Product-Led Growth & Virality Strategy — Print-Flow-360
By Pritesh Yadav 39 min read —
Provenance. Synthesized 2026-06-16 from a 5-stream multi-agent research workflow (PLG/self-serve, viral loops & network effects, activation/aha engineering, growth-experiment discipline, freemium economics), each independently grounded in the actual codebase. Benchmarks are flagged by confidence (
solid/directional/vendor-blog/weak). Treat every external number as a hypothesis to test against our own funnel, not a forecast — most of the “reverse trial converts ~24%” and “branding loop k-factor” figures are vendor-sourced and self-reported. This doc builds on the existing GTM research inreadme/(cross-referenced in its own section); it does not restate it.
Executive Summary
Print-Flow-360 is a textbook product-led-growth candidate — low ACV ($99–$199/mo), a one-person buying committee (Maria the print-shop owner), and a product whose value is self-evident the moment a store goes live and an order lands. The good news: roughly 80% of the PLG machinery already exists in the code (a Plan/Subscription spine with JSON feature flags, PlanLimitsService enforcing per-tenant caps, a usePlanLimits composable that already returns plain-English upgrade reasons, and a SetupChecklist that tracks the exact activation signals a product-qualified-lead needs). The bad news: none of it is wired into a growth motion. Activation lives in localStorage and stops at “preview your store” — it never reaches the actual aha (a real order). Plan limits fail open when no plan is present and fire as hard blocks rather than contextual upsells. There is no funnel instrumentation, no trial countdown, no PQL signal, and zero “Powered by Print-Flow-360” attribution on the most-trafficked surfaces a web-to-print tool could dream of.
Four opinionated conclusions cut across all five streams:
- Reverse trial, not freemium. Start every signup on the top (Pro) plan for 14 days. Freemium converts at ~2–5%; a no-card reverse trial converts ~18–32% (directional) for this profile — and a permanent open free tier is actively wrong here because every live store costs real per-tenant infra (hosting + S3 + PDF compute) with no compensating viral upside.
- The post-trial fallback is a dormant/parking tier, not a usable free product. Keep the owner’s designs, catalog, and pricing rules alive and editable (the habit/lock-in layer), but take the public storefront offline (no hosting cost, can’t cannibalize the cheapest paid plan). The paid fence is the live, order-taking, publicly-hosted storefront — the one thing a shop owner actually pays to get.
- Instrument the funnel before optimizing anything. We literally cannot measure activation today. This is the #1 unblocker for the PQL, the reverse-trial card-capture, and any experiment program.
- Ship the “Powered by Print-Flow-360” attribution loop now — it’s the highest-volume, lowest-effort growth surface we have, and it’s currently 0% built. But be honest with stakeholders: it is sub-viral amplification (k ≈ 0.15–0.4), not a self-sustaining engine, and it is not a network effect.
TL;DR — Highest-Leverage Moves (ordered by impact ÷ effort)
| # | Move | Effort | Impact | One-line why |
|---|---|---|---|---|
| 1 | Ship “Powered by Print-Flow-360 · Build your own store →” on every tenant storefront footer | S | High | Free impressions on the highest-traffic surface we own; zero owner behavior change |
| 2 | Add a remove_branding/white_label flag to the Plan features JSON and gate footer + proof page + exports on it | S | High | One flag monetizes the loop and gives the clearest plain-language upgrade reason (“remove the badge”) |
| 3 | Reframe pre-seeded defaults as “You can already take orders” on the dashboard | S | High | Offline payment + sample catalog + shipping are already seeded; surfacing it compresses time-to-value to minutes |
| 4 | Convert to a no-card 14-day reverse trial on the Pro plan | M | High | ~24% (directional) vs ~4.5% freemium (solid); status enum already supports trial→active |
| 5 | Instrument the activation funnel server-side (signup → product → publish → first-order events + persisted milestones) | M | High | Nothing else can be optimized until this is measured; unblocks PQL, card-capture, experiments |
| 6 | Extend the checklist to the real aha: add “place a test order” + a first-real-order celebration | M | High | Today the checklist ends at “preview” — owners never reach the value moment that predicts retention |
| 7 | Turn hard plan-limit blocks into contextual “value-then-fence” paywalls rendered at the point of friction | M | High | The block UX reads as “broken” (violates §0); make it explain value + offer a path |
| 8 | Build the dormant/parking tier (data alive, storefront offline) as the reverse-trial fallback | L | High | Cannibalization fence + cost-to-serve fix in one move; replaces the dead-end no-plan.vue lockout |
| 9 | Fix the fail-open hole: free/dormant tier must be a real Plan row with explicit caps, never the absence of a plan | M | High | PlanLimitsService returns allowed=true with no plan — a naive free tier would be silently unlimited |
| 10 | Wire the PQL card-capture trigger to fire on first_order_received, not on trial day-N | S | High | Behavioral triggers convert ~2.5× calendar triggers (directional); willingness-to-pay peaks at the aha. Depends on landlord-side subscription billing + card-on-file actually working — see §9 |
| 11 | Run a 30-min weekly growth meeting off an ICE backlog; default to fake-door/qualitative, reserve true A/B for landing + pricing only | S | High | We’re too low-traffic for classic A/B; velocity of cheap experiments is the real lever |
| 12 | Add a thin product-led-sales assist gated on the PQL — an in-app “book 15 min”, no SDR team | M | Med | Captures the ~20–30% high-intent accounts without inverting the unit economics of a $99–199 tool |
1. Product-Led Growth (PLG) & the Self-Serve Motion
Frameworks
| Framework | What it says | How it applies to us |
|---|---|---|
| PLG flywheel + PQL (Wes Bush) | Product is the primary acquisition/conversion/expansion vehicle; a PQL is a user who experienced value in-product, not a form-fill | Define the PQL as “storefront live + first real customer order within 14 days.” The existing SetupChecklist already operationalizes the activation half |
