Store credit
Understand monetary reward lots in Store credit.
Sellerlane includes two related but separate growth systems:
Neither system changes the financial meaning of an order. A discount reduces the eligible selling price; store credit or a gift card is a payment tender; and points remain a non-cash loyalty balance until redeemed into a defined reward. See Financial semantics.
For worked examples that connect these balances to COD, fulfillment, self-service, events, and review rewards, see Order & rewards lifecycle.
Every member has separate balances:
| Balance | Meaning |
|---|---|
| Pending | Earned but still inside a vesting/return hold; not spendable |
| Available | Vested, unexpired points that are not reserved |
| Reserved | Held for an active cart or redemption; still outstanding but unavailable elsewhere |
| Unrecovered | Points that should have been clawed back after the member had already spent them; future earnings repay this debt first |
The append-only points ledger records the original earn, vesting, redemption, expiry, manual adjustment, and reversal. A correction adds a compensating entry; it never deletes history.
The base program controls:
Purchase points use a snapshotted earn basis. A later rule edit does not rewrite points already earned. Full cancellation, void, and terminal full refund always reverse the corresponding earned basis. Partial refunds do so when Partially refunded is included in the program’s configured clawback statuses; a retry is idempotent.
Earn rules can target purchases, spend thresholds, products, punch cards, signup, birthday, newsletter opt-in, reviews, referrals, verified custom actions, and metafield conditions. Each rule can carry a schedule, customer eligibility, frequency cap, and a versioned reward definition.
Only actions Sellerlane can verify should be described as completed actions. A social-network URL open can be rewarded as a social visit/click; it must not be labelled a verified follow unless the network provides a trusted confirmation callback.
Review earns must be rating-neutral. Verification, text/photo/video content, and one-per-order/customer limits are valid gates; positive sentiment or a minimum rating above the universal 1-star floor is not. See Review requests and rewards.
For media rules, “photo” or “video” means at least one unhidden object with a persisted Clean verdict, committed canonical copy, edge publication marker, and Public ready timestamp. The buyer’s attachment window must be closed and every attached object must be terminal before the qualification generation is frozen. Pending, rejected, scan-failed, clean-but-not-public-ready, and hidden objects do not qualify. Base review points and the optional media bonus are posted as separate idempotent lots, so removing the last qualifying media object reverses only unused media-bonus value while the base review earn remains.
This loyalty earn is configured independently from the review-incentive rules
under Reviews. The review-incentive engine chooses exactly one richest matching
review_reward_rules payout (video, then photo, then text/any); multiple review
reward tiers never stack. Loyalty base and media-bonus lots do stack with each
other and may coexist with a discount or store-credit review incentive. If the
selected review incentive is also points, the loyalty rule owns that review’s
points award and the duplicate incentive-points mint is suppressed. Configure
both systems only when that combined policy is the merchant’s intended offer.
The later “add a photo or video” reminder is not a second earning opportunity. Its seven-day upload window begins only after qualification is frozen, and those objects are permanently excluded from the loyalty media-bonus predicate. They still require normal scanning and approval before display.
Tiers can qualify members by lifetime points, eligible orders, or spend and can apply an earn multiplier and birthday points. Tier history is versioned so an account has one current membership while previous memberships remain auditable.
Timed campaigns can add a multiplier or flat bonus and can scope eligibility by product, collection, segment, or tier. Overlapping campaigns follow the published campaign combination rule; the resulting points are posted once with the winning campaign/rule snapshot.
Published redemption rules define the exact exchange:
A redemption reserves points before issuing/applying the benefit. Completion consumes the reservation exactly once; cancellation releases it. Loyalty free-product, gift-card, and store-credit redemption definitions currently fail typed validation and cannot be published. Store credit issued by the separate referral or review-reward engines creates one linked monetary ledger artifact; it is not a second points balance.
The admin only exposes reward kinds with a complete issuer and reversal path. Unsupported catalog definitions cannot be activated as customer-facing promises.
For purchase points, the reversal basis is the refunded or canceled eligible merchandise basis, not the entire gateway refund when that refund also includes tax or shipping.
If the member still has the affected points, the reversal consumes them. If the member already spent them, the unrecovered balance records the debt and future earnings repay it before becoming available. This preserves the original ledger and prevents a refund from manufacturing value.
