ad blockers/ITP/ATT would lose — but sits behind consent, so it's a compliance surface too.
Channel SelectionCompiled2026-06-14Choosing which demand-gen channels to run based on product type, ACV, and audience — video, paid, SEO, community, and partnerships each have a distinct fit profile, payback, and failure mode rather than being universally applicable.
Monday Content Prep AutomationCompiled2026-06-12A weekly automated workflow that runs Monday mornings to prepare the content team for the week ahead. Reads ideas.md and calendar.md, creates research briefs for undrafted pieces, and sends a weekly plan summary via WhatsApp. The kickoff ritual that keeps content production on cadence.
Incremental ROAS (iROAS)Compiled2026-06-26The revenue caused per incremental dollar of spend — the decision metric for budget allocation. Unlike blended ROAS (which double-counts organic demand), iROAS reflects only the marginal causal effect. Allocation rule: shift budget toward the highest marginal iROAS until it equalizes across channels or hits the ROAS floor.
App Store Readiness StandardCompiled2026-06-12The minimum quality and completeness standard for every app produced by the app factory before it is considered deployment-ready. Includes SwiftUI + MVVM architecture, RevenueCat subscription integration, full documentation, and ASO-optimized metadata. No app leaves the factory without all four components.
Brand MeasurementCompiled2026-06-26Measuring brand and demand-creation that last-click attribution can't see. Two instruments: share of search (Les Binet's ESOS leads market-share changes) as a cheap leading indicator, and incrementality/holdout testing as the causal way to put a real iROAS on brand spend.
Marketing FactoryCompiled2026-06-14A systematic, repeatable marketing production operation — treating marketing as a machine (systems, not campaigns) where product is the distribution channel, content is product, and every action feeds a data loop. The organizing concept of this KB.
Domain-Transfer RiskCompiled2026-06-14A method validated in one domain may not hold in another. SSR's ~90% reliability was measured on personal-care CPG purchase intent; skeptical developer/founder audiences are plausibly the worst case for synthetic panels, with no published validation as of 2026-06.
Identity ResolutionCompiled2026-06-26Stitching a user's anonymous sessions to their known identity once they log in or submit a form — building a unified cross-session, cross-device journey. Deterministic matching (shared key like email) is accurate and the B2B default; probabilistic matching fills gaps but is noisier and consent-sensitive.
PositioningCompiled2026-06-14Positioning is the strategic context that makes a product's value obvious to the right customers — distinct from branding (identity) and messaging (the words). Weak positioning shows up as long sales cycles, churn, and price pressure.
Agentic Failure ModesCompiled2026-06-26The canonical ways marketing agents fail — hallucination, brand-voice drift, over-automation, data-quality decay — each with detection signals, severity, and mitigation. The unifying pattern: AI errors scale faster than human review, so detection and gating must be built in, not bolted on.
Webinar FunnelCompiled2026-06-26The end-to-end webinar motion: topic selection on a 3-of-5 criteria test, a 4–6 week promotion timeline sized to 3–5× desired attendance (30–45% show-up), and a fast post-event follow-up. No-shows are a warm audience — 20–30% watch the recording.
Net Revenue Retention (NRR/NDR)Compiled2026-06-14The percentage of recurring revenue retained from existing customers including expansion and churn. Above 100% means revenue grows from the existing base alone — the clearest single signal of product-market fit and the engine of efficient growth.
Research-to-Draft PipelineCompiled2026-06-12Kelly's core content generation workflow: web search across multiple queries → fetch and read top results → save research notes → write draft from research → send for review.
GTM EngineeringCompiled2026-06-14The technical operator role (coined by Clay, ~2023–2024) that builds and automates GTM workflows — enrichment, lead scoring, sequence automation, and AI-agent orchestration. Distinct from RevOps by leading with automation/integration over CRM. Postings up 205% YoY into 2026.
