Sales Funnel skill

Sales Funnel

Design conversion paths, instrument them, find the leak, and ship experiments. This skill assumes you have (or will gather) five inputs: product/offer, ICP, ACV / price point, sales motion (PLG self-serve, sales-led, hybrid, transactional ecommerce), and current funnel analytics. Sales motion drives nearly every decision below — a $30/mo PLG tool and a $120k/yr enterprise contract share almost no funnel mechanics.

Related skills: use lead-scoring for the qualification/routing logic that sits between MOFU and BOFU; use social-media-kit for the TOFU content production this funnel consumes.

---

0. Diagnostic workflow (run this first)

Do not propose tactics before you've measured. Work the funnel in this order.

1. Gather inputs. Product, ICP (firmographic + role + pain), ACV, sales motion, sales cycle length, and current analytics (stage-by-stage volumes + conversion rates for the last 1–3 months).
2. Map the literal stages the buyer actually moves through (not the textbook ones). For PLG: Visitor → Signup → Activated → Paid → Expanded. For sales-led: Visitor → MQL → SQL → Opportunity → Closed-Won. Name each stage by an observable event, not a feeling.
3. Compute step conversion rates and absolute drop-off counts at every transition. Rank leaks by recovered revenue = (stage entrants) × (lift you believe is achievable) × (downstream conversion to revenue) × (ACV). The biggest %-drop is rarely the biggest $ opportunity.
4. Diagnose the top leak. Is it traffic quality (wrong ICP in), messaging (right people bounce), friction (they try and fail), trust (they stall at decision), or follow-up (no nurture)? Each has different fixes.
5. Form a hypothesis in the form: "For [segment] at [stage], [change] will lift [step metric] from X% to Y% because [mechanism]." Tie to a guardrail metric so you don't win the step but lose revenue.
6. Design the experiment (see §8). Pick A/B vs. before-after vs. holdout based on traffic. Pre-register the primary metric and minimum detectable effect.
7. Define the events you must fire to even measure this (see §6). If you can't measure the step, instrument before you optimize.
8. Set guardrails (see §9) — disclosure, consent, claims substantiation — before shipping, especially for scarcity/urgency, pricing, and email.
9. Produce implementation tasks: copy/design changes, event tracking, CRM stage definitions, lifecycle automations, and the analysis query.

Benchmark handling. Treat published "average conversion rates" as priors, not targets. They vary 5–10× by channel, ACV, and motion. Anchor on your own trailing 90-day baseline; only use external benchmarks to sanity-check order of magnitude. Never optimize a metric to a benchmark you can't tie to revenue.

---

1. Funnel stages (generic skeleton)

Adapt these to your motion using §5 blueprints. Stages must map to fired events (§6), CRM lifecycle stages (§7), and a clear owner.

TOFU — Awareness

• Goal: reach the right strangers; build a measurable audience you own (email/list), not just rented reach.
• Content: SEO articles answering buyer search intent, comparison/"vs." pages, short-form video, podcast, original-data reports, free tools. (Produce via social-media-kit.)
• Metrics: qualified sessions (ICP-fit, not raw traffic), new-visitor → known-contact rate, content-assisted pipeline. Watch bounce by source to catch bad traffic quality.
• CTA (segment, don't generalize): see §2 for CTAs by ACV/role. Avoid generic "subscribe/follow" as the only ask.

MOFU — Consideration

• Goal: convert anonymous traffic into known, consented contacts and educate them toward fit.
• Content: gated assets matched to commitment level (§3), nurture sequences, case studies by segment, ROI/comparison content, product-led "aha" demos.
• Metrics: visitor→lead rate, lead→MQL rate, sequence open/click only as diagnostics, content→opportunity influence. Hand off to lead-scoring here.

BOFU — Decision

• Goal: remove the last friction and close. Surface proof, pricing clarity, and a low-risk first step.
• Content: trial/sandbox, tailored demo, proposal, security/compliance pack, references, ROI calculator, pricing page.
• Metrics: SQL→opportunity, opportunity→won, win rate by segment, sales-cycle length, and the specific objection that stalls deals (track lost-reasons).

