PLG checklist: Four phases, each with exit criteria
The temptation with a list this long is to fix everything at once. This page gives you an order instead. Foundation (instrument and baseline), Activation (get users to value), Revenue (convert them, keep them, grow them), and then the loops that compound once the first three work. Move on when you hit the exit criteria, not when the calendar says so.
Phase 1: Foundation
Get instrumentation in place and establish baselines before optimizing anything. Everything in the later phases is measured against what you record here, so a missing baseline costs you a quarter later.
Analytics & Instrumentation
- Put a stranger in front of the product before anything else (Why PLG Fails) Fresh account, nobody helping. A product that cannot be evaluated alone will not be rescued by anything below
- Track these 5 events minimum If you can't measure it, you can't improve it:
- - Signup started/completed
- - Each onboarding step (with drop-off)
- - Aha moment action (the one that predicts retention)
- - Invite sent/accepted
- - Upgrade clicked/completed
- Pick three metrics and ignore the rest for now Activation rate, retention at a window that suits your cadence, conversion rate. Everything else can wait until these three move
- Name events so a new hire understands them object_action format: user_signed_up, feature_used, invite_sent. No abbreviations
Find Your Aha Moment
- Write down what you think the aha moment is (The Aha Moment) Then set it aside. The data will probably prove you wrong
- Split users: retained vs churned at your chosen window You need both groups to find what separates them. Daily tools can read day 7; weekly tools need day 30 or more
- Rank early actions by retention lift Compare users who did X against users who didn't, then test the top candidate by pushing a random group through it. Correlation gives a shortlist, not an answer
- Ask 5 retained users: "What almost made you quit?" The moment before they stayed is more revealing than why they stayed
- Fill in this sentence "Users who [action] within [days] retain at [X]% vs [Y]%"
Baseline Your Metrics
- Know your activation rate today (How to measure it) % reaching aha moment. Record it with a written definition and a date; the comparison that matters is your own next quarter
- Know your retention today (How to measure it) Pick the window that suits your product's cadence, then hold it fixed. Watch the cohort curve, not the single number: a curve that flattens means you have a product
- Know your conversion rate today (Freemium, How to measure it) Report revenue per thousand signups alongside the rate. A rate alone moves when you change the signup form
- Time your own signup with a stopwatch (Time to Value) Fresh email, timer running. That is one sample and a floor, not your TTV: the real clock starts at first contact, and what you report is the distribution
Find Your Biggest Leaks
- Find your worst onboarding step The one with the biggest drop-off. That's where you start
- Watch 10 users fail at that step Session recordings. Note what they try before giving up
- Screenshot every blank screen a new user sees (Empty State Design) Empty states kill activation. Count them
- Name your #1 friction point Not top 3. Pick one. You'll fix that one first
Collect Feedback Early
- Ask whether they got what they came for Outcome questions predict retention; satisfaction scores mostly do not. Ask once they have had a real chance to succeed
- Read every support ticket from the last 30 days Patterns in tickets reveal friction you can't see in analytics
- Ask one open-ended question monthly "What almost made you quit?" or "What's the hardest part?" Rotate the question
Align Your Team on Product-Led Growth
- Define the handoff between product and sales At what signal does sales engage? Write it down. Make sure both teams agree
- Share activation metrics weekly with the whole team Not just product. Sales, CS, and marketing need to see the funnel
- Train CS to recognize expansion signals They talk to users daily. They should know what a PQL looks like
- Kill one metric that doesn't matter If you track twenty, nobody focuses on any of them. Keep the ones a decision actually depends on
Phase 2: Activation
Get more users to the action that predicts retention faster. Activation is the most common of the three ways product-led growth fails, which is why it comes before conversion work rather than after it.