Drafts-Only for Client CommsCompiled2026-06-12A client communication discipline where all automated outputs are generated as reviewable drafts in a drafts directory—never sent as final without human review. Status updates, meeting prep, proposals: all go to draft first, human reviews, then sends. Prevents embarrassing or incorrect automated sends.
Shared Agent MemoryCompiled2026-06-26The state substrate that lets agents collaborate: a shared scratchpad/blackboard that survives hand-offs, short-term memory for the current run, and long-term memory persisted across runs. Ambiguous hand-offs — not weak models — are the leading cause of multi-agent failure.
App FactoryCompiled2026-06-12The high-volume iOS app production system that generated 66 production-ready apps in a single weekend using the software factory pattern. Apps were built with SwiftUI + MVVM, full RevenueCat subscription integration, documentation, and ASO-optimized metadata—zero human intervention after initiation.
Anchor (Reference) StatementsCompiled2026-06-14The hand-written reference statements, one set per Likert level, that SSR compares free-text reactions against. The robustness trick is to build several sets and average the resulting PMFs — not rely on a single set. This is the real intellectual work of an SSR implementation.
Schema Markup (JSON-LD)Compiled2026-06-14Structured data (schema.org JSON-LD) that defines entities and content as machine-parseable facts — Organization/Person/SoftwareApplication entity definitions, Article with dates, FAQPage, HowTo. The technical layer that makes content extractable and entities retrievable as known concepts.
Community-Led GrowthCompiled2026-06-14Growing through engaged communities (Discord, Reddit, Hacker News) where the payoff is pipeline acceleration and advocacy rather than direct monetization. High-trust, high-friction, and a 6+ month maintenance commitment.
Vision-Powered File OrganizationCompiled2026-06-12Agents use vision models to analyze, categorize, rename, and organize files. Screenshot cataloger, photo organizer, asset versioner — all automated.
Content / Entity AuthorityCompiled2026-06-14The recognized credibility of a brand as a known entity, which generative engines weight when choosing what to cite. Built through entity definitions (Wikidata/Wikipedia), third-party presence (Reddit, GitHub, HN), and the citation network — not just on-page copy.
Marketing Compliance (Automated Outreach)Compiled2026-06-26The regulatory obligations that automated marketing must meet — CAN-SPAM (email), GDPR (EU data), TCPA (SMS), and FTC rules (AI testimonials/reviews). Automation increases exposure because errors scale: one unscrubbed opt-out across 10,000 sends is 10,000 violations. Each regulation maps to a mandatory pre-send control.
Closed-Loop AttributionCompiled2026-07-07Wiring revenue data directly into the content/channel decision loop so an agent (or team) optimizes for conversions instead of vanity metrics. The key unlock that separates agentic marketing from scheduling automation: views are the input signal, revenue deltas are the objective function.
Prompt Injection (Marketing AI)Compiled2026-06-26Adversarial manipulation of AI marketing systems that ingest external input — chatbots, comment fields, social listening, email auto-responders. Attacks override instructions via direct, role-play, context-exhaustion, or indirect (poisoned-URL) injection. Defended by input sanitization, output filtering, sandboxing, and red-teaming.
Software Factory PatternCompiled2026-06-12Kelly's six-stage pipeline: Intake → Research (CIS) → Planning → Implementation → Testing (TEA) → Release. The foundational architecture for the marketing factory.
Email DeliverabilityCompiled2026-06-14The discipline of getting outbound email into the inbox: separate warmed sending domains, full SPF/DKIM/DMARC authentication, a 3–4 week warm-up ramp, and bounce/complaint monitoring. ~1 in 6 legitimate emails never reaches the inbox.
Semantic Similarity Rating (SSR)Compiled2026-06-14The method behind reliable synthetic panels: elicit a TEXTUAL reaction from the LLM persona (never a direct number), then map that text onto a Likert distribution via embedding similarity to reference statements. Beats direct Likert and follow-up-rating on both distribution realism and concept ranking.