Retention & Expansion (post-purchase)

• Goal: drive activation → habit → expansion → advocacy. For recurring-revenue businesses this is where most LTV lives.
• Content: onboarding/activation milestones, in-product nudges, QBRs (sales-led), usage-based upsell prompts, referral program.
• Metrics: activation rate, time-to-value, Net Revenue Retention (NRR), gross churn, expansion rate, referral rate.
• On NPS: NPS is a relationship signal, not a funnel conversion metric. Survey at a stable lifecycle point (e.g., 30–60 days post-activation and periodically thereafter), not immediately post-purchase (you measure buying euphoria, not value). Require a usable sample (rough rule: ≥100 responses before reading the score; below that, read verbatims, not the number). Don't trigger experiments off small-sample NPS swings — segment and look at trend, and pair with a behavioral retention metric (NRR/active usage) before acting.

---

2. CTAs by sales motion, ACV, and buyer role

Generic CTAs ("buy now", "download guide") convert poorly because the next reasonable step depends on deal size and who's reading. Match the ask to the commitment a buyer can plausibly make at that ACV and role.

By role (overlay on the above):

• Economic buyer / exec: lead with outcome + ROI + risk reduction → "See the business case", "Book exec briefing".
• Champion / practitioner: lead with capability + hands-on → "Try the sandbox", "See the API docs".
• Technical evaluator: → "Read the security/architecture doc", "Run the POC checklist".
• Procurement/legal: → "Get DPA & SOC 2", "Download MSA template".

Rule: one primary CTA per page that matches the visitor's stage and likely role; secondary CTA offers the lower-commitment fallback (e.g., primary "Book demo", secondary "Get the report").

---

3. Lead magnets by funnel stage

Match the magnet to commitment level and to the data you're allowed to ask for (see §10 — every field beyond email needs justification and consent).

Progressive profiling: don't gate a TOFU checklist behind a 9-field form. Ask for email first; enrich role/company/use-case on subsequent conversions. Each new field should map to a routing or personalization decision — if it doesn't change what you do next, don't ask.

---

4. Channel → stage fit (where traffic enters)

---

5. Funnel blueprints by business type

Each blueprint lists the stages (as events), the key step metric to watch, the top leak to expect, and the highest-leverage fix. Use the §6 event schema to instrument them.

5.1 SaaS PLG (product-led, self-serve)

Stages: Visitor → sign_up → activated (hit the aha action) → paid → expanded (seats/usage up). Key metric: activation rate (signup→activated). This is the master lever in PLG; nothing downstream improves if users never reach value. Common leak: signup→activation. People create accounts but never complete the core action. Fix: define the single activation event from data (the action most correlated with retention), then engineer the onboarding to drive it: progressive setup, empty-state guidance, milestone checklist, "first value in <X minutes" goal. Reverse-trial (full features for 14 days, then downgrade) often beats a feature-limited free tier for activation. Monetization triggers: gate on value (seats, usage, advanced features), not time alone. Fire sales-assist for accounts above a usage/firmographic threshold (route via lead-scoring).

5.2 Sales-led B2B

Stages: Visitor → lead_captured → MQL (mql_qualified) → SQL (sql_accepted by sales) → opportunity_created → closed_won. Key metric: MQL→SQL acceptance rate (marketing/sales alignment) and opportunity win rate by segment. Common leak: MQL→SQL — marketing passes leads sales won't work, or leads rot in handoff. Fix: (1) a written SLA: lead definition, who follows up, in what time window (speed-to-lead matters — minutes, not days), and the bounce-back rule for rejected leads. (2) Score and route with lead-scoring so reps work the best-fit leads first. (3) Track and review lost-reasons monthly; the top reason tells you which BOFU asset/objection to fix (§11). Notes: long cycles → use multi-touch/influenced attribution (§8), never last-click. Build a security/compliance pack early; it unblocks enterprise BOFU.