A referral moves through:
attributed → qualified → held → rewarded, or becomes disqualified.
Percentage advocate rewards are calculated from the snapshotted eligible order basis and rounded to a whole minor unit. Store-credit referral rewards are fixed money amounts. The award, qualifying order, currency, rule snapshot, and linked credit lot/discount remain connected for audit.
Sellerlane separates deterministic invalidity from uncertain evidence:
Every input is interpreted at the qualification anchor. The disposable-domain boolean is snapshotted when attribution is accepted, not recomputed from the customer’s current email. Address comparison uses only non-test, non-canceled advocate orders that were paid or partially refunded and whose order/address rows existed by the anchor. Device and network populations use signup events whose occurrence and ingestion timestamps are no later than the anchor. The complete privacy-minimized evidence result is then stored on the evaluation row with a SHA-256 fingerprint. The database prevents later changes to the anchor, snapshot, or fingerprint.
The case stores privacy-minimized summaries, counts, source classes, and the
window—not raw addresses, IP addresses, user agents, or visitor tokens. These
are first-party observations. Sellerlane does not claim an external provider
risk score, device graph, proxy/VPN verdict, or email-reputation verdict unless
that source is explicitly labelled providerDependent in a future integration.
A signal is not proof of fraud. Shared households, workplaces, schools, networks, and devices are normal false-positive cases. The queue therefore shows evidence as triage and requires an authorized operator to record a reason for approve, reject, or override.
Risk evaluation does not decide whether the referred customer’s already-paid order was paid. Sellerlane captures the complete anchored evidence snapshot once, stores it durably, and then evaluates only that stored snapshot inside a database savepoint. If evaluation or case persistence fails after capture, Sellerlane rolls back only that attempt, commits the paid order/referral qualification, and retains:
evaluation_pending technical hold; andThe periodic evaluator uses persisted exponential backoff and reads the stored
snapshot; it never queries the current customer, address, order, device, or
network population. A successful clean retry removes the technical placeholder
and lets the normal award issuer continue. A retry whose snapshot contains
evidence replaces the placeholder with a real merchant-review case.
evaluation_pending is not merchant evidence and cannot be approved or
rejected from the UI.
If the source snapshot itself cannot be captured, reconstructing it later would
not be an honest as-of query because email and order addresses can change. The
award therefore enters evaluation_capture_failed: a fail-closed, non-spendable
technical state that is excluded from automated evaluation and requires
operational repair/replay from an authoritative source. It is not a fraud
verdict and merchants cannot approve around it.
The issuer never runs evidence queries again. It consumes the completed durable result, so a duplicate payment webhook or a later worker clock cannot change the risk window or roll the paid order back.
Each decision is bound to the exact SHA-256 evidence fingerprint displayed to the operator. A changed signal, count, summary, or window changes the fingerprint and invalidates the stale action. In the normal path, the immutable evaluation snapshot prevents those fields from drifting between retry and decision.
| Decision | Award effect |
|---|---|
| Approve | Removes the fraud hold. A pending award issues only if fulfillment, return-hold, cap, currency, and order-eligibility rules are also satisfied. If an operational race exposed a case after issuance, approval preserves the issued economics; observing evidence alone never reverses value. |
| Reject | This is the only fraud-review decision that cancels or claws back value. It cancels a pending award; for an issued award, Sellerlane disables an unused discount or reverses the unspent store-credit lot. Already used value is preserved in history and requires an explicit audited recovery/manual adjustment rather than a fabricated reversal. |
| Override reject | Applies the same rejection/clawback rules to a previously approved case. |
| Override approve | Restores a false-positive award only when it was canceled before issuance. An issued-then-rejected reward cannot be silently re-issued; use an audited manual adjustment. |
Idempotency keys make client retries replay the one recorded transition and financial effect. Audit history records the prior/new status, action, reason, operator, evidence fingerprint, and timestamp. Viewing cases requires customer access; actions require Manage customers, the same authority used for store-credit value changes.
Monitor, by currency and period:
Never add points units to money. A points liability estimate and a monetary store-credit liability are separate measures.
Store credit
Understand monetary reward lots in Store credit.
Discounts
Configure eligibility and combinations in Discounts.
Review rewards
Collect policy-safe reviews in Review requests and rewards.
Analytics definitions
Measure the systems consistently in Analytics definitions.