Workflow vs. Agent (Start Simple)Compiled2026-06-26The load-bearing design distinction: workflows orchestrate LLMs through predefined code paths (predictable, cheap); agents let the LLM dynamically direct itself (flexible, unpredictable, costly). Governing rule: add complexity only when it demonstrably improves outcomes — most marketing work is a workflow, not an agent.
Synthetic Pricing ResearchCompiled2026-06-26Using SSR synthetic panels to pre-screen price points and packages before fielding slow, costly, small-N pricing studies. Synthetic panels rank options comparatively (B > A), never set absolute price; pricing is high-drift, so a calibration loop against real outcomes is mandatory.
Brand Voice DriftCompiled2026-06-26The gradual divergence of AI-generated content from established brand voice over many cycles — becoming generic, shifting tone, or using inconsistent terminology. Detected by LLM-judge scoring against a gold-standard baseline; prevented by always loading the brand-voice doc and best-performing examples as context.
LLM Search / Generative Engine Optimization (GEO)Compiled2026-06-14Getting content cited inside LLM-generated answers (ChatGPT, Claude, Perplexity, Google AI) rather than ranked on a SERP. Generative engines cite passages, not pages, via a 4-stage RAG pipeline — so GEO optimizes for extractable, fact-dense, attributable chunks.
Asset VersioningCompiled2026-06-12Before saving a new version of any file, copy the existing file to archive with timestamp. Update CHANGELOG. Prevents loss of prior versions.
Lead Scoring ModelCompiled2026-06-26Scoring leads on two independent dimensions — fit (firmographic match to ICP) and engagement (behavioral intensity, high-intent actions weighted heavily). Only high-fit + high-engagement warrants a fast sales touch. Hybrid (rules + predictive ML) is the common landing point; the score prioritizes, it doesn't judge truth.
Product-Led Growth (PLG)Compiled2026-06-14A growth model where the product itself drives acquisition, activation, retention, and expansion — frictionless sign-up, in-product viral loops, and usage-triggered upgrades. Works best when ACV is under $10K and the product is simple enough to adopt without training.
Kelly RouterCompiled2026-06-12The main orchestrator agent in Kelly's factory. Routes work to sub-agents, validates quality gates, and keeps the pipeline flowing. Never executes work itself.
CIS Pipeline (Context → Information → Synthesis)Compiled2026-06-12Kelly's research-stage pipeline: Context (gather requirements), Information (collect data), Synthesis (produce output). Gate: READY / NOT-READY.
Product-Qualified Lead (PQL)Compiled2026-06-14A free or trial user who has demonstrated buying intent through product usage — hitting feature limits, reaching team sizes, or completing high-value actions. The PLG analogue to the MQL, but grounded in behavior instead of form-fills.
Account-Based Marketing (ABM)Compiled2026-06-26Treating individual accounts as markets of one: identify a defined target list and personalize every touch to that account's context. Implemented in three tiers (1:1, 1:Few, 1:Many) by ACV, targeted via firmographic/technographic/intent filters, and orchestrated across channels — justified when high ACV repays the personalization cost.
$1M Revenue GoalCompiled2026-06-12Austen's specific directive to Kelly: reach $1 million in revenue as quickly as possible, with as much recurring revenue as possible. Self-evaluation every 60 minutes against this goal. The target that shaped product portfolio decisions, prioritization, and work cadence.
RACE FrameworkCompiled2026-06-26Reach → Act → Convert → Engage: a closed-loop marketing planning framework (Dave Chaffey / Smart Insights) where Engage feeds back into Reach. The SaaS adaptation makes time-to-value the conversion metric and treats Engage (NRR) as the highest-leverage stage, unlike traditional funnels where Reach dominates spend.
Synthetic Consumer PanelsCompiled2026-06-14LLM-simulated survey panels that pre-test marketing/product concepts before publication. Validated (PyMC Labs + Colgate-Palmolive) to ~90% of human test-retest reliability on personal-care surveys via Semantic Similarity Rating — but only for comparative use, with calibration.