5.3 Ecommerce / DTC (transactional)

Stages: Visitor → view_item → add_to_cart → begin_checkout → purchase → repeat (purchase #2). Key metric: add-to-cart→purchase (checkout completion) and repeat-purchase rate (the real margin driver). Common leak: cart→checkout→purchase abandonment (industry-wide, a large majority of carts are abandoned). Fix: reduce checkout friction (guest checkout, wallets/Apple-Pay/Google-Pay, shipping cost shown early, fewer fields); consented cart-abandonment email/SMS flow; trust signals at checkout (reviews, returns policy, security badges). Build repeat purchase via post-purchase email flows and a consented loyalty program — not dark-pattern subscription traps (§9). Measurement: GA4 ecommerce events below map 1:1 to these stages.

5.4 Agency / professional services

Stages: Visitor → lead_captured → discovery_booked → proposal_sent → closed_won → retainer/expansion. Key metric: discovery→proposal→won; and proposal close rate. Common leak: lead→discovery-call booked (high-friction, high-consideration purchase) and proposal→won (scope/price/trust). Fix: replace "contact us" with a qualifying application + instant calendar booking; use case studies by vertical as the trust mechanism; send tiered/options proposals (good-better-best) with clear scope to combat price objections. Productize a paid "audit/assessment" as a low-risk BOFU entry that converts to retainer.

5.5 Course / creator / education

Stages: Audience → email_subscribed → webinar_registered/free_lesson_viewed → enrolled → completion → advocacy. Key metric: subscriber→customer rate; and for high-ticket, webinar/launch attendance→purchase. Common leak: subscriber→buyer (audience that consumes free content but never buys). Fix: segment the list by intent (interest survey on subscribe); run launch sequences with genuine, substantiated deadlines (a real cohort start date) rather than fake countdown timers (§9); offer a low-ticket tripwire to convert subscribers to buyers, then ascend. Show completion/outcome proof, not just testimonials.

5.6 Marketplace (two-sided)

Stages (per side): Visitor → signup (supply and demand) → first listing / first search → first_transaction → repeat / liquidity. Key metric: liquidity (match/fill rate) and time-to-first-transaction on each side; balance of supply vs. demand acquisition. Common leak: the under-supplied side (a marketplace dies from the cold-start/chicken-and-egg problem). For each new buyer cohort, the leak is search→first_transaction if inventory is thin. Fix: seed/concentrate supply in a narrow niche or geo before scaling demand; subsidize the constrained side; measure per-side funnels separately. Demand acquisition is wasted spend if supply liquidity isn't there to fill it.

---

6. Conversion-event schemas (instrument before you optimize)

If you can't measure a step, you can't optimize it. Define events once, fire them consistently across web/product, and map them to CRM lifecycle stages (§7). All of the below is subject to consent (§10).

6.1 GA4 (recommended events)

GA4 is the standard since Universal Analytics was sunset (UA stopped processing data on 1 Jul 2023; standard properties' historical data was removed thereafter). Use GA4's recommended event names for ecommerce so reports work out of the box.

// Fire these only after Consent Mode/CMP grants analytics_storage (see §10.1);
// before consent, Consent Mode sends cookieless pings rather than full events.

// SaaS PLG (custom events)
gtag('event', 'sign_up',   { method: 'email', plan: 'free' });
gtag('event', 'activated', { milestone: 'first_project_created' }); // your aha action
gtag('event', 'purchase',  { value: 30, currency: 'USD', items: [{ item_id: 'pro_monthly' }] });

// Ecommerce (GA4 recommended events — names matter, GA4 builds funnels from them)
gtag('event', 'view_item',     { currency: 'USD', value: 49.0, items: [/* ... */] });
gtag('event', 'add_to_cart',   { currency: 'USD', value: 49.0, items: [/* ... */] });
gtag('event', 'begin_checkout',{ currency: 'USD', value: 49.0, items: [/* ... */] });
gtag('event', 'purchase',      { transaction_id: 'T123', currency: 'USD', value: 49.0,
                                 tax: 4.0, shipping: 5.0, items: [/* ... */] });

// Lead gen
gtag('event', 'generate_lead', { lead_source: 'gated_report', value: 0 });

Server-side caveat: GA4's recommended path for resilient measurement is the Measurement Protocol via server-side tagging (a server-side GTM container, e.g. on Cloud Run). It survives ad-blockers and ITP/Safari cookie capping better than client-side gtag — but it is not a consent bypass. Server-side hits still require a lawful basis/consent for the user, and you must still honor Consent Mode signals server-side. Send a stable client_id/user_id and never PII in event params. As of Jun 2026, verify event names and Consent Mode behavior at https://developers.google.com/analytics and https://support.google.com/analytics.

