Revenue is a system,
not a headcount.

Chapter 01 · Part I · Day 1

The Revenue Engine: names and owners

1.1 The naming problem

Three terms get used as if they meant the same thing. They do not, and the confusion erases accountability.

Rule: the umbrella is the Revenue Engine. Strategy has exactly one owner, the CRO. RevOps decides how the process and data work. Revenue Systems Engineering decides how to build the systems that run the process. Everyone can be involved in strategy; exactly one seat owns it. If two roles both "own strategy", nobody does.

1.2 The ownership ladder

The engineering-org analogy is the fastest way to internalize the altitudes: a CEO sets company strategy, a CTO sets strategy within engineering, a Head of Engineering makes tactical calls, a senior engineer decides how to implement. None of them re-owns the layer above.

1.3 The four layers

Layer 1: the CRO (strategy and command). The CRO does not do operations. The CRO owns the number and commands the engine:

• Owns the number. Marketing owns leads, CS owns retention, finance owns headcount; the CRO owns revenue end to end.
• Sets the go-to-market motion: product-led, sales-led or hybrid; which markets; how to position; which channels.
• Owns pricing and packaging strategy in service of the revenue goal.
• Aligns the functions into one revenue motion instead of three departments optimizing locally.

The CRO is the general who owns the objective. RevOps is the logistics and intelligence staff who make the plan executable. The general does not personally run logistics.

Layer 2: functional VPs (command the doers).

Layer 3: RevOps (the operational backbone). RevOps is centralized and reports to the CRO. It is not one operations lead per VP. The siloed model (Sales Ops under Sales, Marketing Ops under Marketing, CS Ops under CS) is exactly what RevOps was invented to remove, because each silo ran its own data model and processes broke at every handoff. The result is a matrix: VPs own the people who execute; RevOps owns the rails they all run on. Sales Ops, Marketing Ops and CS Ops sit under the Head of RevOps, typically organized as operations, systems, enablement and insights.

RevOps owns: CRM architecture · data governance and hygiene · pipeline reporting and forecasting · funnel-stage design · tooling standards and procurement · territory and quota design · commission and comp plans · deal-desk process · enablement infrastructure · orchestration of the build layer.

Layer 4: Revenue Systems Engineering (the build layer). The technical function that builds the automated systems executing the lifecycle. "Go-to-market engineering" is the label that stuck, but it sounds like a one-time launch; the function is permanent. "Customer acquisition engineering" would be too narrow, because the job also automates expansion and retention. The precise name is Revenue Systems Engineering: building automated systems across acquire, expand and retain. This framework uses "GTM Engineering" when speaking the market's language and "Revenue Systems Engineering" when precision matters. Chapter 6 covers it in full.

1.4 Role glossary

The sharpest distinction: the SE helps close one customer, in front of the customer. Sales Ops builds the machine all AEs use, behind the scenes. The Revenue Systems Engineer automates the whole pipeline rather than serving any single live deal.

1.5 Three models to lay over the engine

They model different things and are complementary.

• The Bowtie (Winning by Design, Jacco van der Kooij, Revenue Architecture, 2023). Extends the funnel past the close: left side Awareness → Education → Selection; the knot is the mutual commitment; right side Onboarding → Adoption → Expansion. In recurring-revenue businesses most value is created right of the close, which is why customer success is a revenue function, not a cost centre.
• The Revenue Waterfall (Forrester, formerly SiriusDecisions; 2006, revised 2012, 2017, 2021). The canonical model of pipeline stages and handoffs: Inquiry → MQL → SAL → SQL → Close, each with entry and exit criteria and a single owner. The 2017 version models buying groups; the 2021 version makes renewal, cross-sell and upsell first-class opportunity types. RevOps uses it to instrument the funnel.
• The nine building blocks (Diorio and Hummel, Revenue Operations, Wiley 2022). The most research-backed description of what a mature RevOps function should contain, from more than 1,000 surveys and interviews.

Chapter 02 · Part I · Day 1

Throughput and the constraint

Translated from Eliyahu Goldratt's The Goal (1984) and the Theory of Constraints into the revenue engine.

Every revenue engine has exactly one constraint at a time: the stage or resource whose capacity is less than or equal to the demand placed on it. The engine's revenue rate is set by that constraint, not by how hard everything else works. The highest-leverage job in revenue operations is to find the constraint with data, exploit it without spending, subordinate everything else to its pace, elevate it only when the free moves are exhausted, and re-run the hunt when it moves.

2.1 The goal is money: throughput, not activity

Example. An SDR team at 140% of activity quota while AEs carry three weeks of demo backlog is not productive. It is producing inventory the system cannot sell. In the book, the plant's new robots raised measured efficiency on every non-bottleneck machine while the plant kept losing money.

Rule: before fixing anything, write the engine's goal as one throughput number, and classify every metric you report as throughput, inventory or expense. Anything else is a local-efficiency mirage.

2.2 The five focusing steps

Never skip steps. Most bottleneck failures are teams jumping to step 4 (hire, buy a tool, do more of the loud thing) before steps 2 and 3 are exhausted, or never re-running step 1 after a fix.

2.3 Identify: the arithmetic pass

Two laws: an hour lost at the bottleneck is an hour lost for the entire system; an hour saved at a non-bottleneck is a mirage.

How to hunt:

1. Model the engine as stages, each with an owner, an entry criterion and an exit criterion (Inquiry → MQL → SAL → SQL → Close → Onboarding → Adoption → Expansion).
2. For each stage write demand (inflow per week), capacity (realistic weekly output) and WIP (open items).
3. The constraint is where capacity ≤ demand, or WIP grows and ages, or cycle time is longest and lengthening, or conversion collapses as items wait.
4. Cross-check with the Herbie test: in the book's scout hike the troop moves at the speed of its slowest walker. Whose pace does everyone work around? Whose backlog does everyone complain about?

If you cannot write capacity versus demand per stage from your CRM, the first fix is data quality. You cannot find the constraint in a system you cannot measure.

2.4 Exploit (no new spend)

1. Quality-gate before the constraint. Re-qualify everything entering the constraint stage. Constraint time is the engine's most expensive time; never spend it on garbage.
2. Strip admin from the constraint role. CRM entry, notes, scheduling, internal meetings: automate or offload. This is the right use of automation.
3. Protect the constraint's calendar. Meetings schedule around the constraint, never the reverse.
4. Offload to slack resources. Non-bottlenecks have spare capacity by definition. Move parts of the demo, discovery or technical deep-dive to them.
5. Do not build ahead. No pre-building pipeline the downstream cannot absorb.

2.5 Subordinate

• Marketing and SDRs release work at the constraint's consumption rate, not their own maximum. Ten thousand MQLs into an SDR team that can work four hundred is inventory, not revenue.
• Comp pays on constraint outcomes: meetings that become qualified demos, or closed revenue; not meetings booked.
• The constraint's output is the downstream's urgent input. Closed-won moves to onboarding the same week.
• The test for any non-constraint activity: does this feed the constraint what it needs, when it needs it?

