Economics / framework / Free to use

The Referral Reward Promise Review Framework

Review both sides of a referral offer before launch: cohort economics, earned reward obligations, eligibility, reconciliation and a human-approved decision.

Growthcraft Editorial · 2026-10-12. AI-assisted research and implementation. Examples are synthetic; Akshay's personal review is not claimed.

Copyable template

Select and copy the complete template below. With JavaScript enabled, you can edit, copy and download it in the interactive workspace.

REFERRAL REWARD PROMISE REVIEW — original Growthcraft synthesis
DECISION: proposed offer, acquisition cohort, owner and deadline
QUALIFICATION: new-customer definition, qualifying order/payment, refund maturity, exclusions and review appeal
SCOPE: currency, acquisition dates, contribution horizon, reward settlement horizon, extraction timestamp
COST CONTRACT: contribution before rewards; friend benefit treatment; referrer economic cost; overhead allocation; double-count checks
EARNED AWARDS: stable award ID, anonymous acquisition ID, rule version, face value, cost policy, pending/redeemed/reversed status
SCENARIOS: assumed redemption; 100% redemption; break-even threshold; excluded costs
CONTROLS: issuance idempotency; new-customer checks; abuse review; fulfilment owner; support explanation
EVIDENCE: attributed qualified customers versus independently measured incremental effect
GATE: hold / redesign / bounded test; rationale; approver; evidence needed; stop condition
FOLLOW-UP: reconcile reward ledger; document definition changes; preserve customer promises

When to use this framework

Use it before introducing a two-sided referral offer, increasing rewards or interpreting an apparently cheap acquisition channel. It fits a consumer retailer, paid app or early-stage subscription business with a defined qualifying transaction. It is not a forecast of participation, a fraud detector or an accounting policy. A sales introduction without a costed customer outcome needs a separate funnel model.

Five gates from idea to a bounded test

Define qualification. The growth lead and support owner agree what makes an acquisition eligible, when the reward is earned and how reversals are handled. A code redemption is not necessarily a new person. Record the published promise and its version before launch.
Freeze the cost contract. Finance or the commercial owner defines contribution after ordinary variable costs and refunds but before the two reward costs being modeled. Put the friend discount in exactly one place. Identify whether referrer cost means cash, foregone margin, product fulfilment or a documented blended estimate.
Reconcile earned awards. Operations counts all awards earned by the cohort, including those not yet redeemed. Count awards rather than unique advocates; multiple rewards or tiers can break a one-to-one assumption. Keep pending qualification separate from an earned promise.
Stress-test redemption. The analyst models a supported assumption and full redemption. A negative full-redemption balance does not make the offer automatically unacceptable, but it exposes dependence on non-redemption and requires explicit commercial approval. A negative balance even at zero redemption means changing the redemption assumption cannot rescue the model.
Approve the learning plan. The growth owner specifies an offer cap, evidence review and stop condition before a human authorizes a test. Track incremental effects separately. Do not retroactively reduce already-earned benefits to make the spreadsheet look better.

Worked synthetic decision

A cohort contains 200 qualified acquired customers with 30 currency units of first-order contribution each, before a 10-unit friend benefit. There are 200 earned referrer awards costing 20 units each when redeemed, and 1,000 units of programme costs. At 60% redemption, reward cost is 2,400, total programme cost is 5,400 and contribution remaining is 600. Full redemption produces −1,000. The break-even redemption threshold is 75%.

The decision is not to advertise a profitable referral engine. The team first validates whether its redemption assumption is mature, checks whether the 30-unit contribution already includes the friend benefit, and gets approval for the full-redemption exposure. A bounded test can then investigate participation and acquisition quality. The analyst records cost assumptions; support verifies the promised customer experience; the commercial owner signs the risk decision.

Use the worksheet

Copy the template above into the launch brief. Link to access-controlled order and award evidence, not customer identities. Preserve the original offer version and store subsequent corrections as a new review. Assign one person to reconcile the awarded, redeemed and reversed totals; a dashboard screenshot does not replace a reproducible ledger.

Failure modes and limits

A short observation window can understate redemption. Face value and cost can differ. Product returns can remove contribution after an award is earned. Rewards may be shared or used by existing customers. An apparently low cost per tracked acquisition can reflect attribution rather than incremental demand. No universal reward percentage fixes those issues. Hold the decision when a material cost or eligibility definition is missing; investigate the smallest missing input instead of inventing a benchmark.

Sources checked October 12, 2026: Shopify referral programme ideas (visible publication June 7, 2025, despite the 2026 title) and Stripe Checkout discounts (living documentation; publication date unavailable). They support reward-design options and redemption controls, not this original model or any promised performance gain.