| Activation vs PQL (OpenView) | Activation = reached core value; PQL = showed buying intent. Track them as separate events | Activation = checklist 100% (store previewable). PQL = activation plus a real order. Today neither is captured as an event |
| Reverse trial (Verna/Poyar) | Start on the full paid tier, time-boxed; at expiry convert or downgrade to a constrained free state | We already seed trial_days:14. Make the trial start on Pro so owners build habits with white-label/custom-domain, then convert or downgrade |
| Value vs fence gating (Poyar/OpenView) | Value gates cap how much core value you get (scale with success); fences lock peripheral features | max_products/max_stores/max_users are value gates; white_label/custom_domain/api_access are fences. Free keeps the core loop open, fences the growth features |
| PLG growth loops (Reforge) | Self-reinforcing in-product loops (content/collaboration/integration) replace the linear funnel | Our two-level structure creates a content loop: every storefront carries a “Powered by” link. Removing branding is both fence and loop fuel |
| Product-led sales / PQL handoff (ProductLed) | Layer light sales only above an intent threshold; pure self-serve breaks down >~$30K ACV | At $99–199, do not build an SDR team. Add a thin “talk to us” assist triggered only by the PQL |
Benchmarks
| Metric | Value | Confidence |
|---|---|---|
| Freemium free-to-paid (median) | ~4.5% (2–8%) | solid |
| Opt-in no-card trial conversion (median) | ~14% (8–22%) | solid |
| Reverse-trial conversion (SMB self-serve) | ~24% (18–32%) | directional |
| Card-required / opt-out trial conversion | ~44% (35–60%) | solid |
| Best-in-class PLG activation rate | ~33% (band 20–40%) | solid |
| Median NRR (SMB segment, ChartMogul 2024) | ~108% | solid |
| Uplift in odds of fast growth from tracking PQLs | +61% | vendor-blog/weak |
| ACV where pure self-serve breaks down | >~$30K | directional |
What this means for Print-Flow-360
The plumbing is largely built and just needs connecting:
app/Models/Plan.php+app/Models/Subscription.php— JSONfeaturescolumn, status enum{active, trial, expired, charge_failed, cancelled},trial_ends_atlogic. A reverse trial needs no schema change to the status model.app/Services/Subscription/PlanLimitsService.php— enforcesmax_products/max_users/max_storesserver-side, returns plain-language messages,-1 = unlimited. Already wired intoProductController,UserController,StoreController. Critical flaw: it fails open when no plan is found (returnsallowed=true).nuxt/app/composables/usePlanLimits.ts— the client mirror (canUseFeature,checkProductLimit,isTrial,isExpired,hasNoplan, aFEATURE_LABELSmap). It computes reasons but nothing renders an upgrade CTA — this is the natural hook for contextual paywalls.database/seeders/PlanSeeder.php— already a Good-Better-Best ladder: Starter $99, Pro $199 (+ white_label/custom_domain/remove_branding/themes/api/priority_support), Enterprise $0/is_contact_us:true. There is NO permanently-free tier — required as the reverse-trial downgrade target.nuxt/app/components/dashboard/SetupChecklist.vue— tracks logo, payment gateway, first product, shipping, storefront preview. The activation event source; missing only a “first order” step and any analytics emission.nuxt/app/layouts/no-plan.vue— the hard lockout surface. This is the dead-end a reverse-trial downgrade should replace.
Missing: PQL event capture/scoring, in-app trial countdown, contextual paywall component, a free Plan row, a “first order” activation signal, and any sales-assist trigger.
Recommendations
| Move | Effort | Impact | Why |
|---|---|---|---|
| Convert to a reverse trial: start every signup on Pro for 14 days, downgrade (not lock out) at expiry | M | High | Highest-converting model for this profile; replace the no-plan.vue dead-end with a downgrade so an activated-but-unpaid owner keeps a live store and a reason to return |
| Define + instrument the PQL: “store live + first real order within 14 days” | M | High | The most predictive intent signal — it means their customers are transacting. The only honest trigger for any sales assist |
| Turn hard limit-blocks into contextual “value-then-fence” paywalls at the point of friction | M | High | Build one reusable component — new nuxt/app/components/billing/UpgradePrompt.vue — fed by usePlanLimits reasons, rendered inline at each call site that today throws a hard block (ProductController/UserController/StoreController consumers). Value gates soft-warn near the cap; fences show a one-line benefit + Upgrade link. Matches the §0 amber-banner pattern |
| Add a persistent, non-nagging trial-state strip in the admin shell + Action Center | S | High | ”X days left on your Pro trial — what you’ll keep / what you’ll lose.” Render in the admin layout shell (the default authenticated layout that wraps the dashboard, alongside where no-plan.vue is swapped in) so it shows on every page, plus an Action Center widget. Plain-language loss-framing for Maria |
| Keep “Powered by Print-Flow-360” ON for free/trial tiers deliberately | S | Med | Fuels the content loop and creates a genuine upgrade reason. Don’t give branding removal away free |
| Add a thin product-led-sales assist gated on the PQL — no SDR team | M | Med | Capture the ~20–30% high-intent accounts with an in-app “book 15 min”; route multi-store interest to the seeded Enterprise tier |
| Instrument activation rate + time-to-value as first-class metrics off the checklist | S | Med | Leading indicators of the whole funnel; ties to the North Star already chosen in CONVERSION_FUNNEL_RESEARCH |
Risks
- Hard-gating a non-technical owner mid-task reads as “broken,” not “upsell” — violates §0’s “never leave a dead UI.” Paywalls must explain value + offer a path.
- A reverse trial backfires if the downgrade tier is too generous (no urgency, cannibalizes paid) or too stingy (owner loses their live store and leaves angry). Keep the store-data alive but fence the growth features.
- A PQL on a vanity signal (“logged in 3×”) over-counts and wastes sales budget — the only honest PQL is a real order, which depends on the still-incomplete production spine working end-to-end (QUEUE=sync, partial fulfillment, carrier/tracking).
- Any sales motion below ~$30K ACV usually loses money — resist staffing SDRs off PQLs.