2.6 Elevate (only now, with spend)

Hire into the constraint role · automate the constraint's own work (routing, scheduling, triage, proposal assembly, kickoff prep) · parallelize by segment · add a self-serve path for the long tail · cut per-unit friction at the constraint (pricing complexity, contract redlines, procurement). Most "we need more AEs" requests are really unexploited constraint time or unsubordinated upstream volume.

2.7 Repeat, and beware inertia

The constraint moves downstream with every fix. The metrics that told the truth last quarter become the metrics that hide the new constraint. Schedule the re-identification pass the same week you declare a constraint fixed. If a metric has sat unchanged on the dashboard for two quarters, ask what it is hiding.

2.8 The mirage of local efficiency: the robots and the AI-SDR

The plant's robots produced inventory the constraint could not consume. The current version of this trap is automation:

• An AI-SDR that doubles outbound into an AE path that cannot absorb it creates WIP faster, not revenue faster.
• Marketing automation that generates more MQLs than the constraint can process builds a bigger, staler queue.
• A tool that "saves hours" for a non-constraint team saves nothing.

Buy automation for the constraint. Let non-constraints stay slack by design.

2.9 Drum-buffer-rope and smaller batches

• Drum: the constraint sets the beat.
• Buffer: a queue of qualified, ready items in front of it, sized to roughly one to two weeks of constraint capacity: enough to absorb variance, never an infinite backlog.
• Rope: upstream release is tied to the drum.

A buffer too small starves the constraint; a buffer too big decays (leads go cold). Smaller batches cut lead time: continuous outreach batches instead of ten-thousand-contact blasts; daily handoffs instead of weekly dumps.

2.10 The throughput dashboard

Chapter 03 · Part II · Day 1

Choose the entry motion: Lighthouse or Landgrab

From Joe Schmidt IV and Julian Marx, "Lighthouse or Landgrab? How to Pick Your AI Sales Strategy", a16z, July 2026.

There are two coherent ways to enter a market: win a handful of marquee accounts whose adoption de-risks everyone behind them (Lighthouse), or win on arithmetic and sign the largest number of unglamorous accounts (Landgrab). The choice is a property of the market, not of ambition. Picking the flattering one costs a year. As the source puts it: your buyer doesn't purchase the future; they purchase either proof or math.

Run this test before building any pipeline machinery. A Landgrab market with a Lighthouse motion bleeds cycle time; a Lighthouse market with a Landgrab motion bleeds credibility.

3.1 The two playbooks

Neither is a phase you graduate from.

3.2 The selection test: two questions

Q1. How exposed is the individual signing if this goes wrong? Exposure is high when the domain is regulated, the product replaces a system of record, the output is external-facing (filed, sent, published), or the mistake is not recoverable in the normal workflow. High exposure means ROI math does not close; the buyer is solving for defensibility, and only precedent supplies it.

Q2. Does proof travel in this market? Proof travels when there is a legible status hierarchy, buyers share peer groups, conferences, league tables or a rotating talent pool, and "who else uses this?" is asked unprompted. Answer Q2 with evidence: ask won accounts whose adoption they had heard about before signing. If nobody can name a peer, proof does not travel.

3.3 The decision matrix

The off-diagonals matter most. "Hard market" is usually misfiled as "we need better case studies". "Bottom-up" is often misread as Lighthouse because status-dense markets tempt logo-chasing.

3.4 The field test

The buyer's first unprompted question is the cheapest signal you have.

Mixed signals usually mean the segment is drawn too wide. Split it.

3.5 Sequencing: Lighthouse into Landgrab

1. Win a bellwether in one vertical.
2. Saturate that vertical while precedent is still doing work.
3. Step to the adjacent vertical with the same buyer shape, not the largest one.
4. Standardize as you go, or step 3 compounds delivery cost instead of revenue.

The switch signal: buyers arrive with budget allocated and ask "show me a demo" instead of "who went first?"

3.6 What the choice commits you to

Traps. Lighthouse: hostage to the logo · prestige without economics · pilot purgatory (treat a pilot as unsold until it converts) · single-customer capture · precedent that does not transmit. Landgrab: dying of indigestion (no qualification floor) · scaling sales ahead of product · mistaking one dense seam for the whole market · running a velocity motion at a buyer whose real question was "is this safe".

3.7 Examples

Figures are the source authors' claims.

Chapter 04 · Part II · Day 1

Choose channels on evidence

From an evidence audit of Daniel Priestley's channel tier list, expanded to fifteen channels with 2025–2026 studies.

Channels are not ranked; they do different jobs. The evidence supports matching each channel to a job: discovery, demand capture, education, qualification, or repeat contact. It does not support a universal S-to-D league table. Different jobs cannot substitute for each other: an excellent newsletter cannot reach a list that does not exist, and a quiz needs incoming traffic.

4.1 The fifteen channels by job

Ratings are analytical judgments for a small business selling expertise or a considered B2B purchase, over 6–12 months. A = strong candidate to test; B = useful when prerequisites fit; S = reserved for an established capability with demonstrated economics.

Read the limits. Most of these are vendor-observed samples, not randomized tests. Do not divide one channel's reply rate by another's and announce a lift; audiences and denominators differ.

4.2 What to do with it

• Considered B2B: test one route to identifiable buyers (targeted outreach or professional content), one substantial explanation of the offer (long-form demo), and one way for interested people to return (opt-in list). Add a quiz if its answers improve qualification.
• Visually demonstrable consumer offer: Instagram, TikTok, Shorts and paid social move up.
• Urgent local service: search demand matters more than content reach.
• Do not launch everything at once. Choose a bounded comparison suited to the constraint (Ch. 2). One avatar, one offer, one channel until it works.

4.3 Normalize at the customer, not the click

Fully loaded acquisition cost = media + production + prospect/list research
                                + tools + allocated acquisition and sales labour
CAC = fully loaded acquisition cost ÷ new paying customers
Contribution after acquisition = collected revenue − refunds − variable fulfilment − acquisition cost

Example (illustrative arithmetic, not a benchmark). A $27.66 lead that closes at 2% costs $1,383 per customer in media alone. A $66.69 lead that closes at 10% costs $666.90. The cheaper lead produced the more expensive customer.

Example (funnel math). A quiz's visitor-to-lead rate is visitor-to-start × start-to-lead. With a hypothetical 25% start rate and the 40.1% starter benchmark, about 10% of visitors become leads, before any qualification.

Chapter 05 · Part II · Day 1

Design the offer and the economic case

5.1 Offer architecture

Before a conversion system can route anyone, the business must define the commercial choices it is allowed to route them through. Most operators skip this and send more traffic into a choice environment where every path feels equally optional.

