OverSkill · Growth Architecture · Internal Strategy

Three growth systems.
One graph.

Referrals, invite codes, and affiliates aren't three products to build and maintain. They're one relationship—"who brought whom"—paying out in different currencies as we grow. Here's why that's the right architecture, and how it makes our beta advocates wake up already monetized on launch day.

2,781
waitlist signups today
1 graph
not 3 systems
2 phases
beta → public, one flip
FEASIBLE
Whop API confirmed
The reframe

We were about to build the same thing three times.

"Refer a friend to climb the line," "share an invite code to let someone skip," and "earn commission for sending a paying customer" feel like three features. They share one verb: I brought you here. Build that bond once—then let the reward change with the season.

❌ The naive build

Separate referral table, separate invite-code logic, separate affiliate attribution. Three schemas that drift, three places to debug, and a beta advocate who has to re-sign-up as an affiliate at launch—losing the relationship they already earned.

✅ The unified build RIGHT

One attribution edge—"X brought Y"—written once at signup. Referral link or invite-code redemption both write the same edge. A payout layer reads that graph and changes by phase. Three concepts, one system.

🧠 Why it matters

Superhuman's own data: a waitlisted user converts to paid at ~3% vs ~10% live—and that halves every year they wait. The system's real job isn't a giant list. It's identify your hottest advocates and fast-track them in, monetizing the relationship for the long haul.

One graph, three currencies

The rule that keeps it straight

Referral = a position.  Invite = a seat.  Affiliate = a dollar. Same edge in the graph. Different payout, depending on what we have to give at that moment.

🔗

Referrals

Climb the waitlist. The viral growth engine while we're gated. Costs us a seat we were releasing anyway.

Currency · Queue position + milestone unlocks
🎟️

Invite codes

Skip the line entirely. The scarce accelerant—you get a few to hand trusted friends instant entry. Clubhouse mechanic.

Currency · A seat (instant access)
💰

Affiliates

Earn cash. Only meaningful once there's something to buy. Same person, same sharing—now there's revenue to share.

Currency · Whop commission on a sale

Critical: every "I brought you" event—referral link OR invite-code redemption—writes the same parent-affiliate edge. The friend you fast-track becomes your downline. Forever.

The whole strategy in one line

The currency upgrades at every stage

A single person's sharing behavior never changes. What it earns upgrades from position → seat → dollars as they move through the funnel and as we move from gated to public.

01

Climb

Shares referral link, jumps the queue with score-ranked position.

02

Skip

Spends scarce invite codes to fast-track their best friends past the line.

03

Activate

Gets in, builds a real app. Rewards vest on activation—fraud earns nothing.

04

Monetize

At public launch, their beta downline auto-converts to paid affiliate commission.

position seat a built app dollars   ·   same edge, same graph, the whole way
Beta → public, one flip

The same graph, read two ways

We don't rebuild at launch. The payout layer reads the existing edges and switches what it pays. Public launch is a flag flip, not a migration.

Phase 1 — Prelaunch beta

WE ARE HERE
  • Referrals Primary growth engine — score-ranked queue + milestone ladder + honest waves.
  • Invites Scarce accelerant — hand a friend instant access; still counts as YOUR referral.
  • Affiliates Dormant — nothing to buy yet, so no commission to pay.
  • Payout Queue position + credits + free Pro months, all vesting on activation.

Phase 2 — Public launch

THE FLIP
  • Referrals Become 2-sided credit referrals (Dropbox model) — same share habit, new currency.
  • Invites Mostly retire (no line to skip) — survive optionally as "gift credits."
  • Affiliates Turn ON — beta downline edges auto-route Whop commission on real sales.
  • Payout Cash. Because the edge was recorded back in beta, no re-signup needed.
The payoff
A beta user who refers 30 people and burns 5 invite codes is building a 35-person affiliate downline right now. The day OverSkill goes paid, every one of them who subscribes pays that user commission—automatically. No re-signup. No "become an affiliate" step.

Our best beta advocates wake up on launch day already monetized.
War-gamed · the load-bearing question

"But can Whop actually do this?"

The whole vision lived or died on one feasibility question: can we set a parent affiliate at beta signup and have commission auto-route to them on a purchase months later—after any cookie is long gone? We attacked it adversarially. The answer:

✓ FEASIBLE — Whop's API is the reason it works, not the reason it fails

Whop exposes createAffiliate (mint an affiliate for any user by email/userID) plus a programmatic affiliateCode we inject server-side at checkout (data-whop-checkout-affiliate-code). That means our graph drives attribution at checkout time—no surviving cookie required. The "auto-monetized on launch day" claim is buildable.

Feasible if

  • We inject the parent's affiliate code server-side into every checkout the child makes (overrides Whop's last-click cookie).
  • Rewards vest on activation (build in beta / buy in paid) + a money-tier anti-fraud layer.
  • We lazy-mint the legal affiliate status at first conversion—not declare 2,781 affiliates on day one.

Fails if

  • We rely on the buyer still carrying a beta cookie (gone in ~30 days).
  • We assume a DB edge = money flows. Whop pays Whop affiliate records—we must mint them.
  • We auto-enroll thousands at a curated 20–30% rate—that's a margin event, not growth.

Two things only Whop/Ari can confirm: (1) does our injected code beat Whop's last-click cookie when both exist? (2) recurring all_payments behavior. There's no affiliate.* webhook in v1 — we poll listOverrides and diff earnings. All flagged, all solvable.