Reduce Time to Value
- Run your own signup on a slow connection and on a real phone (Time to Value) Chrome DevTools throttling, then an actual handset. Compare mobile and desktop activation for the same cohort; a large gap is a mobile problem, not a funnel one
- Count your signup fields, then cut one Cut every field not needed to deliver first value, then measure completion before and after
- Remove email verification before value Verify after users see value, not before. Test this week
- List everything you ask before the aha moment (Quick Win Architecture) Profile, preferences, integrations. Move each one to after first value
- Kill one blank screen (Empty State Design) Put a worked example beside the space, not inside it. Never fake numbers in a chart
- Cut time to value against your own baseline, quarter over quarter The five-minute figure is a practitioner convention, not a measured threshold. Beat last quarter
Fix Empty States
- Find every "No data yet" message and rewrite it Add what to do and why: "Add your first project to start tracking time"
- Add one template to your deepest blank screen (Empty State Design) Not the first-run screen, which somebody already designed. The one people hit on day three after real work
- Remove choices from your first screen One obvious button, not three options. Decisions kill momentum
- Watch a user who churned in 24 hours Where did they stop? That empty state is your biggest leak
Build Onboarding Flow
- Borrow the interface your users already know (Familiar Interface) Behavior, not appearance: shortcuts and panel positions where the hand expects them. Partial fidelity is worse than none
- Name what your users are afraid of doing (Interactive Sandbox) If you cannot name it, do not build a sandbox. If you can, make the mode unmistakable and make it resettable
- Show users how close they are to value, then check the rate moved (Aha Moment) A handful of steps driven by what they have actually done, not a fixed script. Compare activation before and after
- Test guided onboarding against skippable A forced tour is one more thing between signup and first value. For simple products, letting people skip usually wins
- Ask one question that changes the path (Personalized Onboarding) Only if the answer actually changes what users see. Otherwise don't ask
- Show tooltips one at a time, not all at once (Progressive Disclosure) Trigger on relevant action, not on page load
Set Up Re-engagement
- Trigger onboarding emails on behavior, not elapsed time (Aha Moment) "You have not invited anyone yet" beats "Day 3: here is how to invite teammates", and costs the same to build
- Email users who signed up but didn't activate Trigger on the step they stalled at, and name it. A calendar-based sequence sends the same mail to someone who already succeeded
- Trigger notifications on teammate activity "Sarah commented on your doc" pulls users back. Generic reminders don't
Phase 3: Revenue
Converting users, keeping them, and growing the accounts that stay. None of it works before Phase 2 does: a prompt shown to somebody who never got value is an interruption, not an offer.
Design Upgrade Triggers
- List every moment users want more (Smart Upgrade Triggers) These are your upgrade triggers:
- - A usage or history limit reached during real work
- - Teammate invited (collaboration proven)
- - Premium feature clicked (intent signal)
- Check that limits hit after the aha moment (Aha Moment) If users hit walls before value, you're killing conversion, not driving it
- Move one upgrade prompt from generic to contextual Show it when users hit the limit, not in a sidebar they ignore
- Rewrite your highest-volume upgrade prompt to name the work at risk (Smart Upgrade Triggers) Copy will not rescue a prompt firing at the wrong moment, so check the timing first and the wording second
Implement Pricing Strategy
- Aim for three paid tiers, plus whatever comes before them (PLG Pricing) Entry, middle and top, each with a distinct job. More than that turns the pricing page into a comparison nobody finishes
- Sign up for 3 competitors and note their limits Where do they gate? What do they give away? Take the best ideas
- Know what % of new ARR comes from expansion (Usage-Based Pricing) Whatever it is, know it and watch the trend. Split it before you read it, because a rise driven by price increases is not the product expanding
- Test reverse trial if premium creates dependency (Reverse Trial) Start on premium, then drop to free. Set the window from your product's rhythm, long enough for a habit to form, and test it against your freemium flow, not against a trial
Build PQL Scoring (If You Have Sales)
- Pull your last 50 conversions and find the pattern (PQLs & Product-Led Sales) What did they have in common before converting?
- - Hit a limit or attempted a gated feature
- - Invited 2+ teammates
- - Looked at your security, SSO or invoicing pages
- Pick 3 signals and score them Weight each by how strongly it predicted conversion in your own history, then work down the ranked list instead of setting a hard gate
- Alert whoever follows up as soon as the signal fires Slack ping or CRM task. The intent decays quickly, so measure your own delay from signal to contact
- Respond in hours, not days The signal decays fast. The published response-time conversion figures do not survive checking, but the decay is real
- Track PQL-to-close rate weekly (How to measure it) Compare it against your other lead sources, not against a published figure. Rank your signals by it
Create Retention Loops
- Find what accumulates in your product, and make it worth keeping (Data Lock-In) History, configuration, integrations, reputation. Gravity holds people because leaving costs them something real; friction only makes them resent you
- Show users what they accomplished this week (Habit Loops) Value summaries pull users back:
- - "You saved 3 hours this week"
- - "Alex commented on your project"
- - "You've completed 100 tasks!"
- Build a health score weighted toward outcomes (Reducing Churn) Weight outcome achievement highest, then teams active, accumulated setup, usage trajectory. Activity and sentiment predict least
- Alert CS when health drops, not when users cancel By cancellation it's too late. Catch the warning signs
- Interview the accounts that left, not the ones you saved (Reducing Churn) By the time churn shows up it is already decided. The value is in what it tells you about the next cohort
Enable Expansion
- Name your expansion path: seats, usage, tier, or an adjacent product (Land and Expand) Price rises are not expansion. Pick the path that grows because the customer succeeded. Focus there first
- Find the signal that predicts expansion (PQLs) Approaching a seat limit or usage cap? Colleagues being invited? A second team appearing? Build an alert
- Identify your internal champions by behavior (Bottom-Up Adoption) Users who invite colleagues are doing your selling. Find them and support them
- Check which path your largest accounts actually took (How to measure it) It is often not the one the team planned for, and only some of them compound
Test Pricing
- Ask 10 users what they'd expect to pay (PLG Pricing) Before you set prices. The answer often surprises you
- Test annual vs monthly default Which converts better? Which retains better? They're often different
- Try a higher price on new signups in one segment (PLG Pricing) Read conversion and revenue per account together. Healthy conversion with weak revenue is what underpricing looks like
- Track willingness-to-pay by segment Enterprise buyers need things SMBs do not, and pay for them. Charge accordingly
Cancellation & Offboarding
- Fix failed payments first (Reducing Churn) Card retries, expiry warnings, a dunning sequence. This is churn you lose to plumbing rather than to a decision, and it is the cheapest to recover
- Add an exit survey with 4-5 options (Reducing Churn) "Too expensive", "Missing feature", "Switched to competitor", "Not using enough", "Other"
- Offer a downgrade path before cancel "Would you stay on a smaller plan?" Some will. That's retained revenue
- Show what they'll lose before confirming "You'll lose 2 years of history and 5 integrations." Make the cost concrete
- Make cancellation easy, not hostile Dark patterns create angry ex-customers who warn others. Let them go gracefully
Phase 4: Loops
These compound, which is why they come last: a loop built on a funnel that does not convert simply circulates people who were never going to pay.
