Product-Led Growth: From Activation to Expansion

In July 2026, I was reviewing an AI product’s growth dashboard when one number took over the conversation: approximately 5,000 free daily active users. It looked impressive. It also answered the wrong question—activity is not the same thing as value, and value is not automatically willingness to pay (free credits can make almost any chart look alive).

Product-led growth is a go-to-market model in which users experience meaningful product value before a traditional sales process becomes the main growth driver. It works when users activate, repeat the valuable behavior, and develop paid or account-expansion intent. Signup volume or free daily active users alone cannot establish that PLG is working.

Our analysis measured approximately 5,000 users active daily through a four-stage funnel: valid signup, value-based activation, repeat value, and paid or expansion intent. This July 2026 case study uses Amplitude-style product metrics as a reference, but the self-reported figure remains Evidence C—not an industry benchmark. The diagnostic sequence is the reusable asset.

Key takeaways

• Define activation as observable receipt of core value, not registration or login.
• Measure the sequence from signup to activation, repeat value, and paid or expansion intent.
• Diagnose cohorts and use cases before changing price, adding sales, or buying more traffic.
• Freemium is a packaging choice; PLG is an operating model.
• Add sales assist when product usage reveals team, security, procurement, or expansion needs.

What is product-led growth?

Product-led growth is a go-to-market model that uses the product experience for acquisition, activation, retention, and expansion. Users reach useful value with limited friction, while marketing, customer success, and sales support the same journey. Our analysis applies 4 stages—valid signup, value-based activation, repeat value, and paid or expansion intent—to diagnose whether activity represents durable demand.

PLG does not mean “no sales” or “everything is free.” A trial can exist without PLG, and a product-led company can add human help for complex accounts. The decisive question is whether product behavior reveals value before a sales motion becomes necessary. This 2026 case-study framework works using Amplitude alongside Harvard Business School, OpenView, Atlassian, and Pendo; operating thresholds must still come from each product’s cohorts.

Background sources: Harvard Business School Online on product-led growth, OpenView’s product-led growth resources, and Atlassian’s PLG overview. For the broader channel decision, compare the site’s existing PLG versus sales-led growth guide.

Why is free DAU not enough?

Daily active users measure activity under a chosen definition. They do not prove that users received core value, returned because of that value, or intend to pay. Free usage may include testing, curiosity, internal accounts, one-off campaigns, repeated failures, or consumption driven by expiring credits.

In one July 2026 advisory meeting, a large AI product team examined whether approximately 5,000 free DAU represented genuine product value. The product, company, and participants are intentionally withheld. This is Evidence C: the figure was discussed in a meeting, its precise measurement window and denominator are not publicly verified, and it must not be treated as an industry benchmark.

The useful lesson was methodological. Instead of celebrating the headline, the team reframed the next decision around four questions: Which use cases produced a successful outcome? Which users returned for the same value? Which accounts showed payment or expansion intent? How much activity came from free allowances rather than durable need? The case did not prove that free users were low quality—it showed that DAU alone could not answer the commercial question.

Case note: how the team turned one headline into a diagnostic plan

The advisory plan began by removing staff, test, gifted, and obviously duplicated usage from the denominator. The team would then rank the top use cases and high-frequency accounts, compare activation and repeat behavior by signup cohort, and interview four groups: paid or high-frequency users, activated but unpaid users, new users who never activated, and users who stopped returning.

Each interview was designed around a recent workflow rather than an opinion survey. Participants would share their screen, reproduce a real task, and explain where value appeared or disappeared. The team would compare those statements with successful-output logs, activation timestamps, repeat events, and payment signals. Only after that diagnosis would it test free quota, packaging, onboarding, or sales assist.

This anonymized case preserves the decision process while removing the client identity, internal product details, and commercial information. The approximately 5,000-DAU figure remains self-reported Evidence C; the workflow is a Gingiris advisory method, not a validated industry conversion benchmark.