- Free-tier abuse/cost — free stores still consume storage, PDF compute, and S3. Tie limits to
PlanLimitsService.
2. Viral Loops & Network Effects (beyond referral programs)
Frameworks
| Framework | What it says | How it applies to us |
|---|---|---|
| Viral coefficient (k) & cycle time | k = invites × conversion-per-invite; k>1 = self-sustaining | Honest denominator is shop owners, not end-customers. Inherent k among shops ≈ 0; the customer-of-customer path gives a realistic k ≈ 0.15–0.4. Target sub-viral amplification, not virality |
| Three loop types (inherent / referral / word-of-mouth) | Inherent virality exposes others as a byproduct (Hotmail footer); cheaper than referral, no behavior change | The referral program is specced elsewhere (greenfield). This research owns the inherent-virality layer — currently 0% built |
| NFX network-effects taxonomy | Distinguishes true network effects from one-way distribution loops | We have no genuine network effect today: one shop joining doesn’t make us more valuable to another. A cross-tenant template marketplace would be the first real one |
| Powered-by attribution loop (Hotmail/Shopify/Calendly) | Embed a clickable brand signal in user-facing output; pair with upgrade-to-remove | Directly actionable and 0% present. Closest analog: Shopify’s footer ↔ our branded storefronts; Calendly’s booking page ↔ our public proof-review page |
| Customer-of-customer as a channel | In two-level structures the customer’s customers are a built-in free audience | Every print shop’s end-customer is a free impression. Realistic target isn’t “shopper becomes shop owner” but “business-minded viewer notices us” — concentrate on the proof page + invoices |
Benchmarks
| Metric | Value | Confidence |
|---|---|---|
| Average k-factor, B2B SaaS | ~0.2 (0.3–0.7 = strong) | directional |
| Share of B2B SaaS reaching k>1 | Vanishingly rare (needs collaboration to be foundational) | directional |
| Realistic k for our attribution loops | ~0.15–0.4 (sub-viral) | directional |
| Referral CAC vs others (from PRICING_RETENTION doc) | referral ~$150, inbound ~$200, SEO $480–942, paid search ~$802 | solid |
| Best-in-class referral share of new customers | 20–30% | solid |
| Dropbox K-factor (had built-in shared-folder virality we lack) | 1.5–2.0 | vendor-blog/weak |
| SMB self-serve monthly churn | 3–5%/mo | solid |
What this means for Print-Flow-360
The verified state of play is stark: zero “Powered by” attribution anywhere customer-facing. A grep across frontstore/, designer/src, and nuxt/ found only build artifacts.
StorefrontFooterDynamic.vue— fully tenant-controlled, no platform-owned slot. The prime Shopify-analog surface; needs a non-removable-by-default platform footer line below tenant content.app/Mail/Proof/ApprovalMail.php— the proof link (/proofs/client-review/{id}) is a public, guest-accessible page sent to every end-customer. A business-context viewer (someone reviewing a print job) is a far better prospect than a random shopper. Zero attribution today.designer/src/components/FileExport/*— exports (PDF/SVG/PNG) carry no watermark; shared designs leave the platform invisibly (a wasted Canva-style loop).app/Services/ProofService.php— stamps the tenant logo on proof PDFs (confirming it’s a branding surface), but never Print-Flow-360.- Internal-only “network effects”:
max_stores:3and B2B multi-location create within-account lock-in but do not cross tenants. The “My Designs” library (frontstore/app/pages/profile/designs.vue) is a retention surface, not a shareable template gallery.
Recommendations
| Move | Effort | Impact | Why |
|---|---|---|---|
| Ship “Powered by Print-Flow-360 · Build your own store →” on every storefront footer | S | High | Highest-volume, lowest-friction surface; UTM-tagged, off only when remove_branding=true. Render it as a platform-owned line below StorefrontFooterDynamic.vue’s tenant markup — not inside it — with its styles scoped/inlined so per-theme CSS can’t hide or restyle it (the codebase has a known pattern of theme !important rules hijacking shared components; see the Aurora-sweep memory). Verify the badge survives all three themes at 375px |
| Brand the public proof-review page + proof/order/invoice emails | S | Med | Reaches a business-context viewer; reuses existing email infra. Keep it a single subtle line, never a banner |
| Add an optional small platform credit to designer exports, removable on paid plans | M | Med | Turns shared designs into impressions; removal becomes an upgrade trigger |
Wire remove_branding/white_label into the Plan features JSON; gate all three surfaces on it | S | High | One flag monetizes the loop and is the exact Shopify/Wix “remove the badge” upgrade playbook |
| Build a cross-tenant shared template gallery/marketplace (longer-term) | L | High | The only genuine network-effect play; start curated to avoid cold-start, then open contribution |
| Track attributed signups per surface; report a per-surface k, never a blended viral number | S | Med | UTM-tag footer vs proof vs export distinctly; set internal expectation of k ≈ 0.15–0.4 |
Risks
- Conflating distribution loops with network effects is dishonest — only the template marketplace creates a real moat. Don’t pitch “k>1” for the attribution loops.
- Branding annoys owners who want their store to look like their brand — must be tasteful, small, removable via upgrade, and never injected into checkout/cart trust moments.
- Customer-of-customer conversion is structurally weak — it’s amplification, not a channel. Don’t let it starve the proven founder-led/community + BOFU-SEO motion.
- Template marketplace cold-start + IP/moderation risk — an empty gallery violates §0; seed curated content and add review before opening contribution.
- Per CLAUDE.md, the
remove_brandingflag needs a round-trip test or it joins the silent-lie class (owner toggles white-label, nothing happens).