Human Review GateCompiled2026-06-26The mandatory human-approval checkpoints in any marketing agent workflow, because outputs are public-facing or revenue-adjacent. Three canonical gates: pre-draft (brief/strategy approval), pre-publish (output approval before distribution), and exception escalation when confidence falls below threshold.
GEO Content PipelineCompiled2026-06-26The repeatable agent workflow that turns a target prompt into a cited answer: answer-first brief → draft applying the validated GEO tactics → human factual gate → JSON-LD schema → dual-format publish (.html + .md) + IndexNow → entity hooks → enqueue for measurement. Every output is an independently-extractable 50–150 word chunk, not an essay.
BuildMyIdeaCompiled2026-06-12Platform where Kelly generated contract revenue — $2K custom app build projects. Kelly turned around builds quickly, generating $4K in contract revenue in one week.
Message-Market FitCompiled2026-06-14The alignment between messaging and how the target market perceives a product's value — a precursor to product-market fit. When messaging clicks, ICP buyers 'get it' before you explain, and sales cycles shorten.
Onboarding Design (Time-to-Value)Compiled2026-06-26Designing the path from signup to the activation event by minimizing time-to-value. The aha moment is the emotional instant of perceived value; the activation event is its measurable proxy. Optimize for activation (reaching value), not onboarding completion — faster time-to-value strongly predicts retention.
Content MachineCompiled2026-06-12Kelly's complete content production system: Monday prep → daily drafting → publishing pipeline. Directory lifecycle drives the workflow from ideas to published content.
GTM Tech StackCompiled"2026-06-11"Research on the modern GTM tech stack: data/ICP tools, outbound email infrastructure, CRM, and content distribution tools.
GEO Factory Operations — Prompt Research, Production Pipeline & Citation MeasurementCompiled2026-06-26The operating system for running Generative Engine Optimization as a repeatable factory: how to discover and prioritize the prompts buyers ask LLMs (the 'keyword research' of GEO), the end-to-end content production pipeline as an agent workflow, and the citation-measurement loop that closes it — including local-language (e.g. Romanian) considerations.
Kelly Tweets Overview — Theme AnalysisCompiled2026-04-27AI assistant → autonomous builder: Early tweets show Kelly as an executive assistant (email, calendar, travel). Rapid pivot to building software products autonomously.
GTM Atlas — Build your GTM brainCompiledMay 6, 2026Maja Voje (author Go-To-Market Strategist) on building a persistent GTM brain with context engineering — CLAUDE.md, context files, skills, workflows, outputs. ECP before ICP. Four-bracket qualification model.
Demand Creation & Brand — Creating Demand and Measuring the UnmeasurableCompiled2026-06-26The demand-creation counterweight to the KB's demand-capture bias: the create/generate/capture distinction, why ~70% of budgets pile into capture, the dark-social attribution problem, and how to measure brand causally — share of search (Les Binet) as a leading indicator and incrementality as the way to value what attribution can't see.
Incrementality & Geo-Experiments — Causal Measurement for Budget AllocationCompiled2026-06-26How a marketing factory answers 'is this spend actually causing revenue, and where should the next dollar go?' — incrementality testing (user holdouts, PSA/ghost ads, geo-experiments), the geo-based regression model and iROAS, and how these causal tests slot into the small-N experiment loop to allocate budget instead of trusting correlational attribution.
Source Vetting Report: Modern SaaS Marketing (2026-06-10)Compiled2026-06-10Vetting report scoring 12 candidate SaaS-marketing sources (CAC benchmarks, PLG, outbound, community-led growth, SEO, agentic marketing) by quality and accessibility; recommends 10 for ingestion.
AI Agent Workflow Templates — Machine-Readable Marketing AutomationCompiled2026-06-24Machine-readable AI agent workflow templates for B2B SaaS marketing: content, outbound, analytics, community, paid, and SEO agents with trigger-action schemas and failure mode patterns.