Which PLG metrics should you track?

Use a four-stage diagnostic funnel:

Stage	Operational definition	Common misread
Signup	Valid external account after removing staff, tests, and obvious duplicates	Counting every campaign registration as demand
Activation	Completion of a product-specific event that demonstrates core value	Treating login or page view as activation
Repeat	Repeated receipt of that value within a relevant usage cycle	Using one day’s DAU as retention
Paid / Expansion	Payment, seats, procurement, security review, or another credible expansion signal	Treating free-credit consumption as willingness to pay

This funnel is a diagnostic model, not a universal benchmark. The event definition and time window must come from the product’s job-to-be-done and observed cohorts.

Amplitude’s product metrics guide similarly separates activation, engagement, and retention concepts and emphasizes that teams need metrics tied to product behavior. Use external frameworks to organize analysis, but calculate thresholds from your own instrumentation and cohorts. Sources: Amplitude Guide to Product Metrics and Amplitude PLG guide.

How should product activation be defined?

Product activation is the first observable behavior demonstrating that a user has received the product’s core value. Registration and login are setup events unless they deliver the promised outcome. Our analysis evaluated activation inside a 4-stage funnel rather than treating approximately 5,000 users active daily as proof of value.

A defensible definition contains four parts:

1. Actor: Which user or account performed the behavior?
2. Value event: What outcome did the user receive?
3. Quality condition: What distinguishes success from an attempted or failed action?
4. Time window: How soon after signup must the event occur for the cohort analysis?

For an AI agent, activation might require a completed task with accepted output rather than a prompt. For a workflow product, activation might be a completed workflow with a usable result. For collaboration software, activation may require inviting another participant and completing a shared action. These examples can be instrumented using Amplitude, but they are not prescribed thresholds.

Pendo defines activation in relation to the behaviors that correlate with users realizing value, reinforcing the need for product-specific measurement. See Pendo’s product-led growth glossary.

How do you validate an activation event?

Start with candidate behaviors, then test whether they distinguish stronger cohorts:

• remove internal, test, promotional, and obviously duplicated accounts;
• group users by signup period and relevant acquisition source;
• compare candidates against repeat use and retention;
• inspect whether the action completed successfully, not merely started;
• interview users while reviewing their actual workflow;
• revise the event when it captures effort without value.

Correlation is a diagnostic signal, not proof that the event causes retention. The purpose is to choose an operational event that helps the team improve onboarding and product value, then keep testing it as the product and audience change.

How should cohorts diagnose the PLG funnel?

Analyze signup cohorts through the same sequence: valid signup, activation, repeat value, then payment or expansion signals. Segment by use case and acquisition source before averaging everything together. A channel can produce many signups but few activated users; another can produce fewer signups with stronger repeat behavior.

Pair behavioral data with four interview groups:

1. paid or high-frequency users;
2. activated but unpaid users;
3. new users who did not activate;
4. users who activated and later stopped.

Ask participants to share their screen and reconstruct a recent task. Compare what they say with event logs, successful outputs, return timestamps, and payment events. Interviews explain why a pattern may exist; logs test whether the pattern is common enough to matter.

Is freemium the same as PLG?

No. Freemium describes access and packaging: a user can continue on a free tier subject to limits. PLG describes how the company acquires, activates, retains, and expands users through the product experience. A freemium product can have weak activation, and a time-limited trial can support a strong product-led motion.

Evaluate free access by the behavior it enables:

• Does the allowance let the target user reach core value?
• Can the user repeat that value enough to form a habit or workflow?
• Does the limit distinguish genuine expansion from accidental consumption?
• Are compute cost, abuse, support load, and refunds included as guardrails?

Do not copy a competitor’s credit amount or conversion benchmark without matching product economics, audience, geography, period, and measurement definition.

When should a PLG company add sales assist?