Acquisition That the Product Does Itself
- Check whether non-users already touch your product (Co-Experience) Shared links, invitations, exports, anything a recipient opens. If they do, that is your cheapest channel and it needs no budget
- Publish the security answers before anyone asks (Bottom-Up Adoption) A reviewer who cannot find them contacts sales, and the self-serve motion ends there
- Make the API usable before the sales call, or decide it never will be (API-First) Keys without a contract, or an honest gate. The half-open version wastes both sides' time
- Test whether your product is genuinely broken alone (Network Effects) If one person gets full value solo, you have collaboration features, not a network effect. Stop investing in invitation flows to fix it
- If you open-source anything, decide the license before the first contributor (Open Source) Relicensing later is where this pattern usually breaks, and it breaks in public
- Only run a waitlist if something is arriving at it (Waitlists) Scarcity filters demand. It does not create demand
Virality & Referrals
- Calculate K: (invites per user) × (acceptance rate) (Two-Sided Referrals) K above 1 is self-sustaining and almost nothing in B2B reaches it. Judge referrals on whether referred users retain
- Give both sides something Referrer gets X, invitee gets Y. Rewarding both sides removes the social cost of recommending, which is the mechanism
- Design one screen users want to screenshot (Achievement Sharing) Achievement, stat, or result worth sharing
- Badge the outputs that already travel (Embedded Virality) Invoices, reports, exports, shared links. Check first that a badge does not cheapen the sender
Content & Community
- Let users share templates others can clone (UGC Loop) User-created content scales without your effort
- Find your top 1% by usage and give them a direct line (Community-Led Growth) Recognition, early access, product input. They'll become evangelists
- Build pages at scale, but only where the substance already exists (Product-Led SEO) One page per thing your product genuinely does. A page whose only reason to exist is the query is a doorway page
Integrations & Ecosystem
- Build where the work already happens, not where the requests are loudest (Workflow Embedding) The request queue is unbounded and will never be finished. Find the surface your users sit in all day
- Compare retention: users with integrations vs without (Workflow Embedding) Compare against your own baseline. The published integration-retention figures are unsourced
- Check your integration backlog before building a marketplace (Marketplace Ecosystem) Variety means demand a marketplace could serve. Repetition means you have a roadmap. Build the integrations yourself until you cannot keep up
What should you compare your numbers against?
| Metric | What to compare it against |
|---|---|
| Activation rate | Your own last quarter, same definition. Then split by acquisition channel |
| Retention | Cohort curves over time. Watch whether the curve flattens, not the single number |
| Time to value | Your P50 and P90 together, plus the share who never arrive |
| Free-to-paid conversion | Your own baseline. Revenue per thousand signups is harder to game |
| Net revenue retention | Its own four components, separately. The published bars are worth less than the trend in your own |
| PQL-to-close | Your other lead sources, with a holdout group |
Where the evidence lives. Roughly half these items compress an argument a chapter makes properly, and those carry a link. The sourcing lives there rather than being repeated on every line, so if an item reads like an assertion and has a link, follow it: the citation, the caveat and the reason it is worded that way are all at the other end. The unlinked items are ordinary operational hygiene, and they are not carrying a claim that needs one.
Why there is no benchmark column. Almost every metric above is a ratio whose denominator you define, so two companies with identical products report different numbers depending on what they count as a signup or an active user. The metrics chapter sets out the argument; the practical consequence is that your own dated series is the only reliable comparison, and the published tables this page used to reproduce have been removed.
If you only do one thing
The order above assumes you do not yet know which transition is costing you. If you do know, that one comes first and the rest of this page waits.
The flywheel chapter has the exercise: rank your five transitions against each other, pick the worst, and ignore the rest until it improves. Everything above is what to do once you know which one you are fixing.
A product that sells itself is not one thing you build. It is a stranger finding you, reaching something real without help, coming back on a Tuesday, paying more as they get more, and telling somebody. Five transitions, and you can rank them this afternoon.
Start with the worst one.