6.2 Segment / RudderStack (warehouse-first CDP)

Use the standard spec so every downstream tool agrees. Gate track/identify on consent category = analytics/marketing.

// Identify a known person (after they consent + convert)
analytics.identify('user_123', {
  email: 'placeholder@example.com', // hash or omit if consent not given for marketing
  company: 'Acme', role: 'engineering_manager', plan: 'trial'
});

// Track funnel events (consistent names across web + product + server)
analytics.track('Signed Up',      { plan: 'free', source: 'organic' });
analytics.track('Activated',      { milestone: 'first_project_created' });
analytics.track('Lead Captured',  { magnet: 'benchmark_report', icp_fit: true });
analytics.track('Trial Started',  { plan: 'pro' });
analytics.track('Subscription Started', { mrr: 30, plan: 'pro_monthly' });

Naming convention: Object + past-tense Verb, Title Case ("Order Completed", "Lead Captured"). Pick one tense and casing and enforce it in a tracking plan; inconsistent event names are the #1 cause of unusable funnel data.

6.3 UTM conventions (lock these down)

Inconsistent UTMs destroy attribution. Standardize and validate:

Rules: never UTM-tag internal links (it overwrites the real source); use lowercase everywhere (UTMs are case-sensitive); enforce a builder/spreadsheet so reps can't free-type. Capture first-touch and last-touch UTMs into the CRM (§7) on the lead record.

---

7. CRM lifecycle stages + handoff rules

Funnel events feed a lifecycle model in the CRM. Define stages as states a contact/deal is in, with explicit entry/exit criteria and an owner at each step.

7.1 Lifecycle stages (HubSpot-style; Salesforce equivalents noted)

7.2 Handoff rules (write these as an SLA)

• MQL→SQL: define MQL numerically (score + required firmographics). On MQL, route to a rep via lead-scoring rules. Speed-to-lead: first touch within minutes for inbound demo requests; conversion drops sharply with delay.
• Rejection/bounce-back: sales can reject an MQL with a reason code (not-ICP, no-budget, wrong-timing). Rejected→back to nurture, not deleted. Review reason codes to fix scoring.
• Recycling: closed-lost and gone-cold opportunities re-enter nurture with a timestamp and reason; don't re-pitch identically.
• Source of truth: pick one system as the funnel system-of-record (usually the CRM) so marketing and sales report the same numbers. Sync product/CDP events in; don't maintain two conflicting funnels.

---

8. Attribution & experiment design

8.1 Attribution — pick the model to the motion

Default stance: for sales-led B2B use multi-touch / influenced pipeline plus periodic incrementality tests for paid; for short ecommerce use last-/data-driven but verify big spend with holdouts. With cookie loss (§10), single-cookie click attribution is increasingly unreliable — lean on first-party events, logged-in user_id, and incrementality.

8.2 Experiment design

1. One primary metric, defined as a step conversion rate (not a vanity metric), plus a guardrail (e.g., revenue/visitor, refund rate, unsubscribe rate).
2. Estimate sample size before launch from baseline rate + minimum detectable effect (MDE) + power (0.8) + significance (0.05). If you can't reach the sample in a reasonable window, you don't have the traffic for an A/B test — use a before/after with a holdout or sequence test instead, and say so.
3. Method by traffic:
   • High traffic → randomized A/B (or multi-armed bandit if you must optimize live).
   • Low traffic / long cycle → holdout group, pre/post with control market (geo-holdout), or qualitative + funnel-step analysis.
4. Don't peek: fix the horizon (or use a sequential test designed for peeking). Calling significance the moment p<0.05 inflates false positives.
5. Decision rule pre-registered: ship if primary lifts ≥ MDE and no guardrail regresses; otherwise iterate or revert.
6. Log the result (win/loss/inconclusive + effect size) in an experiment log so you build institutional knowledge, not folklore.