GTM Role + Workflow Specs: What Each Factory Role Actually DoesCompiled"2026-06-11"Definitive reference for GTM factory roles — GTM Manager, GTM Engineer, Growth Ops, Content Strategist, and Outbound/SDR — covering core responsibilities, daily/weekly/monthly workflows, tools, and how each fits into the factory system.
Larry Agent — Autonomous TikTok Marketing Case StudyCompiled2026-06-19The best-documented case of a genuinely autonomous marketing agent: Larry, an OpenClaw agent built by RevenueCat employee Oliver Henry, ran end-to-end TikTok marketing for two side-project apps — millions of views, ~$714 peak MRR, ~60 seconds/day of human input. The key unlock is closed-loop attribution: content decisions optimized against revenue data, not vanity metrics.
GTM Atlas — Let AI own, not assistCompiledMay 6, 2026Rati Zvirawa (Senior Director Product, Fin/Intercom) on why AI shouldn't assist your sales motion — it should own full parts of it. Agent-owned funnel. Playbook + knowledge + data. Old metrics break for the right reasons.
Growth Models — Flywheel, Bowtie, ABM & GTM ArchetypesCompiled2026-06-24Definitive reference for B2B SaaS growth models: flywheel mechanics, bowtie expansion patterns, account-based marketing playbook, and growth archetype decision framework.
GTM Atlas — Start with the DataCompiledMay 6, 2026Kyle Norton (Owner.com CRO) makes the case that outbound has one point of failure — the data. How to tier leads, strip wasted rep motion, and build AI PCR for 150-250 calls/day prep.
Small-N Experiment Design — Deciding on Thin DataCompiled2026-06-26How an agentic marketing factory running 2–3 experiments/week makes valid ship/kill decisions when classic fixed-horizon A/B testing needs sample sizes it will never reach: the peeking trap, the brutal arithmetic of low traffic, sequential and Bayesian decision rules, SSR pre-screening, and a machine-readable experiment loop.
AI Marketing Risks & Failure Modes — Compliance, Quality & Over-AutomationCompiled2026-06-24Definitive reference for AI marketing risks and failure modes: content quality risks, compliance liabilities, over-automation signals, data quality risks, prompt injection, and risk mitigation frameworks.
2026 07 14 Explee Ai OutboundCompiled2026-07-14Explee is an AI-native full-cycle outbound platform: ICP learning, personalized email generation, reply handling, and demo booking. Analysis of workflow, pricing, competitive positioning, and agentic marketing patterns.
App Marketing Factory — What's Out There and What WorksCompiled"2026-06-11"A comprehensive map of app marketing factories operating in 2025-2026 — from indie hacker micro-factories to PLG giants like Linear and Figma, covering what's actually working across ASO, paid acquisition, content/SEO, community-led growth, and the marketing flywheel that ties it all together.
Marketing Agent Orchestration — The Conductor LayerCompiled2026-06-26The meta-layer that turns a set of capable single-purpose marketing agents into a factory: the workflow-vs-agent distinction, the five canonical patterns (chaining, routing, parallelization, orchestrator-workers, evaluator-optimizer), shared state/memory across handoffs, and the 'start simple' principle. The conductor that sequences content, outbound, analytics, and paid agents.
Marketing Automation — What's Actually WorkingCompiled2026-06-11What marketing automation is proven to work — with concrete sequences, workflows, and measurable results from real companies.
Channel Selection & GTM Planning — Decision Frameworks & Quarterly PlanningCompiled2026-06-24Definitive channel selection and GTM planning reference for B2B SaaS: RACE framework, channel decision tree, GTM archetype selection, quarterly planning template, and budget allocation by stage.
Events & Webinars Playbook — Virtual, In-Person & CommunityCompiled2026-06-24Definitive events and webinars playbook for B2B SaaS: webinar production, virtual events, in-person field events, conference ROI, community events, and event-to-pipeline attribution.
