The funnel is the product

For most of software's history, the person who bought the product and the person who used it were different people, and the buyer decided. Enterprise software was sold over golf, evaluated in RFPs, and installed onto users who had never asked for it. The entire marketing playbook — the ads, the analyst briefings, the steak — aimed at the signature, not the seat.

Then a generation of products walked in through the side door. Dropbox spread because sharing a folder was an invitation. Zoom spread because every meeting made guests into users. Slack spread team by team, under IT's radar, until "we already use it" was the business case. In each case adoption came first and purchase came second — the users arrived before the buyer did, and by the time money changed hands the evaluation was long over.

That's the coup, and it has a name: product-led growth. In one line: the product is the funnel. Awareness, interest, evaluation, conviction — the stages a funnel walks a stranger through — happen inside the product, self-serve, at whatever hour the user shows up. The demo isn't a calendar invite; it's the signup button.

Programmer's version: sales-led distribution is a synchronous call to a human — high latency, business hours only, doesn't scale past headcount. PLG is exposing the API publicly with great docs: the docs are the sales team, and every curious developer serves themselves. The funnel-and-loop machinery of the last chapter didn't disappear — it got compiled into the product.

The PLG motion has three verbs, run in order:

• Land. A stranger becomes a user in minutes, alone, for free or close to it. No form asking for their phone number, no "book a demo," no procurement. The bar is brutal: value before commitment, or they're gone.
• Habit. The product earns a slot in someone's week. This is the substrate everything else grows on — a product used weekly is a rehearsal for the renewal conversation nobody has to have. No habit, no expansion; churn is just the absence of habit, invoiced later.
• Expand. The account grows itself: more seats as coworkers get pulled in, higher tiers as usage hits limits, more spend as usage-based pricing meters what's already happening. The upgrade isn't a sales event; it's a threshold.

Land-and-expand is the classic name, and it's really a loop: each landed user invites the next, each expanded team justifies the next tier. It's a while loop with the customer's own org chart as the iterator. The pricing model is a load-bearing part of the motion — per-seat and usage-based pricing mean the contract grows without anyone renegotiating it, which is why PLG companies obsess over pricing pages the way sales-led companies obsess over proposal decks.

Notice what marketing does in each verb: for land, it buys and earns the traffic and then gets out of the way; for habit, it runs lifecycle nudges (the channel work of the email chapter); for expand, it arms the internal champion. Less persuading, more path-clearing.

"Free" is doing strategic work in this motion, and it comes in three deliberately different shapes:

• Freemium. A permanent free tier — Spotify with ads, Figma's starter files, Slack's message history limit. The free tier is a product-shaped ad that never stops running. Its job is distribution and network effects, not revenue.
• Free trial. Full product, ticking clock — the 14-day SaaS standard. The clock filters for intent: only people with a live problem start a timer. Its job is urgency and qualification.
• Reverse trial. Start everyone on premium, then let them fall to a free tier when the trial lapses. The user experiences the ceiling before choosing the floor, so the upgrade decision is a memory, not a promise.

Now the arithmetic that governs all three. Free-to-paid conversion in freemium products runs at roughly 2–5%. Read that number again: the business model is that 19 of 20 users never pay you. That only works under two conditions, both non-negotiable. First, free users must be nearly free to serve — marginal cost rounding to zero is the whole trick. Second, the free 95% must be doing something for you: inviting coworkers, publishing artifacts with your watermark, filling the search index, being the network the paid users pay to reach. Every free user is either distribution, a referral node, or a cost. A free tier whose users are none of the above isn't a funnel — it's a subsidy with a dashboard.

Programmer's version: freemium is a loss leader with telemetry. You're paying hosting costs to run the world's most honest ad — the product itself — and collecting usage data that tells you exactly who's ready to buy. Which is also why the choice of flavor is a strategy decision, not a default: pick freemium for network products, trials for tool products with sharp intent, reverse trials when the premium features are the story.

Run the economics yourself — same product, three flavors, two sliders.

If the product is the funnel, then the funnel's top is the first session, and the metric that rules it is time-to-value: how many minutes stand between "created an account" and "felt the point of this thing." Practitioners call the felt-the-point event the aha moment. Facebook's early-growth folklore made the form famous — get a new user to seven friends in ten days and they stay — and every PLG team since has hunted its own version: the first file shared, the first message in a channel with three coworkers, the first query that returns something useful.

Why this metric above all others? Because it sits at the top of the leak stack. A signup who never activates is a lead you already paid for, wasted; and every improvement to activation is inherited by every downstream stage for every future cohort. In last chapter's terms, activation is where the worst leak usually lives — and unlike a paid channel, fixing it doesn't raise your bill.

The work itself is unglamorous friction accounting. Every signup field is a toll. Every empty state ("Welcome! Your dashboard is empty") is a dead end wearing a smile. Every setup cliff — invite your team, connect your data source, install the agent — is a step where some fraction of users falls off, forever. The craft is to drag the aha moment earlier: templates instead of blank canvases, sample data instead of empty dashboards, single sign-on instead of forms, defaults instead of decisions.

Onboarding is the new landing page. The conversion craft that Part V's closing chapter applies to pages — cut friction, front-load value, remove choices — moved inside the product, and the stakes went up: a landing page that loses you gets another chance from another ad; a first session that loses you is usually the last session.