---

9. Persuasion vs. dark patterns — guardrails

Funnels move money and exploit psychology; that puts them squarely in scope of consumer-protection law. Persuasion is fine; deception is illegal and brand-damaging. As of mid-2026, regulators (US FTC, EU under the Digital Services Act / Unfair Commercial Practices Directive, and others) actively pursue dark patterns. Verify current rules with counsel for your jurisdictions.

Allowed (honest persuasion):

• Real social proof (true counts, real reviews — and you must be able to substantiate them).
• Genuine scarcity/urgency (actual stock level, a real cohort start date, a real promo end date).
• Anchoring with real reference prices; good-better-best tiering; risk-reversal (real money-back guarantee you honor).
• Clear, prominent CTAs and benefit-led copy.

Prohibited (dark patterns — do not implement):

• Fake scarcity/urgency: countdown timers that reset, "only 2 left" when untrue, fabricated "12 people viewing".
• Forced continuity / subscription traps: hard-to-cancel subs, hidden auto-renewal, pre-checked upsells. (The FTC's federal Click-to-Cancel rule was vacated in July 2025 and replacement rulemaking is underway, but ROSCA, FTC Act Section 5 enforcement, and state auto-renewal laws (California and others) still require clear auto-renewal disclosure, express consent, and a simple cancellation path; EU law requires cancellation be as easy as signup.)
• Confirmshaming ("No, I don't want to save money"), disguised ads, bait-and-switch pricing, drip pricing (hiding mandatory fees until checkout — increasingly explicitly illegal).
• Sneaking items into carts; roach-motel flows; trick questions in opt-ins.

Claims substantiation: any performance/ROI/health/earnings claim ("2× your conversions", "lose 10 lbs", "$10k/mo") must be truthful and substantiated before you publish it; testimonials must reflect typical results or carry a clear disclosure. For regulated offers (financial, health, supplements, credit, crypto, gambling), add the legally required disclosures and get professional/legal review — this skill is not legal advice.

---

10. Privacy-safe measurement & email compliance (mid-2026)

The cookie-and-form playbook from 2019 is non-compliant today. Bake consent and data minimization into the funnel from the start.

10.1 Consent & tracking

• Consent Mode / CMP first: load a consent management platform; do not fire analytics/marketing tags before consent where consent is required (EU/EEA/UK under GDPR + ePrivacy; and increasingly US states). Google requires Consent Mode v2 to use Google audiences/measurement for EEA traffic; configure ad_storage, analytics_storage, ad_user_data, ad_personalization signals. Verify current requirements at https://support.google.com/analytics and https://support.google.com/google-ads.
• Third-party cookies are unreliable, regardless of any single browser's roadmap: Safari ITP and Firefox block them by default, ad-blocker usage is high, and Chrome continues to give users blocking controls. Design for a post-cookie world now: prioritize first-party events, logged-in user_id identity, server-side tagging (with consent), and consented email as your durable identity graph. Do not architect attribution around a 3p cookie surviving.
• Data minimization: collect only fields tied to a decision (§3). Hashing emails, IP anonymization/truncation, and short retention windows reduce risk. Map where lead data flows (CDP, CRM, ad platforms) and ensure a lawful basis for each.
• US state laws (CPRA/California, plus Colorado, Virginia, Texas, and a growing list) require honoring opt-outs of "sale/sharing" — including treating cookie-based ad targeting as "sharing." Support Global Privacy Control (GPC) signals; offer a "Do Not Sell or Share My Personal Information" path.

10.2 Email & messaging consent

Operational rules: maintain a suppression list and never email opted-out/bounced contacts; double-opt-in is best practice for EU/Canada lists; segment by consent scope; keep proof of consent (timestamp + source). SMS marketing has its own consent rules (US: prior express written consent; carrier rules apply) — treat it like email-plus.