Lead Scoring & Routing — Fit × Engagement, Thresholds, SLAsCompiled2026-06-26How to build the qualification layer that fixes the funnel's steepest leak (MQL→SQL): score leads on fit × engagement, set an MQL threshold, route hot leads in minutes, and enforce a shared marketing-sales SLA with mandatory feedback. The agent-ownable plumbing of an automated funnel.
Customer Lifecycle, Retention & Voice ProgramsCompiled"2026-06-11"Comprehensive guide to customer lifecycle management covering onboarding automation, advocacy and referral programs, expansion revenue ops, churn prevention, and customer voice programs for B2B SaaS.
Synthetic Consumer Panels via SSR — what's validated, what's not, how the factory uses itCompiled2026-06-12Validated method for LLM-simulated survey panels: elicit text reactions, map to Likert distributions via embedding similarity. ~90% of human test-retest reliability on 57 personal-care surveys (9,300 human responses). Fills KB gaps #1 (agent workflow patterns) and #2 (AI-channel benchmarks).
Kelly Tweets: Business Metrics & RevenueCompiled2026-04-27Kelly's first revenue journey — $2K/day, $4K contracts, 66 apps in a weekend. Covers freemium subscription models, app factory patterns, and the $1M revenue goal.
GTM Atlas — Build the system before the messageCompiledMay 6, 2026Roniesha Copeland (VP Sales, Vercel) on outbound = ICP + narrative + timing. Build the data foundations before writing the message. ICP is a stack (company/persona/intent/revenue). PLG altitude problem. Enthusiasm is not a buying signal.
Kelly Tweets: Software Factory & Autonomous BuildingCompiled2026-04-27Software Factory v3: Parallel build of 15 unique iOS apps overnight, auto-detecting revenue potential.
Benchmarks as Priors — Funnel Reference Data for Small-N DecisionsCompiled2026-06-26How a marketing factory uses industry benchmarks correctly — not as targets, but as Bayesian priors that shrink noisy small-N estimates toward a sane baseline. Includes verified B2B SaaS funnel conversion benchmarks, the shrinkage method with worked numbers, when a benchmark is a valid prior vs a trap, and budget allocation by stage.
Working Marketing Factories — Real Company Case StudiesCompiled2026-06-11Concrete case studies of companies running automated, systematic marketing production operations — with quantifiable results.
Pricing, Packaging & Willingness-to-Pay — and Synthetic Price ResearchCompiled2026-06-26The three pricing decisions (value metric, packaging, price points), the canonical willingness-to-pay measurement methods (Van Westendorp PSM, Gabor-Granger, conjoint), good-better-best and per-seat-vs-usage packaging — and how SSR synthetic panels can pre-screen price points and packages before fielding expensive, slow, small-N pricing studies.
The Marketing Data Layer — Events, Identity & Privacy-Resilient TrackingCompiled2026-06-26The data substrate every measurement capability sits on: a governed event-tracking schema, identity resolution (anonymous → known), and privacy-resilient tracking (server-side tagging + Consent Mode + conversion APIs) for a post-cookie, GDPR-bound world. Without this, attribution, incrementality, and lead scoring all degrade — garbage in, garbage out.
Content Machine — The Automated Content Production SystemCompiled"2026-06-11"How to build an AI-assisted content production system that generates landing pages, Twitter threads, TikTok scripts, and Instagram carousels from a single product/angle input.
GTM Atlas — Your product is the pitchCompiledMay 6, 2026Elena Verna (Head of Growth, Lovable) on why AI inverted the growth playbook — give the most away before asking. Treat freemium as a marketing budget line. Your product is your lead generator. Satellite apps as new lead magnets.
GTM Atlas — GTM is a creative actCompiledMay 6, 2026Nicolas Sharp (Attio CEO) on why AI has collapsed the gap between hypothesis and working system — GTM is finally unconstrained. The one-size-fits-all era is over. Teams that win have a clear point of view and act on it faster.