3. Activation & Aha-Moment Engineering (“store live + first order in 7 days”)
Frameworks
| Framework | What it says | How it applies to us |
|---|---|---|
| Setup / Aha / Habit model | Setup = first config; Aha = the emotional “this is valuable” realization; Habit = repeated value | Our checklist engineers only Setup. Aha = a real order lands in admin. Habit = recurring orders via the Action Center. The aha and habit stages are 0% engineered |
| Empirical activation methodology (Lenny/Amplitude) | Candidate moments → regress vs retention → confirm causally. Facebook’s “7 friends” was largely correlational | We can’t run this yet (no events). “Store published + first order within 7 days” is a strong hypothesis — validate once instrumented; don’t hard-code 7 days as causal |
| Fogg B=MAP | Behavior = Motivation × Ability × Prompt; for low-ability users, cut friction beats adding motivation | Maria is high-motivation, low-ability. The codebase already cuts friction (seeded offline payment + sample catalog). Remaining cliffs: pricing-rule config + the gateway step — make both skippable |
| AARRR activation + PQL behavioral trigger | The activation event doubles as the PQL signal to time card-capture at peak willingness-to-pay | Requires first_order_received to be an observable event — today it is not. This is the prerequisite for the reverse-trial card-capture |
Benchmarks
| Metric | Value | Confidence |
|---|---|---|
| Median B2B SaaS activation rate | ~37% | solid |
| Avg onboarding-checklist completion | 19.2% avg, 10.1% median | solid |
| Target time-to-value (median) | ~1.5 days; winners <10 min | directional |
| Early-activation retention multiplier | ~3–5× more likely to retain at D30 | vendor-blog/weak |
| Value-less user churn | >98% who never reach value churn within 2 weeks | vendor-blog/weak |
| Pre-loaded sample data effect | Roughly doubled activation in cited cases | vendor-blog/weak |
What this means for Print-Flow-360
SetupChecklist.vue+useSetupChecklist.tsare built and mounted (dashboard line ~30) with a progress bar, plain-language labels, completion banner. But it’slocalStorage-only (pf360_setup_checklist_v1) and stops at “preview storefront” — it engineers Setup but never the Aha.CONVERSION_FUNNEL_RESEARCH§4 explicitly calls for a missing “place a test order → You made your first sale!” step.- No activation event taxonomy.
app/Helpers/AuditLogger.phplogs admin mutations + login, but there’s nosignup → first_product → store_published → first_orderstream — the doc’s #1 build item. - The “demo store” is shallow.
StoreOnboardingService::seedSampleCatalog()seeds 4is_demoproducts + 11 CMS blocks + published system pages + a welcome coupon + shipping + an offline Cheque/Bank-Transfer payment method. So a store can take an order on day one with zero config — but this is never surfaced to the owner. - Friction-removal already done but underused: because payment + shipping are seeded, those checklist steps are technically already satisfiable, yet the checklist still shows them as to-dos, adding false friction.
- No activation columns on
Store(nofirst_order_at/activated_at). Celebration plumbing exists (showSuccess, the completion-banner pattern) — no new UI infra needed.
Recommendations
| Move | Effort | Impact | Why |
|---|---|---|---|
| Instrument the activation funnel server-side FIRST (events + persisted milestones) | M | High | Activation can’t be engineered or A/B-tested while it’s in localStorage. The doc’s explicit #1 item; unblocks everything downstream |
| Extend the checklist to the aha: add “place a test order” + a real first-order celebration | M | High | Today owners never reach the value moment. Distinguish test vs real order; fire full-screen “You made your first sale!” on the real first order |
| Reframe pre-seeded defaults as “You can already take orders” | S | High | A store is live on day one. Mark payment/shipping “Done (using defaults — upgrade later)” so Maria isn’t falsely blocked. Friction removal > motivation |
Replace shallow is_demo products with a guided “add my first real product” wizard | M | Med | Pre-loaded realistic data ~doubles activation; let the owner make their first product personally theirs in <2 min; keep demo data one-click removable |
| Make the checklist server-truth + cross-device, ≤5 interactive steps, progressive disclosure | S | Med | Completion averages only 19.2%; short server-backed checklists complete best and survive device changes |
Wire the post-aha PQL trigger: prompt card-capture right after first_order_received | S | High | Behavioral triggers convert ~2.5× calendar triggers; willingness-to-pay peaks at the aha |
| Validate the activation metric empirically once data accrues; don’t hard-code “7 days” | S | Med | Regress days-to-first-order vs D30/D90 retention to find the real inflection |
| Connect the Action Center as the post-activation Habit hook | M | Med | Activation must hand off to a habit loop or the tenant churns |
Risks
- Vanity-milestone trap — a green checklist with zero orders is a false-positive activation signal (the Facebook “7 friends” correlation-vs-causation pitfall).
localStorage-only state silently lies across devices — an owner can look “un-onboarded” on a second device (violates the silent-lie rule).- Over-celebration / fake success — firing “first sale!” on the owner’s own test order cheapens it; distinguish test vs real.
- Sample products leaking into a live store confuse the owner’s real customers; need a clear “remove demo data” affordance.
- QUEUE=sync means naively firing analytics inline adds latency to order creation — fire async/queued.
4. Growth-Experiment Discipline for a low-traffic, pre-PMF SMB tool
The honest constraint: we cannot run classic A/B tests on the core funnel. A 2-week test on a 5% baseline detecting a 20% lift needs ~25,000+ users per arm — months at our scale. The right discipline for a tiny team is experiment velocity using cheap, low-traffic-tolerant methods, not more A/Bs.
