Designing a Coupon and Discount System

A promotional system manages discount codes, flash sales, and loyalty rewards without abuse or race conditions.

Coupon Types

Percentage discount: 20% off entire order
Fixed amount:        $10 off orders > $50
Free shipping:       Waive delivery fee
BOGO:               Buy one get one free
Flash sale:         50% off for 2 hours
Loyalty points:     1 point per $1 spent

Data Model

coupons(
  code         VARCHAR UNIQUE,
  type         ENUM('percent', 'fixed', 'shipping'),
  value        DECIMAL,
  min_order    DECIMAL,
  max_uses     INT,
  used_count   INT,
  user_limit   INT,        -- uses per user
  valid_from   TIMESTAMP,
  valid_until  TIMESTAMP,
  product_ids  UUID[]      -- restrict to products
)

coupon_usages(coupon_code, user_id, order_id, used_at)

Validation Flow

User applies coupon_code:
1. Does coupon exist and is active?
2. Not expired (valid_from ≤ now ≤ valid_until)?
3. Usage limit not reached (used_count < max_uses)?
4. User hasn't exceeded per-user limit?
5. Minimum order value met?
6. Applicable to cart items?
→ Apply discount

Race Condition Prevention

Flash coupons with 1000 uses available — 50K simultaneous requests.

-- Redis Lua script (atomic)
local current = redis.call('INCR', 'coupon:FLASH50:count')
if current > 1000 then
  redis.call('DECR', 'coupon:FLASH50:count')
  return 0  -- rejected
end
return 1  -- accepted

Sync Redis counter to DB asynchronously.

Anti-Abuse

• One code per email/device fingerprint
• Rate limit coupon validation attempts
• Flag accounts creating multiple emails
• Velocity checks: >5 coupons in 1 hour = suspicious

Flash Sale Architecture

Scheduler triggers at sale start time
    ↓
Update product prices in Redis
    ↓
Invalidate product cache
    ↓
Push notification to users

Conclusion

Coupon systems require atomic counters (Redis) for race-free usage tracking, validation pipelines for anti-abuse, and scheduled systems for flash sales.