Skip to main content

Evidence and OCR cashback

Evidence is a proof uploaded by a member (purchase receipt, ticket, invoice, delivery note) to demonstrate an action that should be rewarded.

Evidence lifecycle​

  1. Upload — the member snaps a photo of the ticket from the app/PWA.
  2. OCR — the system extracts data automatically: merchant, amount, date, products.
  3. Validation — checked against campaign rules.
  4. Approved / Rejected — the member is rewarded or notified of rejection.
  5. Closure — data is recorded for reporting and audit.

OCR​

IncentIA supports several OCR providers configurable per tenant:

  • Gemini Flash (Google) — default provider, good quality/cost ratio.
  • External API — your own OCR service.
  • Manual — no OCR; an admin reviews and validates by hand.

OCR returns a confidence score. If it exceeds the configured threshold (e.g. 70%), evidence is auto-approved. Otherwise, it goes to manual review.

Extracted data​

Typically:

  • Merchant (name, address if visible).
  • Purchase date and time.
  • Total amount.
  • Currency.
  • Items (when legible): product name, quantity, unit price.

Validation rules​

Per campaign you can configure:

  • Minimum/maximum amount of the ticket.
  • Allowed merchant (whitelist or regex pattern).
  • Required products (must appear in items).
  • Time window (only tickets from the last 48h; dates within campaign range).
  • Quota per member (max N tickets per person).

Duplicate detection​

The system stores a perceptual hash of each image plus extracted data. If a member tries to upload the same ticket twice, it's rejected automatically.

Reward​

Once approved, evidence triggers:

  • Points to the member's balance.
  • Cashback (if the campaign is cashback) — a pending disbursement is created.
  • Notification to the member on their preferred channel (email, WhatsApp, push...).

Manual review​

Evidence that doesn't auto-approve (low confidence, rules not clearly met) goes to a review queue:

  • Tenant admin sees them in their panel.
  • Partner admin can review centrally for all their brands.
  • Every review is logged (who approved, when, why it was rejected).

Shopper Collector (automation)​

Alternative to manual upload: you connect your POS / ecommerce, and purchases arrive automatically as "pre-approved evidence" linked to the member. No photo, no OCR — we take the data straight from POS to program.

Supported by Shopify and WooCommerce webhooks, or custom integration with your ERP.

Anti-fraud​

The fraud rules module detects suspicious patterns:

  • Same member uploading many tickets in a short time.
  • Tickets with suspiciously identical amounts.
  • Same image with minor variations.
  • Irregular purchase hours.

When an alert fires, rules can block the member, require manual review, or just log the incident.