Optimizing SMS Delivery Speed | MacraSMS Developer Guide

Understanding SMS Latency

SMS delivery involves multiple hops: your application → MacraSMS API → carrier SMSC → recipient device. Each hop adds latency. Understanding where delays occur helps you optimize effectively:

Hop	Typical Latency	Optimization Opportunity
App → MacraSMS API	50-200ms	Connection pooling, geographic proximity
MacraSMS Internal Processing	10-50ms	Pre-validation, cached routing tables
MacraSMS → Carrier SMSC	100-500ms	Direct routes vs aggregators, load balancing
Carrier SMSC → Device	200ms-30s+	Network congestion, device status, roaming
Total End-to-End	500ms-60s+	Cumulative effect of all above factors

Reality Check

You cannot control carrier-side latency. Focus optimization efforts on the first three hops where you have direct influence. Target sub-second API response times as your primary KPI.

Route Selection Strategy

Choose the right route type based on message priority:

Direct Routes (Priority): Use for OTPs, transactional alerts, time-sensitive notifications. Higher cost but lowest latency (200-800ms average)
Grey Routes (Budget): Acceptable for marketing campaigns where 5-30 second delay is tolerable. Lower cost but variable latency
Smart Routing: Configure automatic fallback from direct to grey routes when direct paths are congested or unavailable
Carrier-Specific Routes: Some carriers perform better on specific routes. Test and maintain per-carrier route preferences

Best Practice

Tag messages with priority levels in your API calls. Our platform automatically selects optimal routes based on priority tags and real-time network conditions.

Message Segmentation

Long messages are split into segments, each sent separately. Optimize segmentation to reduce overhead:

Stay Under 160 Characters: Single-segment messages avoid reassembly delays at carrier level
Use GSM 7-bit Encoding: Supports 160 chars/segment vs 70 chars for Unicode. Avoid emojis and special characters unless necessary
Pre-Segment Long Messages: If multi-segment is unavoidable, pre-segment client-side to reduce server processing time
Avoid Concatenated Headers: Some carriers add extra bytes for concatenation. Test actual segment counts before deployment

// Character count impact on segments
const gsm7Message = "Your OTP is 123456. Valid for 5 minutes."; // 42 chars = 1 segment
const unicodeMsg = "Your OTP is 🔐123456. Valid ⏱️5 min.";     // 38 chars = 1 segment (Unicode)
const longGsm7 = "Your order #12345 has shipped via courier..."; // 161+ chars = 2+ segments

// Always prefer GSM 7-bit when possible
function getEncoding(message) {
    const gsm7Regex = /^[A-Za-z0-9 \r\n@£$¥èéùìòÇØøÅåΔ_ΦΓΛΩΠΨΣΘΞÆæßÉ!\"#¤%&'()*+,\-.\/:;<=>?¡ÄÖÑÜ§¿äöñüà^{}\\€~]*$/;
    return gsm7Regex.test(message) ? 'GSM7' : 'UCS2';
}
            

Batch Sizing & Throttling

Bulk sends require careful batching to avoid overwhelming systems:

Volume	Recommended Batch Size	Delay Between Batches
< 100 messages	Single request	N/A
100 - 1,000	100 per request	100ms
1,000 - 10,000	500 per request	500ms
10,000 - 100,000	1,000 per request	1s
> 100,000	2,000 per request	2s + monitor queue depth

Rate Limit Warning

Exceeding rate limits triggers HTTP 429 responses that add significant retry overhead. Always implement token bucket or leaky bucket algorithms for outbound throttling.

Connection Pooling

Reuse HTTP connections to eliminate TLS handshake overhead:

HTTP Keep-Alive: Enable persistent connections with keep-alive headers
Connection Pool Size: Maintain pool size matching your concurrency requirements (typically 10-50 connections)
Idle Timeout: Set 30-60 second idle timeout to prevent stale connections
Max Lifetime: Rotate connections every 5-10 minutes to avoid carrier-side timeouts
Health Checks: Implement periodic ping tests to detect and remove dead connections

Async Processing Patterns

Never block user-facing requests on SMS sends:

Queue-Based Architecture: Push messages to Redis/RabbitMQ/Kafka queue immediately after validation
Worker Processes: Dedicated workers consume queue and make API calls independently
Immediate Acknowledgment: Return 202 Accepted to client instantly, provide message ID for status tracking
Webhook Notifications: Use webhooks instead of polling for delivery status updates
Dead Letter Queue: Route failed messages to DLQ for investigation without blocking main queue

// Example: Async send pattern with immediate acknowledgment
app.post('/send-sms', async (req, res) => {
    // Validate and enqueue immediately
    const messageId = await messageQueue.enqueue({
        to: req.body.to,
        message: req.body.message,
        senderId: req.body.senderId,
        priority: req.body.priority || 'normal'
    });
    
    // Return immediately - don't wait for API response
    res.status(202).json({
        message_id: messageId,
        status: 'queued',
        estimated_delivery: '< 5 seconds'
    });
});

// Separate worker process handles actual API calls
worker.on('message', async (msg) => {
    try {
        const result = await macraSMS.send(msg);
        await updateStatus(msg.id, result);
    } catch (error) {
        await deadLetterQueue.add(msg, error);
    }
});
            

Monitoring Performance

Track these metrics to identify bottlenecks:

API Response Time: P50, P95, P99 latencies for /send endpoint
Queue Depth: Number of pending messages in queue (alert if > 1000)
Error Rate: Percentage of 4xx/5xx responses (alert if > 1%)
Throughput: Messages sent per second vs target capacity
DLR Latency: Time between send and delivery report receipt
Retry Rate: Frequency of retries indicating transient failures

Carrier-Specific Tips

Kenyan carriers have unique characteristics:

Carrier	Peak Hours	Optimization Tip
Safaricom	8-10AM, 5-7PM	Use dedicated OTP route during peak hours. Standard routes experience 2-3x latency
Airtel	12-2PM, 6-8PM	Batch sizes >500 trigger throttling. Stick to 200-300 per batch during peaks
Telkom	9-11AM, 4-6PM	Smaller network = faster off-peak. Schedule non-urgent sends outside peak windows

Need Performance Tuning Help?

Our engineering team can audit your integration and recommend carrier-specific optimizations.

Request Performance Audit