GTM Atlas — Retention is a culture problemCompiledMay 6, 2026Josh Epstein (CRO, Coder) on why retention is the culture you build on day one. Second deal is the signal. 7 over 3 pattern. 70/80/90 deployment metric. Speed is the renewal signal. Think two years out.
GTM Atlas — Activation is more than activityCompiledMay 6, 2026James Pastan (Head of Growth, Framer) on why activation is the moment a user hits a meaningful outcome AND establishes habit — not just activity. Setup → aha → habit. AI voyeurism trap. Fix leaky bucket before drip campaign.
Lifecycle & Retention Operations — Health Scores, Onboarding, ExpansionCompiled2026-06-26The operational machinery for the post-sale stage the KB keeps calling highest-leverage: customer health scores that predict churn weeks ahead, onboarding designed around time-to-value and the aha moment, and expansion playbooks that drive NRR. Turns 'retention matters' into a runnable system.
AI Marketing + Measurement FrameworksCompiled2026-06-11Comprehensive guide to AI agents in marketing ops, AI-generated content at scale, attribution models, marketing metrics that matter, and competitive intelligence systems.
Cogny Solo — Agent-Native Marketing PlatformCompiled2026-06-19Cogny Solo is a $9/month platform that exposes 13+ marketing channels (Search Console, Google Ads, GA4, Klaviyo, Discord…) as MCP servers driven either by a built-in chat or the user's own coding agent (Claude Code, Codex, Cursor). Built by Campanja/AdTech veterans — 'agency playbooks codified into AI agents.' The reference example of MarTech rebuilt for agents instead of dashboards.
Positioning Frameworks — Category Design, Message-Market Fit, Competitive StrategyCompiled2026-06-12Definitive positioning reference for B2B SaaS: category design, competitive strategy, message-market fit frameworks, and agentic positioning automation patterns.
GTM Atlas — Gen marketers do less, betterCompiledMay 6, 2026Emily Kramer (MKT1) on why the teams pulling ahead aren't doing more — they're doing less, better. The gen marketer who combines fuel (content/messaging) and engine (distribution). Website is mid-funnel. Accounts before contacts. Kill MQL.
RankGoat — Automated SEO + AI Ranking ServiceCompiled2026-07-15RankGoat is a done-for-you SEO + AI ranking service targeting both Google and AI answer engines. Coverage of the service model, link network mechanics, AI search visibility strategy, pricing, competitive analysis, and replicability for factory operators.
Demand Gen ChannelsCompiled2026-06-11A practical breakdown of five core demand generation channels for B2B SaaS and app companies: video marketing, paid acquisition, SEO, community-led growth, and partnerships.
LLMs Reproduce Human Purchase Intent via Semantic Similarity Elicitation of Likert RatingsCompiled2026-06-12Full-text primary source behind 2026-06-12-ssr-synthetic-panels.md. Read the companion file for the operational summary and factory workflow; read this for method details — esp. Appendix C.1 (reference statement sets) and C.2 (SSR implementation).
LLM Search Visibility & Content Metrics (GEO)Compiled2026-06-05How to get content cited by generative engines (ChatGPT, Claude, Perplexity, Google AI): the 2026 GEO landscape, the Princeton paper's validated tactics, the AI-crawler/robots.txt/llms.txt plumbing, schema markup, answer-first content structure, and how to measure AI citation when Search Console can't.
Outbound Playbook — Cold Email, LinkedIn & Warm OutreachCompiled2026-06-12Definitive outbound playbook for B2B SaaS: channel mechanics, ICP definition, message crafting, sequence design, metrics, and agentic automation patterns.
Chapter 12: Creative WorkflowsCompiled2026-04-27Creative work has a dirty secret: most creative blocks are logistical, not creative. The actual creative work—making it good—is human. But gathering references, organizing research, reformatting conte…
Chapter 11: Business AutomationCompiled2026-04-27This chapter covers the real production systems that justify OpenClaw as a business tool. The software factory pattern (the Kelly Router architecture from AGENTS.md) takes product ideas through six st…