This is not dark-pattern pricing. If a sales call is the right services path, attach it to a specific diagnostic or teardown, not a generic "contact us".

Cold-start variant. When entering from zero, give away value that is cheap to reproduce, give it a native path to the next qualified person, and put the paid boundary where capacity is scarce: operated infrastructure, implementation, customization, accountability, access or judgment. The free layer must be measurable against qualified demand, and paid conversion must trace back to the free asset.

5.2 The value justification

The goal is not to describe what the solution does. It is to show, in economic terms, where value leaks today, why current tools or habits do not fix it, how the solution changes the decision, and what the return is under conservative assumptions. Four links, always in this order:

the leak (or missed opportunity) → the decision gap causing it → the mechanism that fixes it → the value created

Principles

1. Lead with economics, not features. Features matter only after the leak is clear.
2. Separate control from optimization. Existing systems may enforce policy and report activity; that does not mean they produce the best outcome.
3. Use research directionally. Benchmarks show plausibility, not the result in this case.
4. Model scenarios, not a point. Conservative and aggressive ranges are more credible than one precise number.
5. Explain the process, not just the promise.
6. Show why the format matters. If value depends on context and frequency, say why a static rule, spreadsheet or one-off review is not enough.
7. Calibrate to the buyer's pull. Established-position buyers respond to vivid downside, loss of control and audit exposure. Constrained-but-ambitious buyers respond to credible upside, speed and visible progress.
8. Premium perception is a trust bridge, not the value. Buyers judge competence before they finish reading the economics. Every visible surface (screenshots, reports, onboarding, proof) should look capable of the buyer's standard of work. Use it to make real value easier to believe, never to make weak value feel expensive.

The document, section by section

1. Title and target. [Initiative] for [Audience]. Target: generate [X] in value within [period].

2. Baseline assumptions. Volume, revenue, cost base, known leak, known opportunity, data caveats and dates. Use observed data where possible; label estimates.

3. The leak. Name it in operational language: unnecessary discounts, avoidable churn, dropped follow-up, underpriced services, low conversion of existing demand.

4. Why the current approach does not solve it. Current systems answer the compliance or visibility question; they do not answer the optimization question.

   Current approach tells you	Proposed solution tells you
   "This discount is within policy."	"This discount was likely unnecessary given this customer's history."
   "Here is a report of discounts given."	"Here are the deals where margin was lost with no likely volume gain."

5. Directional evidence. Then the required sentence: the research supports the direction; the model below is an internal scenario estimate on conservative assumptions.

6. Scenario model.

   Example (reference case: discount optimization at a multi-location distributor).

   	Conservative	Aggressive
   Annual discounts given	$1,200,000	$1,500,000
   Unnecessary share	15%	40%
   Recoverable margin	$180,000	$600,000

7. Why it needs a dynamic system. The best decision depends on customer history, product mix, timing, channel and seller behaviour. A fixed rule cannot capture that.

8. How it works. Data intake → analysis and scoring → recommendation delivered in the workflow → feedback and learning.

9. Why it is credible. Established methods; the work is integration and validation, not research.

10. Guardrails. Hard constraints (cost floors, contracts, law) · statistical guards (small samples, variance) · business-logic checks · anomaly detection. Restraint builds trust.

11. Combined value and the first lever. Roll up opportunities; name the fastest, cleanest lever first (pure margin recovery, low disruption, easy measurement).

The standard. The reader should conclude: the leak is real, the current method is insufficient, the solution fits the problem structurally, the assumptions are conservative, the upside is material, and the risk is controlled.

Chapter 06 · Part III · Day 2

Revenue Systems Engineering (GTM Engineering)

From Clay's Head of GTM Engineering, Everett Berry, on Stories of Scale*, with market and role research.*

Go-to-market is no longer a headcount problem solved by hiring more SDRs and AEs. It is a systems problem solved by engineers who build, instrument and operate the revenue engine itself. At its best, the same engineering profile sits on both sides of the table: inside the company generating pipeline, and forward-deployed inside customer accounts proving the system works.

Market context. LinkedIn postings for "GTM Engineer" went from ~1,400 in mid-2025 to 3,000+ in January 2026 (ZoomInfo Pipeline). Compensation bands now overlap senior software-engineer bands at well-funded SaaS companies (Apollo), a sign the role is budgeted from engineering, not sales.

6.1 What it is, and what it is not

"RevOps is the factory that builds and maintains your revenue system; GTM engineering is one of the specialized machines inside that factory." (Revenue Operations Alliance)

6.2 The core job in one analogy

You make pickles in New York and want to grow:

Every step used to be a human SDR's job. Now each is a workflow that runs continuously and escalates to a human only after the system has done its work.

6.3 The two-team model and the flywheel

• Internal team: builds the systems that find, score, enrich and route inbound and outbound pipeline.
• Forward-deployed team: diagnoses a prospect's go-to-market stack and builds a working workflow inside the prospect's data, before the sale, solving their exact problem. The build is the proof of concept; the artifact survives the contract and becomes onboarding.

internal team builds pipeline systems
        → qualified enterprise leads
        → forward-deployed team lands them by building inside the customer's data
        → customer-side learning about which plays work
        → fed back into the internal systems (better targeting, signals, plays)
        → loop

The loop closes only because both teams share the same skill profile; sales-style learnings do not survive the handoff into engineering systems.

6.4 Hiring and the buyer-as-seller principle

A GTM Engineer must (1) know go-to-market well (pipeline math, ICP, signal vs noise), (2) know the tooling deeply, and (3) be a good salesperson, internally or externally. Hire from your own user community: Clay's first GTM Engineer was already a power user. In markets where the product is a system and the buyer is a practitioner, the best motion is practitioner-to-practitioner: the people selling are the same kind of person buying, which collapses time-to-trust to a single working session.

6.5 The signal loop and the stack

Engineered plays run on signals, not cadenced lists. The canonical shape:

external signal → filter → enrich → personalize → send → log

Example. Pull every company that posted a "VP of Sales" or "Head of Revenue" job in the last 30 days → filter to Series A/B with 50–300 employees → enrich. One signal yields 200–400 highly relevant accounts a month. A team's real product is a library of these loops.

The tools are interchangeable; the orchestration logic is not. Leverage compounds in the workflow library.

6.6 The enrichment cascade (the one primitive worth building well)

Try provider A; on a miss, fall through to B, then C; pay only for the hit.

cascade.run({
  field: "work_email",
  input: { name, company_domain, linkedin_url },
  providers: [ { id: "provider_a", cost: 1 }, { id: "provider_b", cost: 2 }, { id: "provider_c", cost: 3 } ],
  policy: { stop_on_first_hit: true, max_spend_credits: 5, confidence_floor: 0.85, parallel_first_n: 2 },
});

It must guarantee six things: per-provider success billing · confidence normalized across providers · idempotency (no double billing on retry) · partial-failure recovery · budget visibility before escalating · observability (which providers were tried, which hit, latency, cost, confidence) so hit rates can re-rank providers over time. Most cascades a real team needs can be built from five to eight self-serve, API-first providers.