Frameworks
| Framework | What it says | How it applies to us |
|---|---|---|
| ICE (Sean Ellis) | Impact × Confidence × Ease (1–10); omits Reach deliberately | Primary backlog tool. Reach is meaningless pre-PMF, so ICE beats RICE |
| RICE (Intercom) | (Reach × Impact × Confidence) / Effort | Explicitly NOT recommended here — Reach is guesswork at our traffic. Keep for later-stage roadmap only |
| PIE (Goward, CRO-specific) | Potential + Importance + Ease (1–10) | Use only for landing-page + pricing-page CRO — the only true-A/B-capable surfaces |
| North Star + input-metric tree | One value metric, decomposed into a small MECE tree the team can move | NSM already chosen: “store live + first order within 7 days.” Inputs: signups → checklist completion → store published → first order → days-to-first-order |
| Guardrail metrics | Counter-metrics that must not regress while optimizing | trial-to-paid %, refund/chargeback rate, support tickets, churn, CLS, money-path error rate |
| Painted-door / fake-door / Wizard-of-Oz | Surface a CTA for an unbuilt feature; measure intent. Tolerates low traffic | The core technique for us. Test referral, new CMS blocks, B2B-approvals interest before coding |
| Sequential / always-valid p-values | Lets a small team peek + stop early without inflating false positives | Use a sequential/Bayesian tool (GrowthBook/Statsig) when we do run a true A/B |
| Experiment-velocity culture (Kohavi) | Only ~1/3 of well-designed experiments win; throughput compounds, not hit-rate | Optimize for cheap experiments per month, not win-rate. Expect most bets to fail; kill fast |
Benchmarks
| Metric | Value | Confidence |
|---|---|---|
| Share of well-designed experiments that improve the key metric | ~1/3 positive, 1/3 flat, 1/3 negative | solid |
| Experiment failure rate at elite orgs | ~70% (MS) to ~90% (Google/Netflix/Airbnb) | directional |
| Annual experiment volume at mature orgs | >10,000/year each | solid |
| Min traffic for a fixed-horizon A/B (5% baseline, 20% lift) | ~25,000+ users per arm | solid |
| ICE/PIE scoring scale | each dimension 1–10 | verified |
What this means for Print-Flow-360
- Storefront customer analytics EXIST (
useGtm.ts+gtm.ts: GA4 e-commerce events) — but they measure the end-customer funnel inside a tenant store. That’s the wrong level for our growth experiments. - Tenant business reporting EXISTS (
StoreAnalyticsService.php) — KPIs for the shop owner, not experiment/event analytics. store_analytics_eventsexists but is thin — its only writer is a generic tenant recorder; nothing emits structured activation events.- No landlord-side funnel/cohort analytics, no feature-flag library. Grep found zero Pennant/GrowthBook/Statsig/PostHog/Amplitude/Segment. The raw data to reconstruct the funnel exists in
Subscription/Tenant/Order; only the aggregation layer is missing. config/pdf_service.phpuses simple on/off flags — a usable kill-switch pattern, not variant assignment.
Recommendations
| Move | Effort | Impact | Why |
|---|---|---|---|
| Instrument the funnel BEFORE any experiment (signup, store-published, checklist-step, first-order, trial, upgrade, cancel) | M | High | You can’t experiment on a funnel you can’t measure |
| Adopt the existing North Star; publish a one-screen input-tree + guardrails dashboard | S | High | Reuse the prior decision; give the weekly meeting a target and a “do no harm” floor |
| Run a 30-min weekly growth meeting off an ICE backlog; default to fake-door/qualitative, reserve true A/B for landing + pricing | S | High | Velocity is the lever; we’re too low-traffic for funnel A/B |
| Seed the backlog with concrete experiments (fake-door referral tile; 5 onboarding interviews; PIE landing-headline A/B; cohort upgrade-timing test) | M | High | Turns discipline into a runnable list grounded in the prior docs |
| Use a sequential/Bayesian tool when a true A/B is justified; never peek on fixed-horizon | M | Med | Avoids the dominant low-traffic statistical error |
| Add a kill-switch feature-flag layer (Laravel Pennant or extend the config-flag pattern) | M | Med | Enables painted-door tests + gradual rollout per §0 “no half-built UI” |
Risks
- Peeking / optional stopping inflates false positives — the dominant low-traffic failure. Use sequential methods or strict pre-registered horizons.
- Underpowered tests declared “no difference” — run a power calc first; if infeasible, don’t A/B.
- Fake-door backlash — showing Maria a feature that 404s violates §0. Always capture intent gracefully (“want early access?”).
- Optimizing a local metric that hurts a global one — guardrails are mandatory.
- Vanity-metric drift — GTM events measure the tenant’s customers, not our funnel. Keep the two levels strictly separate.
5. Freemium Economics — when free helps vs cannibalizes
The verdict: for a low-ACV, high-churn SMB vertical SaaS where every live store consumes real per-tenant infra (always-on hosting + per-tenant S3 + PDF/image compute), a permanent open free tier is the wrong default. Freemium converts 2–5%, so 20–50 free users must be carried per payer — and unlike Slack/Dropbox/Canva, our free users are not a viral channel (their shoppers don’t become our customers) and our cost-to-serve is high.
