The verdict in three sentences
A Black Friday peak is 10x traffic concentrated in a few hours: your checkout must hold peaks of 500 orders/minute with latency under 800 ms and a success rate kept above 95 %. The three enemies are double payments (no idempotency), operator API saturation (no rate limiting or queue), and total outage (no fallback). Well architected, a peak becomes a non-event; poorly prepared, it costs hundreds of orders.
The bottlenecks, quantified
Checkout is not a single system but a chain: front-end, internal API, database, then a call to the aggregator that calls the mobile money operator. The weakest link sets capacity. Here are the typical bottlenecks and their 2026 limits.
| Link | Typical capacity | Limit that breaks first | Countermeasure |
|---|---|---|---|
| Front / CDN | Very high | Rarely | Static cache |
| Checkout API | 100-300 req/s | CPU, DB connections | Autoscaling |
| Database | 50-200 writes/s | Locks, connections | Pool + read cache |
| Aggregator API | Per contract | Provider rate limit | Queue |
| Operator API (Wave/OM) | Variable quotas | Operator throttling | Retry + backoff |
The key point: the mobile money operator often imposes a request cap. Exceeding it generates cascading 429 errors. The queue smooths traffic to never exceed the quota.
The architecture that holds the peak
Four mechanisms turn an unmanageable spike into a controlled flow. Each has a quantified target.
| Mechanism | Role | 2026 target | Effect if absent |
|---|---|---|---|
| Async queue | Smooth traffic | 500 orders/min absorbed | Cascading 429 errors |
| Idempotency (unique key) | Prevent double debits | 0 double payment | Customers charged twice |
| Product/stock cache | Relieve the DB | < 50 ms reads | Saturated DB |
| Outbound rate limit | Respect operator quota | 0 overruns | Operator blocking |
| Multi-operator fallback | Service continuity | Switch < 2 s | Total checkout outage |
| Retry with backoff | Absorb transient errors | 95 %+ final success | Lost orders |
Idempotency is non-negotiable: each payment attempt carries a unique key, so if the network replays the same request twice, only one debit is recorded. During a peak on an unstable network, this is what prevents double debits and disputes.
Mini case study
A fashion shop in Dakar prepares its 2026 Black Friday. Normal traffic: 12 orders/minute. Peak forecast: 480 orders/minute for 20 minutes, about 9,600 orders. Without a queue, the operator API throttles at 300/min and the success rate falls to 70 %: about 2,880 failed orders, which at an 18,000 FCFA average basket is over 51,000,000 FCFA of sales at risk. With the Kolonell queue smoothing to 300/min and retrying the rest, the rate climbs back to 96 %: fewer than 400 lost orders. Difference: about 44,000,000 FCFA saved in a single evening.
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
FAQ
What is idempotency in a payment?
It guarantees that the same request replayed (due to a network timeout) debits only once. A unique key is tied to each attempt. Without it, a customer can be charged twice during a peak.
Doesn't a queue slow the customer down?
Marginally: a few seconds of waiting beats an error. It smooths traffic below the operator quota and keeps a success rate above 95 % rather than a collapse to 70 %.
Do I really need a multi-operator fallback?
For a critical peak, yes. If Wave throttles, switching part of the traffic to Orange Money in under 2 seconds prevents a total checkout outage and saves sales.
How do I test my capacity before the peak?
Run a load test simulating 500 orders/minute two weeks ahead. Measure latency, error rate, and breaking point, then size the queue and cache accordingly.
Can caching cause stock oversells?
Yes if done poorly. You cache the product display but validate real stock at debit time, atomically, to avoid selling a sold-out item.
Let's talk about your project. We audit your checkout and prepare it to hold your next peak without losing an order. WhatsApp +221 77 596 93 33.
Mohamed Bah
Fondateur, Kolonell
Passionate about digital and entrepreneurship in Africa, Mohamed has been helping Sénégalese businesses with their digital transformation since 2020. Founder of Kolonell, he believes every SME deserves a professional and accessible online présence.