6.7 Example: how Clay backed into the role

Five silent years of building with revenue near zero (2017–2022) → launched as "a spreadsheet that pulls data from anywhere" (Feb 2022) → noticed the technical operators who got it immediately and repositioned around them → AI let the same users write personalized outreach (late 2022) → Rippling signed, first $1M ARR (March 2023) → hit the enterprise wall; refused to hire traditional sales and named the role "GTM Engineering" (2024) → the first GTM Engineer sold nothing for six months while finding the motion → landed Figma and OpenAI (end of 2024) → crossed $100M revenue and 16,000+ customers (2025). The role was a forced response to a product too systems-shaped to demo with slides.

Chapter 07 · Part III · Day 2

Lead pipeline operations: five gates

Distilled from Salesforce and HubSpot lifecycle, scoring and import documentation, enrichment and sending vendors' operator docs, and practitioner RevOps sources (2025–2026).

A lead is not a row in a list. It is a record that earns its way through five gates, and outreach readiness is a measured state of that record, its sender, its message and its batch: never a document's claim, and never the fact that a list was bought.

The pipeline is designed to be operated by agents under a human approval gate. Each stage names what the agent does, what the human decides, and what evidence opens the gate. Two operating rules:

• Loading is a script, not an agent. It parses, keys, matches, normalizes, writes and prints its counts. An agent is used only to repair a failed run. A hand-loaded row has no reproducible evidence.
• Use a schema-constrained classifier for every bounded classification (title → seniority, free tag → vocabulary, same-person / different / review, reply type, each copy-lint rule), not a chat model. Declare the option set before the call and threshold the confidence: below the threshold goes to a human review queue, never to an action.

7.1 The six stages and their gates

PROCURE ─G1─▶ LOAD ─G2─▶ ENRICH ─G3─▶ PRIORITIZE ─G4─▶ READY ─G5─▶ outreach
   ▲                                                                   │
   └────────────────────── MEASURE (continuous) ◀──────────────────────┘

Stages move forward only. A record goes back only on an explicit event with a reason: a bounce returns it to Enrich; a disqualification returns it to Prioritize; withdrawn consent moves it to a suppressed terminal state.

7.2 The record

Every field an agent writes carries provenance: source, provider or method, timestamp, cost. On conflict, "most recently verified wins" for contact fields, and "human overwrite wins forever".

• Identity keys, in match order: email (lowercased) → platform + handle → canonical profile URL → phone (E.164) → fuzzy name + company (adjudicated; only an above-threshold same-person answer merges). Same company, different people are distinct records. Merges keep the losing key as a secondary value.
• Company and role: company name, domain derived at load, title, separate seniority, ISO country, headcount band. A name plus a domain or company name is what an email finder needs.
• Reachability: route (email / LinkedIn / DM / phone / none), email status, verified-at, provider of record, evidence, next action when no route.
• Priority: fit grade (A–D), intent score (decays), tier, hold reason, disqualification reason, cohort (immutable), tags.
• Consent and legal: legal basis per jurisdiction (never blank at load), consent per channel with evidence, unsubscribes per channel, suppression reason.
• Lifecycle: stage, entered-at per stage, source, touch count, last touch and channel, stop reason.

Anti-pattern: a label that governs priority or eligibility living in a free-text notes column. If something governs a decision, it is a field with a controlled vocabulary.

7.3 Procure → G1

1. Write the ICP as testable fields before sourcing. A row that cannot be graded is parked with the missing input named.
2. Search wide, enrich narrow. People search is near-free; enrichment costs per hit.
3. Build to segment, not volume: 200–500 verified rows in one segment beat 5,000 generic rows. Decide up front the 30–50% of the addressable list you will ignore; reserve research for the top ~20%.
4. Every source carries a label and a yield (accepted ÷ rows, verified ÷ accepted, replies ÷ contacted).
5. Define suppression sets before sourcing: do-not-contact, existing customers, competitors, prior bounces and unsubscribes, anyone in a live cadence.
6. Freshness: re-verify any batch older than 90 days. B2B contact data decays roughly 2–3.6% a month; email addresses change for about 37% of people within a year.

G1: source label present · legal basis derivable · batch ≤ 90 days or re-verified · every row has an identity key · ≥ 85% of rows carry the grading inputs · suppression applied and counted.

7.4 Load → G2

1. Parse properly (RFC 4180: quoted fields can contain commas). A parser tested only on its own sample will refuse real data.
2. Key, then match in order. Loads are upserts on mapped fields only: never delete, never blank a filled field with an empty one.
3. Normalize: lowercase emails, handles without "@", domain from URL, ISO country, E.164 phone, canonical title with separate seniority.
4. Quarantine, don't reject. Rows missing every key go to quarantine with a reason and a count. Silent refusal is how a 251-row list imports zero rows and nobody notices.
5. Suppress at load: suppressed rows still enter, at a terminal stage with a reason, so they are never re-sourced.
6. Legal basis and consent are written at load. (Legitimate interest generally covers B2B professional addresses; some jurisdictions, Germany for example, require consent even for B2B.)

G2: duplicate rate < 5% · required-field completion ≥ 85% · every accepted record has stage, source, cohort and legal basis · quarantine and suppression counts reported with reasons.

7.5 Enrich → G3

Cascade order: free inference first (email pattern from a known domain) → providers cheapest-per-success first → stop at the first result a separate verifier confirms (a finder's "valid" is not deliverability) → escalate only on miss → stop on budget. Check each job's input prerequisites before the call; a missing input routes the record to the job that supplies it.

Route order: verified email → found-and-verified email → LinkedIn → DM. Email wins because it is the only route with a verifiable delivery state and no account risk.

Re-verify addresses untouched for 60–90 days; re-enrich after six months. Phone reveals cost 5–8× an email; give them only to tier-A records. Build a coverage and economics map per job (hit rate, latency, failure rate, cost per accepted result) from real receipts before buying another data subscription.

G3: every record has a route with evidence or a named next action · email validity ≥ 95% among email-ready · cost per accepted result recorded · no call without its prerequisite · batch budget held.

7.6 Prioritize → G4

Two axes, never blended:

• Fit grade (A–D): explicit, from ICP fields. Never decays.
• Intent score (0–100): implicit, from signals (an explicit contact signal, engagement, a trigger event, a reply). Decays to zero at 30 days or on a 30-day half-life.