Sales-led marketing qualifies leads by guessing: this person downloaded a whitepaper, works at a company of the right size, has a plausible title — call them. That's the MQL, the marketing-qualified lead, and it's a demographic prior wearing a lanyard. It says who might buy. It cannot say who is ready.

PLG replaces the guess with a reading. The product-qualified lead is an account whose usage crossed a threshold: five seats active, the integration connected, the free tier's limit hit twice this month. Behavior beats firmographics because behavior is the purchase decision, mid-formation. The lead scored itself; the dashboard just noticed.

And here is the plot twist of the last decade: the winning motion turned out to be a hybrid. The PLG icons — Slack, Figma, Datadog — all built serious enterprise sales teams. Not to replace self-serve, but to sit on top of it: sales calls the accounts that are already lit up, armed with usage data, walking into conversations the product already started. Product-led sales: prospecting replaced by telemetry, cold calls replaced by warm thresholds.

Which re-answers the question this Part keeps asking — what is marketing's job here? Four things: make the product findable (someone must have heard of you before they try you); make the path frictionless; seed the loops that turn users into more users; and arm the champion — the internal fan who sells you upward needs the security page, the ROI one-pager, the deck they can forward to their VP. That last one is classic B2B persuasion, aimed at an audience of one, delivered through a user instead of a rep.

If PLG has a single scoreboard number, it's net revenue retention. The question it answers is beautifully blunt: take only the customers you had a year ago — no new logos allowed — and ask what their revenue is now. Churn and downgrades pull the number below 100%; seats added and tiers upgraded push it above. One number, netting the whole land-habit-expand motion.

The watershed is 100%. Below it, your revenue is a melting ice cube and new sales exist to out-shovel the melt. At exactly 100% you're on a logo treadmill — every dollar of growth must be bought fresh. Above it, something almost unfair happens: the company grows with zero new customers, compounding out of accounts it already won. NRR of 120% doubles revenue from the existing base roughly every four years, before the first new logo of the year signs. That compounding is why investors treat NRR as the tell for whether the expand loop is real — it's the interest rate on everything you've already sold.

Programmer's version: new-logo sales is O(n) — every unit of growth costs another unit of effort. Expansion revenue is the closest thing go-to-market has to compound interest: the base grows itself while you sleep, and the exponent is set by product quality, not ad spend. Pull the slider and watch what five years does to the same starting point.

Every motion this Part covers has honest boundaries, and PLG's are sharper than its fans admit. The motion fails wherever its one precondition fails: a self-serve path to real value. No path, no PLG. The common blockers:

• Value needs implementation. If the product only shines after a six-week integration, data migration, or org-wide rollout — ERP, core banking, anything sold as "a journey" — there is nothing for a lone curious user to try. The trial can't happen, so the motion can't start.
• Trying requires permission. In compliance-led buying — healthcare, defense, banks where installing software is a security event — the "under IT's radar" adoption story is a firing offense, not a funnel.
• The user can't become the buyer. Teachers adopt, districts buy; doctors prefer, hospital systems procure. When there's no bridge from usage to signature, adoption piles up as goodwill that never converts — unless you deliberately build the bridge (champion kits, usage reports the user can carry upstairs).
• No frequency, no habit. Products used once a year — tax filing, visa paperwork — can't build the weekly habit the expand loop feeds on. They can still self-serve; they can't compound.

And one failure mode is self-inflicted: PLG theater. A sales-led company bolts a free tier onto a product that can't deliver value self-serve, announces a motion change, and gets the worst of both worlds — free users who churn at the setup cliff, and a sales team now competing with its own giveaway. The free tier didn't create a funnel; it created a discount with extra steps.

It's tempting to file product-led growth as the chapter that retires the rest of the book — who needs advertising when the product sells itself? Resist the filing. PLG repeals nothing:

• Someone must have heard of you to try you. Self-serve funnels still start with mental availability — the memory structures of Part I decide which free product the curious user types into the search box. "The product sells itself" is true only after distribution and memory have done their invisible work.
• The two clocks still tick. In brand-vs-activation terms, PLG is a spectacular activation machine — the fastest, cheapest harvest mechanism software has found. It does very little for the slow clock. The PLG companies that became generational brands also bought billboards, ran Super Bowl ads, and built characters — once the loop proved out, they poured money into memory.
• Craft still compounds. Time-to-value is landing-page craft turned inward; the champion deck is Ogilvy's long copy for an audience of one; the pricing page is a positioning statement with numbers on it.

What PLG genuinely changed is the order of trust. For a century, marketing's job was to earn belief before the experience: promise, then prove. For software, the order flipped — prove, then charge. Try-before-trust. The promise didn't disappear; it moved into the product's first ten minutes, where it's kept or broken in front of instruments.

The checklist, compressed: Is there value in minutes, self-serve? If not, this isn't your motion. Which flavor of free, and why — freemium for networks, trials for intent, reverse for premium stories? Is the aha moment found, instrumented, and dragged as early as possible? Are PQL thresholds wired to a human who calls warm accounts? Is NRR above 100 — is the expand loop real, or is the ice cube melting?

One thread left dangling: this chapter's curves — activation decaying with minutes, conversion by flavor — are exactly the kind of claims a team will happily believe without checking. The discipline for checking them is a craft of its own, with its own ways of lying to you. Next: experiments, done honestly.