Frameworks
| Framework | What it says | How it applies to us |
|---|---|---|
| Freemium vs Trial vs Reverse-Trial | Freemium needs massive scale + near-zero marginal cost; reverse trial = full premium then auto-downgrade | High per-tenant cost + modest scale disqualifies pure freemium. Reverse trial is the right primary motion |
| Cannibalization / fence design | Free that’s “good enough” lowers conversion; the fence forces an upgrade to capture scaling value | The fence = the live, publicly-hosted, order-taking storefront. That’s the whole reason to pay |
| Give away the habit, charge for the scale | Free = the daily habit + lock-in; charge for the revenue-scaling layer | Habit layer (keep alive even when dormant): design studio + “My Designs” + catalog + pricing rules. Scale layer (charge): the store that actually sells |
| Gating taxonomy (feature/usage/seat/time) | Best practice: gate on the value metric (Slack’s 10k-message history) | Hybrid: time-limit = 14-day reverse trial; paid fence = live-storefront gate + usage caps (orders/mo, storage). Avoid gating the designer — that kills the habit |
| Cost-to-serve / gross-margin gate | Freemium works only if payers cover the COGS of carrying free users | Our free user is unusually expensive (always-on store + S3 + PDF). The strongest argument for taking the dormant store offline — a dark store costs ~zero |
| Strategic purposes of free | Justified only if it serves acquisition / virality / network-seeding / education / sales-assist | Virality + network effects do not apply here. Free’s only defensible job is keeping a churned/hesitant owner warm — which a dormant tier does without the live-hosting cost |
Benchmarks
| Metric | Value | Confidence |
|---|---|---|
| Freemium free-to-paid (good / great) | ~2–5% / ~6–8% | solid |
| Free-trial free-to-paid (blended) | ~8–12% good / ~15–25% great | directional |
| Reverse-trial conversion | ~7–11% good / ~14–21% great | directional |
Reconciling the two reverse-trial figures in this doc. §1 cites ~24% (18–32%) and this table cites ~7–21%. They are not the same measurement: the higher band is self-reported, best-profile SMB self-serve (the optimistic vendor case); the lower band is the narrower 2026 blended data across all reverse-trial products. Plan against the conservative ~8–15%, treat ~24% as the ceiling if everything goes right, and — as everywhere in this doc — test it on our own funnel before trusting either. | Free users carried per payer at 2–5% conversion | ~20–50 | solid | | SMB self-serve monthly logo churn | ~3–5%/mo (22–39% annual) | solid | | NRR ceiling at ARPA <$10/mo | only ~2.7% of companies exceed 100% NRR | solid |
What this means for Print-Flow-360
PlanLimitsServicefails OPEN with no plan (['allowed'=>true], “graceful degradation”). A reverse-trial downgrade to a no-plan state would be silently unlimited — the opposite of a capped free tier. A real free tier must be a real Plan row with explicit caps.Subscription::isActiveForAccess()returns false the moment a trial expires — there is no downgrade-to-free path today. Trial expiry = total loss of access. The “permanent free tier” the prior docs assume is unbuilt.AccountStatusService::approveTrial()seeds a 15-day trial (prior docs want 14); no auto-transition-to-free on expiry.- The
planstable is bare — no structured limit columns (orders/mo, storage); thefeaturesJSON is the only place limits live and only 3 keys are read. - Cost-to-serve is real and per-tenant:
FileHelper.phpconfigures per-tenant S3;pdf-service/(sharp + PDFKit + BullMQ) does heavy work; the storefront is always-on. A live free store consumes all three; a dormant store consumes only idle storage — the technical basis for “data-alive, store-offline.” - No referral/network machinery exists — removing the “free seeds virality” justification entirely.
Recommendations
| Move | Effort | Impact | Why |
|---|---|---|---|
| Keep the no-card 14-day reverse trial as the primary motion; do NOT ship open freemium | S | High | Freemium converts 2–5% and forces us to carry 20–50 expensive free users per payer with no viral upside. Align AccountStatusService::approveTrial() 15→14 days — this touches a trial/money path, so per CLAUDE.md it needs a round-trip test asserting the new trial length persists and isActiveForAccess flips on the right day |
| Build the post-trial fallback as a DORMANT/PARKING tier: data alive, storefront offline | L | High | Cannibalization fence + cost fix in one move. Mechanism, grounded: add a value to the subscription status enum (app/Models/Subscription.php — currently {active, trial, expired, charge_failed, cancelled}; needs a Postgres CHECK-constraint widen per §6 of CLAUDE.md) e.g. dormant; isActiveForAccess() treats it as admin-read-only; and the public storefront goes dark via a tenant-level storefront_suspended flag checked in the InitializeTenancy middleware chain (or the storefront’s tenant-resolution guard), returning a branded “store is offline” page rather than 404. Both halves need tests |
| Make the paid fence = “live, publicly-hosted, order-taking storefront” and gate it explicitly | M | High | Give away the habit (designer + catalog), charge for the scale. Reuse the PlanLimitsService 403-message pattern for a “publish storefront” + “accept order” gate |
| Fix the fail-open hole before any free tier ships | M | High | A naive free tier would be silently unlimited — the silent-lie/cost-leak bug class. Seed an explicit Free plan with caps; add orders_per_month/storage_mb columns |
| Use a hybrid gate reconciled with Good-Better-Best (Free=offline; Solo=1 live store + caps; Growing(hero)=higher caps + designer + B2B-lite; Multi-Location=per-location) | S | Med | Time-limit + live-storefront fence + usage caps. Charge per location, never per seat or GMV |
| Wire loss-aversion reactivation + dunning off the dormant tier | M | Med | ”Your store is offline — your designs are saved. Go live in one click.” The cheapest re-conversion path |
| Instrument cost-to-serve per tenant before claiming any LTV:CAC | M | Med | Free’s payback depends entirely on free-user COGS, which is high here. Margin-adjust LTV |
Risks
- Open free live storefront cannibalizes the cheapest paid plan — a low-volume one-person shop gets everything for free; the classic “gave away the farm.”
- Fail-open enforcement means a sloppy rollout grants unlimited products/users/stores — a §0 silent-lie + unbounded cost leak.
- Cost-to-serve runaway — the 95–97% who never pay still burn hosting + storage + compute.
- Over-gating the design studio kills the habit/lock-in — gate the storefront/orders, not the designer.
- Taking a previously-live store offline reads as punitive if handled badly — frame in plain language with one-click reactivation and clearly preserved data.
- Assuming freemium “seeds virality” here is a category error — a free shop’s shoppers never become our tenants.
6. The “Powered by Print-Flow-360” Attribution Loop — our most distinctive growth surface (with an honest reality check)
This is the single most distinctive thing about our growth model, so it deserves its own section — and a dose of honesty.
The opportunity is genuinely unusual. We operate, for free, the highest-traffic acquisition surfaces a web-to-print SaaS could want:
- Every branded tenant storefront — pageviews from the shop’s own marketing, SEO, and repeat customers (Shopify’s footer analog).
- Every public proof-approval link (
/proofs/client-review/{id}) — sent to every end-customer, landing a business-context viewer reviewing a print job (Calendly’s booking-page analog). This is the best prospect of the three: someone already thinking about print production. - Every designer export (PDF/SVG/PNG) — shared designs that currently leave the platform invisibly (Canva’s UGC analog).