Four different things, four different fields: order (who goes first) · hold (temporary: "not until X") · suppression (terminal: never) · disqualification (controlled vocabulary, reviewed weekly). Within a tier, reachability orders; it never sets the tier. Recalibrate monthly: a tier that converts no better than the tier below it is a wrong model, not a wrong list.

7.7 Ready → G5

• Contact gate: a resolved route with evidence · not suppressed · not held · required fields present · consent permits the channel · not in a live cadence anywhere · no other-channel touch the same day · every template token resolvable.
• Sender gate, email: SPF, DKIM and DMARC aligned · one-click unsubscribe headers · mailbox warmed ≥ 14 days (≥ 30 for a new domain) · inbox placement ≥ 80% on seed tests · 30–50 sends a day per fresh mailbox, 100–150 warmed, ≤ 500 per domain.
• Sender gate, LinkedIn: 20–25 connection requests and 30–50 messages a day; 10–15 a day for new accounts for two weeks; pause when acceptance drops > 20% week over week or pending invites exceed 30.
• Message gate: a versioned template with named tokens · passes the copy lint (Ch. 9) · physical address, truthful headers, real unsubscribe · under 80 words for step one · copy approved by a human before the pilot.
• Batch gate: pilot first, then measure, then scale. Automatic pauses: hard bounce > 2% pauses, > 5% stops · spam complaints > 0.1% pause, 0.3% stops · placement < 80% pauses · catch-all > 10% refuses the batch.
• Dispositions are rules: reply → a human · bounce or unsubscribe → stop that channel forever · auto-reply is not a reply · a conversion stops every sequence. Process opt-outs within 10 business days everywhere, immediately where feasible.

Readiness is a command's word, never a document's. Only a check that reads the fields and receipts may say "ready". The human approval is a stage change on the record, and the approval itself sends nothing.

7.8 Measure

1. Per sourcing cohort, at equal maturity: procured → accepted → reachable → tier A/B → contacted → replied → positive reply → converted. State cohort age and the observation window with every rate.
2. Unit costs: per accepted lead, per verified contact, per contacted, per converted. A zero denominator is an undefined cost, never a zero cost. Show cash spend and labour-inclusive totals separately.
3. Data-quality composite, weekly: duplicates < 5%, completion ≥ 85%, email validity ≥ 95%, stale records < 20%, quarantine backlog.
4. Sender health, daily. Model health, monthly. Rewrite service levels quarterly.
5. Never combine outcomes of different kinds into one claim.

Benchmarks (dated, external): cold-email reply average 3.4% (2025), top decile 10.7% · LinkedIn DM reply 10.3%, connection acceptance 29.6% (H1 2026) · cold-email cost per lead $150–$300 (2025).

Chapter 08 · Part III · Day 2

Signals: outreach at the moment of need

The highest-leverage target is not merely a person who matches the ICP. It is a person who just experienced a visible event that makes the offer timely. Static targeting scores people against a profile and then tries to manufacture urgency with copy. Signal-triggered targeting reverses the order: the urgency comes first.

8.1 Services have only one acquisition surface: the trigger

Nobody window-shops for a freelancer. Services have no browsing-intent surface, so the trigger is the only reliable way in, and the trigger class differs by type of seller:

Design implication: a services marketplace built on search and listings waits for demand that rarely arrives fully formed. A routing layer that detects the moment beats a directory.

8.2 The mechanic

public signal → event classification → need-window inference → offer mapping
   → personalized message → compliant channel route → send control → attribution feedback

The middle step is the important one. Ask not "does this person match?" but "what can this person now justify buying that they could not justify last week?" If several offers fit, choose the one with the strongest time sensitivity and lowest explanation burden. Acknowledge the moment before routing to the offer, especially for family, home, health or career events. New signal classes run review-first; a lane moves to automatic sending only once signal, copy and channel are trusted, and never for sensitive events or regulated offers.

8.3 Scoring

{
  "event_type": "new_pet",
  "signal_source": "instagram_post",
  "confidence": 0.86,
  "recency_hours": 12,
  "need_window": "first_30_days_of_pet_ownership",
  "offer_category": "pet_insurance",
  "channel": "review_first_dm",
  "risk_level": "medium",
  "reason": "recent public post indicates new pet ownership; insurance relevance is time-sensitive"
}

8.4 Compliance and taste

Do not infer protected characteristics · do not imply surveillance; be contextual, not creepy · write inferences as observations, not claims about private facts · no automated sends in regulated categories without approved copy · never contact people about grief, medical distress, minors or crises · respect platform terms and privacy law · suppress on complaint, block, unsubscribe or low confidence.

Why it works: demographic personalization says "people like you buy this". Signal personalization says "this just became relevant because of what changed". The first feels like targeting; the second, done with restraint, feels like good timing.

Chapter 09 · Part III · Day 2

The message: artifacts that survive triage

From Adrianna Lakatos's cold-outreach workshop at Founders Inc, May 2026.

Cold outreach is not a writing problem and not a volume problem. It is an artifact-design problem: the message must be specific enough that exactly one person, or their inbox agent, recognizes it as relevant in under three seconds, and frictionless enough that the reply is "hell yeah" or "no", never "I don't know what you're asking".

Cold outreach is a respectful one-to-one request for attention from someone who owes you nothing. Two consequences: volume cannot rescue a bad artifact (if it could have gone to 500 people, the reader knows they are number 501), and silence is the default. The job is friction removal, not persuasion; persuasion happens once the conversation exists.

9.1 The five-part subject-line rubric

Three of five is the minimum; five of five for accounts that matter. Score the subject, because it decides whether the body is read at all (~40–60 characters of preview).

9.2 Hell yeah or no

Three landing zones: hell yeah (they know how to say yes), no (they know how to say no), ambiguous (they never reply). A "no" is as good as a "yes": both close the loop. The usual cause of ambiguity is hedging. Always over-resolve the ask, even at the cost of looking blunt.

9.3 The eight-mistakes lint

One fail means rewrite.

1. Forgettable subject line.
2. Copy-paste opener (survives swapping the name).
3. Too long (over ~30 seconds to read).
4. Forgets they owe you nothing ("looking forward to your reply").
5. Ghosting yourself (treating no reply as a closed door).
6. All about you.
7. Too formal.
8. Every platform treated the same.

9.4 The three questions

Before sending, answer each in one sentence: What do you want? (not "to chat": "15 minutes on Tuesday at 2pm Pacific to hear this pitch") · Who can help? (named, and why them) · Where do they actually read?