10.3 Payments

For checkout/billing funnels, handle payment data via a PCI-compliant processor (e.g., Stripe) — never collect raw card numbers yourself. For subscription/billing funnel mechanics and dunning, see the stripe-billing skill if present in the catalog.

---

11. Objection handling (BOFU)

Surface objections early, then arm sales/copy with proof-led responses. Always track lost-reasons so you fix the objection that actually loses deals. Below are field-tested responses grouped by objection type — adapt specifics to your product; never make an unsubstantiated claim (§9).

Price / budget ("Too expensive", "No budget")

• Reframe to ROI/cost-of-inaction: "What's the cost of [status quo] over the next year?" Quantify with their numbers, not yours.
• Tiering: offer good-better-best so "too expensive" becomes "which tier."
• Payment terms / pilot: annual vs. monthly, a paid pilot, or phased rollout to fit a smaller initial budget.
• Don't lead with a discount — it trains buyers to wait and signals your price is fake. If you discount, exchange it for something (annual commit, case study, multi-seat).

Trust / proof ("How do I know it works?", "Never heard of you")

• Segment-matched proof: case study from a similar company/role; reference call; a quantified outcome.
• Risk reversal: money-back guarantee, opt-out pilot, SLA — and honor it.
• Reduce perceived risk of the first step: free sandbox, no-card trial, short pilot.

Timing ("Not now", "Next quarter", "We're too busy")

• Diagnose real vs. stall: "If budget/time weren't a factor, would this be a fit?" If yes, it's a timing problem; if no, it's a fit/value problem — solve the right one.
• Cost of delay: quantify what waiting a quarter costs.
• Lower the activation effort: "We do the setup; you need ~2 hours total." Recycle to nurture with a dated follow-up if genuinely later.

Authority ("I need to check with my boss/team")

• Multithread: ask to include the economic buyer; offer an exec-briefing asset tailored to them.
• Arm the champion: give a one-page internal business case they can forward (don't make them rebuild your pitch).
• Map the buying committee early so this objection never surprises you.

Integration / switching cost ("Will it work with our stack?", "Migration is painful")

• Show the integration (docs, native connector, API) and a migration path/tooling.
• Concierge migration / onboarding for higher ACVs; quantify time-to-value.
• De-risk with a parallel pilot so they don't rip-and-replace blind.

Security / compliance ("Is our data safe?", "We need SOC 2 / DPA")

• Have the pack ready: SOC 2 / ISO report, DPA, sub-processor list, pen-test summary, data residency options.
• Route to technical/security evaluator with the architecture doc; don't make sales improvise security answers.
• This objection blocks enterprise BOFU — build the materials before you go upmarket (§5.2).

Competitor / status quo ("We already use X", "We'll build it ourselves")

• Differentiate on the dimension they care about, not a feature checklist; use a fair "vs." comparison.
• Build-vs-buy math: total cost of building + maintaining vs. your price + time-to-value.
• Switching support: migration help + a side-by-side pilot to prove the delta.

Process: tag every closed-lost deal with a reason code, review monthly, and feed the top reasons back into (a) BOFU content/assets, (b) the relevant blueprint fix in §5, and (c) lead-scoring (if you keep losing on fit, you're scoring fit wrong).

---

12. Implementation checklist (output of this skill)

When you finish a funnel design/audit, produce these artifacts:

• Stage map (events) with current step conversion rates + ranked leaks by recovered-$.
• Top 1–3 hypotheses with primary metric, MDE, guardrail, and method (§8).
• Tracking plan: event names + properties for GA4/CDP (§6), consent-gated, with a server-side note where relevant.
• UTM convention doc + builder (§6.3).
• CRM lifecycle definitions + MQL→SQL SLA + lost-reason codes (§7).
• Lead-magnet/CTA matrix tuned to motion+ACV+role (§2, §3).
• Compliance checklist: CMP/Consent Mode v2, GPC/opt-out path, email consent model per region, suppression list, claims substantiation (§9, §10).
• Experiment log started; first test scheduled.