Today, 100% of this exposure is wasted — zero attribution on any of it. Adding “Powered by Print-Flow-360 · Build your own store →” to these surfaces, gated by a single remove_branding plan flag, is the highest-ROI growth move in this entire document: free impressions on surfaces we already run, requiring zero behavior change from the shop owner, that simultaneously create a clean, plain-language upgrade reason (“remove the badge — go Pro”).
The honest k-coefficient reality check
Do not let anyone pitch this as a “viral engine” or a “network-effect moat.” Here is the unvarnished math:
- The honest denominator is shop owners, not end-customers. A shop owner rarely inherently invites another shop owner, so inherent k among shops ≈ 0.
- The customer-of-customer path (an end-customer sees the badge → becomes a shop owner) has a tiny conversion rate — most end-customers are buyers of printed goods, not aspiring print-shop owners.
- Cycle time is long — someone has to decide to open a print shop.
- Net realistic k ≈ 0.15–0.4: sub-viral amplification, not self-sustaining growth. It lowers blended CAC and compounds the founder-led/referral motion — it does not replace a channel.
The one-line back-of-envelope (so this is a defensible estimate, not a vibe — all inputs are guesses to replace with measured UTM data): say an active shop’s badge earns ~500 badge-impressions/month across its storefront + proof links; assume ~1% click through to our landing page (5 visitors), and that ~1–2% of those business-context viewers are plausible shop-owner prospects who eventually sign up (~0.05–0.1 signups/shop/month). Over a ~6-month consideration window that’s ~0.3–0.6 referred signups per active shop — i.e. k ≈ 0.15–0.4 once you discount for long cycle time. Every term here is a placeholder; the point is the shape (well under 1), not the digits. Measure impressions→clicks→signups per surface and recompute.
And critically: this is a one-way distribution loop, not a network effect. One shop displaying our badge does not make the product more valuable to another shop. The only move that would create a genuine two-sided/data network effect is a cross-tenant template marketplace (more shops → more shared templates → more valuable to all shops). That’s the long-term moat; the badge is the cheap, immediate amplifier. Label them honestly to stakeholders or you’ll set false expectations.
Where to concentrate the badge: the proof page and invoices (business-context viewers) over the storefront footer (random shoppers) for conversion quality — but ship the footer first for volume. UTM-tag each surface distinctly so you measure which loop actually converts shop owners, and report a per-surface k, never a blended viral number.
7. How this fits the existing GTM docs
This doc extends four existing research docs. Read them; don’t re-derive them.
| Doc | Where we AGREE | Where we EXTEND or DISAGREE |
|---|---|---|
readme/CONVERSION_FUNNEL_RESEARCH_2026-06-15.md | No-card 14-day reverse trial; North Star “store live + first order in 7d”; ≤5-step go-live checklist; funnel instrumentation as #1 build item | EXTEND: the doc’s “permanent thin free tier” fallback should be reframed as a dormant/parking tier (storefront offline), not an open freemium — because free live stores cannibalize the cheapest plan and burn infra. EXTEND: the checklist must reach the aha (first real order + celebration), not stop at “preview.” EXTEND: “7 days” is a hypothesis to validate empirically, not a causal constant |
readme/PRICING_RETENTION_REFERRALS_STRATEGY_2026-06-15.md | Good-Better-Best with a hero middle tier; dunning; per-location pricing; referral as a future build | EXTEND: before building the referral system, ship the cheaper inherent-virality (Powered-by) loop — it needs zero owner behavior change and is complementary. EXTEND: wire remove_branding as the fence that monetizes both the loop and the Pro tier |
readme/ACQUISITION_CHANNELS_2026-06-15.md | Founder-led outreach + print communities primary; BOFU SEO + free design tool secondary | AGREE & bound: the customer-of-customer attribution loop is amplification, not a channel — it feeds the proven motion, never replaces it. Don’t let it starve founder-led outreach |
readme/GTM_01_ICP_AND_POSITIONING_2026-06-15.md | ICP = “Maria,” non-technical SMB print-shop owner, one-person buying committee; category = online storefront + design studio | EXTEND: every PLG decision here (reverse trial, value-then-fence paywalls, plain-language trial countdown, dormant-tier framing) is downstream of “Maria is high-motivation, low-ability” — friction removal beats persuasion at every step |
8. Sequencing — Phase 0 → 1 → 2 (with hard gates)
The TL;DR is ordered by impact ÷ effort, not by dependency. Shipping in raw impact order is dangerous here — e.g. dropping a reverse-trial downgrade into the current fail-open PlanLimitsService would silently grant unlimited usage, and calling “first real order” a PQL is dishonest if the production spine can’t fulfill that order. Build in this order; each gate is a “do not ship the next phase until this is true.”
| Phase | Ship | Gate that must hold first |
|---|---|---|
| Phase 0 — Foundations (no growth promises yet) | (a) Server-side activation events + landlord funnel/cohort dashboard (§10 #1–2); (b) fix the PlanLimitsService fail-open hole — no-plan must mean capped, not unlimited; (c) confirm landlord-side subscription billing + card-on-file works end-to-end; (d) per-tenant cost-to-serve instrumentation | Nothing. This is the prerequisite layer. |
| Phase 1 — Cheap amplifiers + activation (no model change) | (a) “Powered by Print-Flow-360” footer + remove_branding flag (with its round-trip test); (b) extend checklist to the real aha + first-order celebration; (c) reframe seeded defaults as “you can already take orders”; (d) contextual UpgradePrompt replacing hard blocks; (e) weekly ICE growth meeting | Phase 0 done: events exist (so activation/aha is measurable), and the fail-open hole is closed (so gating is safe). |
| Phase 2 — Model change + structural moats | (a) Convert to the no-card 14-day reverse trial (15→14, with test); (b) build the dormant/parking tier as the downgrade target; (c) wire PQL card-capture on first_order_received; (d) thin product-led-sales assist; (e) (longer-term) cross-tenant template marketplace | Phase 0(b) + 0(c) done: the reverse-trial downgrade requires a real capped Free/dormant plan (never the fail-open absence of a plan) and working subscription billing. PQL honesty gate: do not call “first real order” a PQL until the production spine (QUEUE=sync, partial fulfillment, carrier/tracking) can actually fulfill it end-to-end. |
The three gates, stated bluntly:
- Instrument before optimize — no activation/experiment work lands before Phase 0 events exist.