9.5 Examples from the workshop

• Works: the Mark Cuban reply. A high-school student opened with shared context (an alumni tie to a program Cuban funds), a self-aware tone, and one specific ask: a place on the cap table. Put any membership overlap in the first six words.
• Works: the paid conversation. A marketer at Lovable assumed the founder would reply, offered to pay for the conversation, and was specific about the founder's product. If the reader's time is the bottleneck, price it openly.
• Works despite a typo. A follow-up that misspelled the recipient's name but showed obvious homework (inferred her firm from her email domain). Research signal beats polish.
• Fails: the $50 gift card. A small bribe wrapped around a generic pitch reads as an insult. Earn attention or buy it transparently; never the middle.
• Works: "Have you given up on this?" (Chris Voss) for a lead gone dark. Either reply is useful: it reopens the loop or surfaces the real objection.
• Right rubric, wrong route. A metric-dense investor subject scored well and still failed, because senior investors rarely open cold founder mail. The same artifact forwarded by a mutual connection is many times stronger.
• Fails: personality without specificity. "If this flops I'm switching to interpretive dance" tells the reader nothing.
• Works: the shared identity. "D3 player seeking to make labeling easy" sent to a coach. Something true about you that mirrors something true about them.
• Works: the 300-character rule. Farza Majd said publicly he answers every email under 300 characters with a clear ask. A 297-character email, with a PS apologizing for going over, got a reply. Targets broadcast their contact protocol; respect it visibly.

9.6 Craft at scale

• Target contact signal. Record, verbatim with source, any public statement of how a target wants to be contacted. Populate it for the top 20% of a list before sending.
• Segment, then write. Three hundred recipients means three groups of one hundred with three messages, never one blast. Segment by role, channel, trigger and mutual-connection tier until each message reads as written for its group.
• Write for two readers. Inbox agents now read first. Put the clear ask in the first sentence so a 20-token summary is accurate and relevant.

9.7 Follow-up: the value-add cadence

At least one value-add follow-up after every "no" before retiring a contact. In the cited case, a value-add reply to "I'm no longer at that firm" was forwarded internally and the deal eventually closed.

Chapter 10 · Part IV · Day 3

Conversion System Engineering

Any business that sells on the internet (store, SaaS product, services firm) is not a page with a button. It is an instrumented conversion system that reads each visitor's behaviour, infers where they are in the buying journey, fires the one intervention that visitor needs next, and keeps them after the sale. E-commerce is simply the most instrumented version; the logic is identical for software and services, and the leaks it closes (bounce at the top, churn at the bottom) are larger and more expensive in software.

10.1 Arrival is not intent

A visit carries almost no information. What separates a buyer from a bouncer is behaviour after arrival: what they view, how long, how often, where they came from, what they did before. The whole system does one thing: turn observed behaviour into an inferred intent state, and that state into the right action, in real time.

The five abstractions

• A1. Behaviour is the signal. Intent is revealed, not declared.
• A2. Resolve identity. One human across ad click, email open, phone visit. For B2B, resolve the account and stitch the buying committee.
• A3. Stage-matched intervention. Proof for the new, reassurance for the skeptical, an offer for the ready, a reason to return for the lapsed.
• A4. Value-exchange capture at the moment of intent. Ask only those whose behaviour says they are close, and pair the ask with something worth having. A captured non-buyer is recoverable.
• A5. Retention is revenue. A churned subscriber is a refund on your own acquisition spend.

10.2 The pipeline: Sense → Segment → Serve → Retain

          ┌──────────── CONNECTED DATA / IDENTITY LAYER (one profile per human) ────────────┐
          ▲                 ▲                    ▲                    ▲
   SENSE ───────▶  SEGMENT ───────▶      SERVE ───────▶       RETAIN ──────┐
   every signal,   intent state from     the one matched      loyalty, subs,  │
   real time       behaviour             intervention         post-purchase   │
          ▲                                                                   │
          └───────────────────────── lapsed customers return ─────────────────┘

It is a pipeline of decisions per person, not a funnel everyone walks once.

10.3 From attention to revenue

Every attention asset needs a downstream path attached. Views, likes and follows are attention signals, not intent.

attention asset → self-selected next action → message-matched path → value-exchange capture
   → proof / lifecycle nurture → Sales Event → retention / expansion

• Message match. The unit of analysis is the pair (creative_id, landing_path_id): the page repeats the hook's pain words, persona, offer frame and visual motif. Kill or iterate the losing pair, not the creative alone.
• Capture before the bounce. Engage first, then ask: a calculator result, teardown, template, quiz result, diagnostic, early access. A popup before value trains visitors to close it.
• Follower-to-conversation. Where a platform permits it, a new follow triggers one short reusable video asking "What are you looking for help with?" The reply turns an ambiguous signal into a declared problem. A follow is not consent for email or SMS.

The attribution contract. Every path writes to the same chain:

source_id → creative_id → landing_path_id → capture_event → lifecycle_touch_id → Sales Event → retention / expansion event

A Sales Event is the commercially meaningful outcome for this business: purchase, subscription, booked call, paid preorder, trial activation, signed proposal, renewal, expansion. The event changes by business model; the chain does not.

10.4 The sell-density dial

The leak nobody sees on a dashboard is the ask that was never made. Out of fear of looking salesy, published content drifts to all value and no ask. Selling is talking about the product: one piece that names what you sell, who it is for and the next action. Make sell-density (the share of published pieces carrying a direct ask) an explicit setting. A defensible starting point is every second piece asks. Install it as a time-boxed test (for example, 30 days) so it reads as an experiment, not an identity change, then tune the ratio with attribution data. Defend the floor at the human approval step, where asks get softened out one at a time.

10.5 The owned-channel campaign flywheel

harvest live signals → one campaign concept → channel-native versions → send and observe
   → promote proven messages into lifecycle flows → audit the flows → next harvest

The weekly harvest, four whats: what content did we make · what happened inside the company · what did customers and prospects ask · what did the community create. One concept, channel-native executions: email carries the full argument; SMS and WhatsApp carry the smallest useful hook; push and in-app attach to a live event; replies are an intended output. Each channel needs its own consent.

Campaigns are the test rail; flows are the compounding rail. Never fill welcome, abandonment, onboarding or win-back flows with untested copy. Ship as a campaign, judge it by its job, promote the winner into the flow with a version and date, keep measuring. Audit installed flows weekly for stale claims, broken triggers, decaying performance and unpromoted winners.

10.6 Creator supply: the source of trust-bearing creative

Testing 50–500 assets a month needs a procurement machine, not a calendar. A creator is any node more trusted than the brand alone: customers, users, experts, employees, founders, partners, niche operators.

source trust nodes → qualify → seed → brief → collect → clear rights and disclosure
   → decompose into variants → publish on matched paths → attribute Sales Events → rebook winners, retire losers

Pay is a variable, not the system: product seeding → modest fixed fee per approved asset → commission or rev-share → long-term partner economics for proven converters. Judge creators by rights-cleared assets, Sales Events influenced, revision burden and repeatability, not follower count.

Example: PROMIX. Ad-intelligence tools counted roughly 960 active Meta ads (one tool) against far fewer (another, ~34 new creatives a week). The transferable point is not the count; it is a high-throughput creator supply chain: many small trust nodes, partnership-ad permissions, one asset multiplied into many variants, and rebooking by outcome.