Sales assist is a human motion triggered when product behavior reveals needs that self-service cannot efficiently resolve. Useful signals include multiple active teammates, seat growth, security or compliance questions, procurement requirements, deployment complexity, and sustained account-level usage. Our analysis treated sales assist as the fourth-stage response to paid or expansion intent—not as a substitute for activation. The July 2026 case study used Amplitude-style behavioral evidence rather than a sales-created lead score.

Sales assist should respond to demonstrated value, not compensate for a broken activation path. First, weak activation calls for better positioning, onboarding, and first success. Second, strong activation with weak repeat use calls for investigation of task frequency, output quality, and workflow fit. Third, strong repeat use with weak payment calls for tests of packaging, limits, price anchors, and purchasing context. Scale sales only when the pattern is sufficiently consistent to justify the cost.

What should the team do in the next two weeks?

A two-week growth experiment is a decision system that changes one major variable. Our 2026 case-study analysis works using Amplitude or equivalent product analytics to preserve the baseline, checkpoints, and outcome:

Field	Example question
Decision	What will we change after this test?
Hypothesis	Which audience, pain, and action should create which result?
Primary metric	What single measure decides the experiment?
Guardrails	What abuse, cost, complaint, refund, or brand risk must stay bounded?
Sample and budget	What minimum evidence and cost are acceptable?
Checkpoints	What will be reviewed at 2h, 24h, 7d, and when relevant 30d?
Decision rule	What causes Continue, Adjust, or Stop?

Day 1 establishes the baseline. Days 2–3 prepare instrumentation. Days 4–10 run the test without changing multiple variables. Days 11–12 interview high-intent and lost users. Day 13 reviews evidence; day 14 makes the decision.

This is the Gingiris Growth Experiment OS, version 1.0. It is a decision framework, not a claim that every PLG experiment should use identical durations. Teams still validating the problem can pair it with the product-market-fit checklist and the B2B SaaS growth playbook.

Common product-led growth mistakes

• Calling signup or login an activation event.
• Combining internal, test, gifted, and external users in one denominator.
• Averaging incompatible use cases and acquisition sources.
• Scaling paid acquisition before activation and repeat value are stable.
• Hiring sales to compensate for unclear product value.
• Treating free usage, Stars, or traffic as proof of commercial intent.
• Copying an external benchmark without its product, sample, period, and definition.

Frequently asked questions

What is the difference between signup and activation?

Signup creates an account. Activation records the first observable receipt of core product value. A login is usually setup; a successfully completed outcome tied to the product promise is a stronger activation candidate.

Can a product have high DAU without product-market fit?

Yes. Activity can come from free access, novelty, promotions, internal use, or unsuccessful attempts. Product-market fit requires broader evidence, including meaningful activation, repeat value, retention, and willingness to pay or expand.

What is a good activation metric for an AI product?

The metric should capture a successful, value-producing task for the target use case. It should distinguish accepted output from an attempted prompt and be validated against repeat behavior in the product’s own cohorts.

Which metric proves PLG is working?

No single metric proves it. A credible pattern connects valid signup, value-based activation, repeated value, and paid or expansion intent. Teams should monitor the transition rates and evidence behind each stage.

Does PLG eliminate sales?

No. Product behavior can qualify and educate users before sales engages. Sales assist becomes valuable for multi-user rollout, procurement, security, deployment, and expansion needs.

How often should activation be reviewed?

Review it when the product promise, onboarding, target segment, packaging, or instrumentation changes, and on a regular product analytics cadence. A previously useful event can become a vanity metric as the product evolves.

Sources and evidence note

This guide combines public PLG frameworks from HBS Online, OpenView, Atlassian, Pendo, and Amplitude with the Gingiris PLG Signal Diagnosis and Growth Experiment OS. The approximately 5,000 free-DAU example is an anonymous 2026 meeting sample (Evidence C, self-reported and not independently verified); it is included only to illustrate the diagnostic method.