- Close fail-open before any free/dormant tier — or the tier is silently unlimited (a §0 silent-lie + unbounded cost leak).
- Working fulfillment + working billing before the order-based PQL — or we’re triggering an upsell on a value moment we can’t deliver or can’t charge for.
9. Sources
Internal (build on, do not duplicate):
readme/CONVERSION_FUNNEL_RESEARCH_2026-06-15.mdreadme/PRICING_RETENTION_REFERRALS_STRATEGY_2026-06-15.mdreadme/ACQUISITION_CHANNELS_2026-06-15.mdreadme/GTM_01_ICP_AND_POSITIONING_2026-06-15.mdreadme/ACTION_CENTER.md,readme/ONBOARDING.md,readme/B2B_MODULE.md
Codebase grounding:
app/Models/Plan.php,app/Models/Subscription.php,database/seeders/PlanSeeder.phpapp/Services/Subscription/PlanLimitsService.php,app/Services/Admin/AccountStatusService.phpnuxt/app/composables/usePlanLimits.ts,nuxt/app/composables/useSetupChecklist.ts,nuxt/app/components/dashboard/SetupChecklist.vue,nuxt/app/layouts/no-plan.vueapp/Services/Onboarding/StoreOnboardingService.php,app/Helpers/AuditLogger.php,app/Helpers/FileHelper.phpfrontstore/app/components/storefront/layouts/StorefrontFooterDynamic.vue,frontstore/app/pages/profile/designs.vueapp/Mail/Proof/ApprovalMail.php,app/Services/ProofService.php,designer/src/components/FileExport/*frontstore/app/composables/storefront/useGtm.ts,app/Services/Analytics/StoreAnalyticsService.php,app/Models/Store/StoreAnalyticsEvent.php,config/pdf_service.php
External (treat numbers as hypotheses):
- Wes Bush, Product-Led Growth — productled.com/book/product-led-growth
- OpenView — activation vs PQL, +61% fast-growth correlation; value-vs-fence gating (Kyle Poyar)
- Elena Verna / Kyle Poyar — Reverse Trial (amplitude.com/blog/reverse-trial; growthunhinged.com)
- Reforge — PLG activation loops; “The Hidden Freemium Advantage”
- NFX — network-effects taxonomy (nfx.com/post/network-effects-manual)
- ChartMogul — SaaS Conversion Report; First Page Sage / Userpilot / ADV.me — trial & freemium benchmarks
- Lenny’s Newsletter — activation methodology; “why SaaS freemium playbooks don’t work in AI”
- Mode / Geckoboard — Facebook “7 friends” correlation-vs-causation
- Ronny Kohavi — experiment win-rates & volume (abtasty 1000-Experiments-Club; bytepawn)
- ICE/PIE/RICE — productlift.dev, statsig.com, weblics.agency; fake-door — userpilot.com, amplitude.com
- Low-traffic A/B — absmartly, vwo, convert.com; sequential — growthbook.io
- Cost-to-serve / unit economics — CloudZero, Monetizely “hidden costs of freemium”
10. Open Questions / What to Instrument FIRST
Instrument first (in order) — nothing below the line can be optimized until these exist:
- Server-side activation events —
signup,first_product_added,store_published,first_order_received,trial_started,upgraded,cancelled. Fire async (QUEUE=sync caveat). Persist milestone timestamps onStore(or atenant_onboarding_progressrecord). - A landlord-side funnel + cohort dashboard — signups → checklist completion → store published → first order → days-to-first-order, with the guardrail metrics overlaid. Events without a dashboard are dead weight.
- Per-tenant cost-to-serve — storage GB, PDF/image job volume, hosting cost — so any LTV:CAC or free-tier-payback claim is margin-adjusted and honest.
Current baseline: unmeasured. We do not today have signup, activation, trial-conversion, or churn numbers for our own (landlord-side) funnel — there is no instrumentation (see §4). Every impact estimate, conversion %, and k-factor in this doc is therefore a directional target, not a delta off a known baseline. The first job is to make the baseline real; only then are “lift” claims meaningful.
Hard dependency to verify before promising behavioral card-capture (#10 in the TL;DR):
- Does landlord-side subscription billing + card-on-file actually work? The reverse-trial PQL move assumes we can capture a card and charge the tenant for their Print-Flow-360 subscription (this is the SaaS billing, distinct from the shop’s own customers paying for prints, which run through the per-tenant payment gateways). Memory flags severe money-path bugs in the customer-side gateways; the landlord-side subscription-charge path must be confirmed end-to-end (card capture → store → recurring charge → dunning) before we trigger card-capture on
first_order_received, or we promise a conversion we can’t collect.
Open questions to resolve with data, not opinion:
- What is the real activation inflection? Is it “first order within 7 days,” 72 hours, or “3 products published”? Regress against D30/D90 tenant retention once events accrue. Don’t hard-code 7 days as causal.
- What is the actual attribution-loop k? UTM-tag footer vs proof vs export; measure signups attributed to each. Confirm or revise the 0.15–0.4 estimate.
- Does the dormant tier reactivate anyone? Track reactivation rate from offline-store dunning before investing further in it.
- What’s the right dormant-tier generosity? Too generous → cannibalizes Solo; too stingy → angry churn. A/B the offline-vs-read-only framing once volume allows; until then, run 5 qualitative interviews.
- Does the production spine actually close the loop? A PQL on “first real order” is only honest if order → fulfillment → tracking works end-to-end (known gap: QUEUE=sync, partial fulfillment, carrier/tracking). Verify before trusting the PQL.
- Will the template marketplace clear cold-start? Decide the curated-seed strategy and moderation/IP policy before committing to the only genuine network-effect play.