Example: IM8. Transferable: the offer makes subscription clearly better than one-time purchase; differentiated creative goes to matched pages; capture happens early; follow-up repeats proof. Not transferable: celebrity equity and unverifiable ad-platform performance claims.

10.7 The two leaks

Leak 1, bounce and drop-off: landing bounce · signup or checkout abandonment · activation drop-off (signed up, never reached the moment the product first delivers its core value; the most ignored leak in SaaS) · trial-to-paid drop.

Leak 2, churn: involuntary (failed cards; often 20–40% of churn and recoverable with no product change) · early churn (almost always onboarding) · late churn and value decay · silent churn (stopped using, not yet cancelled) · contraction.

Bounce tactics, in order

1. Define the activation moment ("sent first campaign", "connected first repo") and instrument it.
2. Message-match the page to the source.
3. One call to action, sub-two-second load, value legible in five seconds. Published A/B results point the same way: a plain account-creation screen beat a designed one (Ring); "#1 most downloaded book summary app" beat an aspirational headline; "Premium" beat "Boss Up, Go Premium"; "one month" beat "30 days" (Audible). Concrete beats aspirational; plain labels beat brand voice on the closing screen.
4. Progressive capture: no card before value.
5. Capture the drop-off as an event and fire a recovery sequence.
6. Treat signup → activation as its own measured funnel.

Churn tactics, in order

1. Smart dunning first.
2. Cancellation save flows: pause, downgrade, switch plan, talk to a human, targeted save offer.
3. Usage-decay early warning, then a proactive save, before the renewal date.
4. Onboarding to activation (most early churn is decided in week one).
5. Expansion at usage thresholds: the cheapest revenue there is, and what pushes net revenue retention above 100%.

Example: services agency. An account repeatedly reads the "SOC 2 readiness" page and a case study → trigger a tailored teardown ("how we would close your gap") with a strategic-call CTA. A proposal unopened for five days → a specific follow-up sequence. Proposal abandonment is the agency version of cart abandonment and one of the most under-worked leaks in services revenue.

10.8 If you build one thing first

Drop-off recovery plus dunning. Capture one abandonment event (signup_started_not_finished, activated_not_upgraded, proposal_sent_not_signed, cart_abandoned) and fire one matched recovery sequence; if there is recurring billing, add dunning right behind it. It recovers revenue from traffic you already paid for, needs only one sensing event, reuses an existing channel, and works in every vertical.

The slot is durable; the vendor is not.

10.9 Who owns what

┌──────────────────────────────────────────────────────────────┐
│ REVENUE OPERATIONS: strategy · segments · metrics · process  │  owns WHAT and WHY
├──────────────────────────────────────────────────────────────┤
│        SENSE  →  SEGMENT  →  SERVE  →  RETAIN                │  the system running
├──────────────────────────────────────────────────────────────┤
│ GTM / GROWTH ENGINEERING: integrations · data · identity ·   │  builds and runs HOW,
│ triggers · agents · the plumbing under every stage           │  underneath all of it
└──────────────────────────────────────────────────────────────┘

GTM Engineering is called Growth Engineering in e-commerce. It is not a station in the funnel; it is the substrate under it.

Chapter 11 · Part IV · Day 3

Lifecycle messaging

From Greg Jones's lifecycle email teardown (May 2026), the Green abandoned-action sequence, and practitioner reply-handling patterns.

Email is not a discount slot or a newsletter afterthought. It is the owned lifecycle surface where a business turns attention into trust, trust into intent, intent into revenue, and customers into repeat customers, by sending stage-matched proof, education, human updates, offers and support-derived answers. It applies once a person is reachable through consent: opt-in, signup, checkout, demo request, RSVP, account.

11.1 One email, one buyer-state problem

The common misuse: discounts only, sending only when you want something, content separated from revenue, recipient state ignored, best public content left to die on the feed. Instead, every email resolves one buyer-state problem: "who are you?", "I don't trust you yet", "is this for me?", "I need one last reason", "I bought once and have no reason to return", "I haven't reached value", "I'm drifting".

Principles: email is owned attention, not free attention · your best public content is pre-tested email material · human voice beats corporate camouflage · proof usually beats discount (a discount answers "is it cheap enough?"; proof answers "will it work for me?") · the archive is inventory · the support inbox is demand research · education sells without always asking · one email, one job.

11.2 Buyer states and email jobs

11.3 Where the content comes from

Tag every asset by stage, objection, product, ICP and proof type. Assemble the calendar from tagged inventory, not last-minute invention.

11.4 The flows

• Welcome: day 0 identity and promise with the best first artifact · days 2–3 proof or best work · days 5–7 the next surface (product, demo, community, trial).
• Education: category point of view → mistakes → how to evaluate → applied example → soft next step.
• Skeptic: built from support questions, sales objections, reviews and guarantees. The best one does not say "trust us"; it shows the exact concern was anticipated.
• High-intent: the abandoned-action sequence. Fast trigger while intent is warm · the direct CTA high and made to feel small ("takes less than 15 seconds") · proof matched to the likely blocker · rotate angles across follow-ups (value, education, social proof, comparison, urgency) · one visible next step. Swap "cart" for any abandoned commitment: signup, demo, trial, proposal, onboarding, RSVP.
• Cold-reply speed-to-lead. A reply to cold outreach is the highest-intent, fastest-decaying state in the system: webhook on reply · respond in under 90 seconds (a drafted answer grounded in the offer and in replies that converted before) · a phone number in the reply goes straight to a human to call · no booking → a 5–7 touch sub-sequence · CRM state updated live, and a booking pulls the lead out of every sequence · booked → a show-rate sequence, because the sale is the show, not the booking.
• Onboarding: the first-win email matters more than the thank-you email.
• Retention and expansion: renewal proof before the renewal date, account review before the expansion ask, usage milestones.
• Save and win-back: diagnose the likely reason; "we miss you" with no new value trains people to ignore you.

11.5 Operating it

• Weekly, mine support and sales language. Output: one support-answer email, one landing-page fix, one sales answer, one product note if the friction is real.
• Flows vs campaigns. Flows are the machine (behaviour-triggered); campaigns are timely injections. Healthy systems run both.
• Measure by job, not opens. Welcome: next open, first click, reply · Proof: bookings, checkout returns, proposal movement · Conversion: purchase, booked call, signed proposal · Onboarding: activation · Retention: renewal, churn reduction · Expansion: upgrade, seat add.

Chapter 12 · Part V · Day 3

A high-accountability sales culture

From Carles Reina, VP of Revenue at ElevenLabs, describing how a ~90-person go-to-market team moved from 10% to 40% outbound in under a year. This is one documented operating model; calibrate its intensity to your market and culture, and keep comp tied to the constraint (Ch. 2).