Phase 1 engine

The reward ladder

Cheap + instant early (drive the zero-to-one share); rare + expensive late (margin-protected). Credits cost ~$0.01 retail but near-zero COGS and only redeemable by building—so they pull people into the product, not toward a cash-out.

MilestoneRewardWhy it sharesMargin guard
1 referralBadge + 250 credits + visible jumpZero-to-one is the whole gameCredits = near-zero COGS, in-app only
31,000 credits + 7-day Pro previewGoal-gradient kicks in; taste of paidTime-boxed = the upsell, not a giveaway
5Skip to next wave + "Builder" badgeAccess is what they actually wantCosts only a seat we were releasing
101 month Pro (vests on activation) + founding price lockFounding price creates a future payerVests on build — filters freebie-seekers
25 · Founder3 months Pro + name on wall + founder DiscordStatus + unity — recognition is the rewardRare by construction; badge/Discord = $0
50 · Champion6 months Pro + build-with-founders callsExperiential beats merch at the topUnpaid evangelists worth far more
100 · Legend1 yr Studio + real affiliate rev-share + wallAspirational halo over the whole ladderReal rev-share — NEVER vague "equity"

The Legend tier's "real affiliate rev-share" is the unified graph paying out — your top beta advocates graduate straight into the paid affiliate system. Same edge, cash currency.

The psychology — honestly

Persuasion, not dark patterns

We lean hard into Cialdini—but every scarcity and urgency claim is literally true. For a "Builder, not Magician" brand, truth is the persuasion: real scarcity builds trust, while fake scarcity provokes the strongest backlash of any dark pattern (and is an active 2026 FTC/EU enforcement target).

Scarcity

Real seat caps + real dates. "Wave 3 = 200 seats, Fri 9am." Never a fake counter.

Loss aversion

"3 people passed you this week"—only because position is genuinely zero-sum.

Goal gradient

"You're 1 referral from Pro preview"—only when literally true. Discrete endpoints.

Social proof

"1,144 builders already shipped apps." Real numbers, real activity.

Reciprocity

The friend gets credits + a head-start first. People share to give.

Unity

Founder / Champion / Legend identity. A real community of builders.

Because edges become money

The fraud guard that makes unification safe

In beta, a fake referral only costs us a position—cheap. But these edges auto-convert to paid affiliate edges. So a ring that games the leaderboard now holds paid edges. One rule closes most of it:

🔒 Vest on activation

Beta rewards vest only when the referee actually builds an app. Affiliate commission only pays when they buy. A fake signup can't build and won't buy—so fraud earns nothing in either phase. This is the single most powerful anti-fraud rule, and unification makes it apply end-to-end.

🛡️ Money-tier stack

On top of vesting: graph-overlap checks (email/IP/device/payment-instrument) at mint · commission math so a self-purchase is never net-positive · first_payment + retention gate for beta-seeded edges · velocity/ASN clustering at edge-write · Whop's refund auto-clawback · terms-based clawback + archive authority.

Commission rate, tiered by edge type: beta auto-enrolled → 10–15%, first_payment (high cannibalization on people who'd buy anyway); intentional/content affiliates → 20–30%. The launch-day magic still lands at 10–15% — automatic is the delight, not the rate.

What to build, in order

The sequence falls out of the model

Fix the wave-grant footgun first

The current "admit the next wave" path stamps access but never sets status:active—it would silently strand everyone. Must fix before any release lever is pulled. (Same bug class we just fixed in the per-person grant.)

The unified attribution edge

Write ONE permanent "X brought Y" edge—on referral link AND invite-code redemption. This is the spine of the whole system.

Score-ranked queue + share loop

Replace flat "+25 spots" with a transparent score. One-tap share + pre-written copy + personalized OG card at the post-signup moment (highest-leverage share point).

Vest-on-activation + anti-fraud Layer 1

Disposable-email block, device fingerprint, velocity caps, ring detection—before scaling traffic, so the list stays clean.

Prove Whop attribution in sandbox

The one load-bearing test: mint a test parent by email → inject their code at a fresh checkout from a never-touched device → confirm it credits them AND overrides last-click. Pass = the vision is proven. (Then it's all policy we control.)

Pre-wire the Phase-2 flip

Lazy-mint Whop affiliate at first conversion + polling reconciliation (no webhook in v1). Build it so public launch is a flag flip, not a rebuild.

How we'll know it's working

The North Star is conversion velocity, not list size

The only slide that matters: "if we release X off the waitlist, we get Y paying customers." That Y/X ratio is the whole game. Release fast (<30 days) or conversion decays.

K > 1.0
Viral coefficient, sustained (stretch 1.5–3, Robinhood territory)
20–30%
Share rate (% who refer ≥1)
> 20%
Waitlist → paid, if released <30 days
< 3–4 wk
Experienced wait for top advocates (anti-congestion governor)

The hidden self-destruct (proven in queuing-theory research): referral-priority queues collapse via congestion + perceived unfairness if too successful. Guard = release cadence proportional to inflow. The wait is a tool, not a warehouse.

The bottom line

One graph. Two phases. The right route.

Don't build referrals, invites, and affiliates as three systems that drift apart and force your advocates to start over at launch. Build one attribution graph, a phase-aware payout layer, and vest-on-activation as the fraud guard that works end-to-end. Your beta advocates build their downline now—and wake up monetized on launch day. It's feasible, it's fraud-resistant, and it's honest.