12.1 Compensation as a filter

• The 20× rule. Quota at 20 times base salary (a $100K base carries a $2M quota), against a SaaS norm of 6–10×. It self-selects for high performers and keeps teams lean; about 80% of reps hit it. Be honest with early hires that the ratio may be adjusted; the honesty itself filters.
• Double commission on upsells. Pay both the original AE (continued commission and quota retirement) and the CSM (paid on net revenue retention). Two people work every expansion, and territorial conflict, the thing that kills expansion revenue, disappears.

12.2 Pipeline reviews: public, monthly, drilled

• Monthly, remote, segmented by role and region; about 7–8 minutes per rep: closed deals, current pipeline, expected closes in the next 30 days.
• The drill-down: while reps present their best deals, pull up random deals and ask for details. It exposes inflated pipelines, shallow grasp and filler.
• Public praise and public criticism, with public acknowledgement when someone improves. Call out luck explicitly: "you hit your number through luck; next month you won't."
• The blocker list. End every review with "what are the blockers and how can I help?", then publish the condensed list by region so the whole company sees where revenue is stuck.

12.3 Forecast with deliberate negativity

If a deal could be $500K, enter it at $24K. It counters rep inflation, builds board trust through consistent over-delivery, and works as a forcing function: reps see they lack pipeline and build roughly twice as much.

12.4 Leaders outbound personally

The leader is the SDR-in-chief, cold-reaching CEOs and handing meetings to the team. It proves outbound works at every level, removes the "doesn't work for us" excuse, and keeps the leader's knowledge first-hand. Inbound to outbound is a culture change first: set a public goal, publish a weekly report naming each AE's and SDR's outbound activity, call out misses consistently until it is a running joke, and make the consequence of repeated misses known.

12.5 Non-fits versus pipeline builders

• Non-fits: never became product experts, lack outbound drive, or mismatch the culture. Part with respect (2–3 months' severance), and help them land well.
• Pipeline builders: below 50% of quota on paper while working real enterprise cycles in hard regions. Example: a hire from AWS below 50% later reached over 200% as the enterprise pipeline matured. The random-deal drill-down is how you tell the two apart; spreadsheet management cannot.

Chapter 13 · Part V · Day 3

The operating rhythm and the scoreboard

13.1 One cadence for the whole engine

13.2 One scoreboard

Three scoreboard laws: never combine outcomes of different kinds into one claim · a zero denominator is undefined, not zero · if a number does not say where the constraint is and how it is doing, it is decoration.

Chapter 14 · Part VI · Day 3

Installation sequence

Install in this order. The order bends to the constraint: if your constraint is at onboarding, steps 8 and 11 move to the front.

1. Name the owners. One strategy owner (CRO), one centralized RevOps, a build function. (Ch. 1)
2. State the throughput number and map the flow, every stage with an owner and entry and exit criteria. (Ch. 2)
3. Run the constraint arithmetic. One constraint. Exploit and subordinate before spending.
4. Choose the entry motion per segment. Lighthouse, Landgrab, bottom-up or hard market. (Ch. 3)
5. Write the offer architecture and the value justification. (Ch. 5)
6. Pick a bounded channel test with a minimum useful result, a budget and a time limit. (Ch. 4)
7. Instrument sensing and the attribution contract before optimizing anything. (Ch. 10)
8. Build the first lever: drop-off recovery plus dunning.
9. Stand up the lead pipeline with its five gates; run a pilot batch and read it before scaling. (Ch. 7)
10. Add signal loops and the message lint. (Ch. 8, Ch. 9)
11. Install lifecycle flows from proven campaigns, starting with speed-to-lead and the abandoned-action sequence. (Ch. 11)
12. Set comp, reviews and forecasting so they pay on constraint outcomes and expansion. (Ch. 12)
13. Re-run the constraint pass and repeat. The engine is a loop, not a project.

Chapter 15 · Part VI · Day 3

Master diagnostic

Answer yes or no. Every "no" names the chapter to work on.

Foundations

• Does exactly one person own revenue strategy and the number?
• Is RevOps one centralized team reporting to that person?
• Can you write demand, capacity and WIP for every stage today?
• Can you name the single current constraint, and was it confirmed by arithmetic?

Go-to-market

• Do you know whether your buyers ask "is it safe?" or "what does it cost?" first?
• Is your motion the one your quadrant permits, per segment?
• Is each channel assigned a job, with a minimum useful result defined in advance?
• Is CAC fully loaded, and compared at the customer rather than the lead?
• Does your offer have a default path, a comparison path, a value stack, a risk reversal and a commitment step?
• Can you show the leak, the decision gap, the mechanism and a conservative-to-aggressive value range?

Pipeline

• Is your pipeline a library of signal loops rather than a list?
• Does every record carry provenance, separate fit and intent, and separate hold, suppression and disqualification fields?
• Does a command, not a document, decide readiness, and does a breach pause sending without a human?
• Does every message pass the five-part rubric and the eight-mistakes lint?
• Does every "no" receive one value-add follow-up?

Conversion and retention

• Have you defined and instrumented the activation moment?
• Does every creative have a matched landing path, and is each pair measured?
• Does every path write to the same attribution chain ending in a Sales Event?
• Is drop-off recovery running? Is dunning running?
• Is sell-density an explicit setting with a floor?
• Does every lifecycle email do one job for one buyer state?
• Is a cold reply answered in under 90 seconds?

Team and rhythm

• Are AEs and CSMs both paid on expansion?
• Do pipeline reviews drill into random deals and end with a published blocker list?
• Is the forecast deliberately conservative?
• Is the scoreboard led by throughput and constraint health, with activity as context only?

Chapter 16 · Part VI · Day 3

The short list of anti-patterns

1. Calling the whole revenue organization "RevOps", or giving strategy two owners.
2. Hiring or buying tools before exploiting and subordinating the constraint.
3. Automating non-constraint stages and calling the extra inventory productivity.
4. Choosing Lighthouse because it is flattering.
5. Launching every channel at once; ranking channels by cost per lead.
6. Selling features instead of the economic case.
7. Treating a bought list as a pipeline; scaling before the pilot is read.
8. Blended scores that hide whether a record fits or is interested.
9. Readiness declared by a document.
10. Volume used to rescue a bad message.
11. Treating every visitor, subscriber or customer as the same state.
12. Publishing attention with no downstream path, or never asking for the sale.
13. Discounting when the blocker is belief.
14. Stopping the model at the close; treating customer success as a cost.
15. Comp that rewards local activity instead of constraint outcomes and expansion.
16. A dashboard that has not changed in two quarters.

Appendix A

Glossary

Appendix B

Sources and lineage

This framework is assembled from twelve component frameworks and their primary sources.

Company figures quoted from sources are those sources' claims and were not independently audited. Illustrative calculations are labelled as such.