Multi-Channel Alerts: Email, SMS, Push & In-App Notifications

Picture this: a client places an order worth 1 million rubles, but the notification never arrives—the manager loses the deal because of a hung RabbitMQ. We build multi-channel notification systems that handle peak loads of 10,000 messages per minute and guarantee 99.9% delivery. Users choose their channels: Email, SMS, Push, In-App. The key tasks—centralized settings storage, sending queue, deduplication, and retry on failures. Over 5 years, we've deployed 50+ such systems for large e-commerce and SaaS platforms. Our notification system is designed to be a robust multi-channel notification system that ensures delivery even under heavy load.

Problems we solve

Without a queue, notifications get lost during peak loads: the database can't handle thousands of inserts, external APIs return 503 errors. Invalid push tokens accumulate—each notification to such a token wastes budget. Users complain about spam if settings don't allow disabling a channel. Our architecture solves all this.

Queue architecture: Guaranteed Delivery

The system is built around a queue (Redis or RabbitMQ). Each notification is a queue job. If the external service is temporarily unavailable, the job retries with exponential backoff. After three failures—it goes to a Dead Letter Queue for manual inspection.

[Event: OrderShipped]
         ↓
[NotificationService]
  ├── Check user preferences
  ├── Email: enqueue → SendGrid/Mailgun
  ├── SMS: enqueue → SMSC/Twilio
  ├── Push: enqueue → Firebase FCM
  └── In-App: save to DB → WebSocket push

Channel comparison: Email, SMS, Push, In-App

Push notifications achieve 3x higher open rates than email, making them better for urgent alerts. According to Firebase documentation, push notification delivery rates exceed 90%. SMS is 10 times faster than email for critical alerts, but costs more.

Implementation in Laravel

class NotificationService
{
    public function notify(User $user, string $type, array $payload): void
    {
        $prefs = NotificationPreference::where('user_id', $user->id)
            ->where('notification_type', $type)
            ->first();

        $defaults = [
            'email_enabled' => true,
            'sms_enabled'   => false,
            'push_enabled'  => true,
            'inapp_enabled' => true,
        ];

        $channels = array_merge($defaults, $prefs?->toArray() ?? []);

        if ($channels['inapp_enabled']) {
            $notification = Notification::create([
                'user_id' => $user->id,
                'type'    => $type,
                'title'   => $payload['title'] ?? null,
                'body'    => $payload['body'] ?? null,
                'data'    => $payload['data'] ?? [],
            ]);
            broadcast(new NewNotificationEvent($user, $notification))->toOthers();
        }

        if ($channels['email_enabled'] && isset($payload['email'])) {
            SendEmailNotificationJob::dispatch($user, $type, $payload['email'])->onQueue('notifications');
        }

        if ($channels['sms_enabled'] && $user->phone && isset($payload['sms'])) {
            SendSmsNotificationJob::dispatch($user, $payload['sms'])->onQueue('notifications');
        }

        if ($channels['push_enabled'] && isset($payload['push'])) {
            SendPushNotificationJob::dispatch($user, $payload['push'])->onQueue('notifications');
        }
    }
}

Provider configuration

Email: templates and providers

For Email we use SendGrid (or Mailgun) with ready-made Blade templates. Each notification type is a separate mailable class. It's important to set up open and click tracking.

SMS: SMSC.ru and Twilio

For Russia—SMSC.ru, HTTP request with login and password. For international—Twilio. Both are wrapped in a queue with retries on 500 errors.

Push: Firebase FCM and Web Push

Firebase Cloud Messaging sends notifications to Android, iOS, and Web (via Service Worker). We remove invalid tokens from the database after FCM response. For Web Push, we use VAPID keys.

Client-side implementation: Web Push and notification center

Web Push: Service Worker subscription

async function subscribeToPush(): Promise<void> {
  const registration = await navigator.serviceWorker.ready;
  const subscription = await registration.pushManager.subscribe({
    userVisibleOnly: true,
    applicationServerKey: urlBase64ToUint8Array(import.meta.env.VITE_VAPID_PUBLIC_KEY),
  });
  await api.post('/api/push-tokens', {
    token: JSON.stringify(subscription),
    platform: 'web',
  });
}

React: real-time notification center

function NotificationCenter() {
  const { data, refetch } = useQuery({ queryKey: ['notifications'], queryFn: fetchNotifications });
  const unread = data?.filter(n => !n.read_at).length ?? 0;
  useEffect(() => {
    const echo = window.Echo.private(`notifications.${currentUser.id}`)
      .listen('NewNotificationEvent', () => refetch());
    return () => echo.stopListening('NewNotificationEvent');
  }, []);
  return (
    <div className="notification-center">
      <button className="bell" aria-label={`Notifications: ${unread} unread`}>
        <BellIcon />
        {unread > 0 && <span className="badge">{unread > 99 ? '99+' : unread}</span>}
      </button>
      <ul className="notification-list">
        {data?.map(notification => (
          <li key={notification.id} className={notification.read_at ? '' : 'unread'}>
            <span>{notification.title}</span>
            <time>{timeAgo(notification.created_at)}</time>
          </li>
        ))}
      </ul>
    </div>
  );
}

The Danger of N+1 Queries in Notification Systems

When fetching the notification list, don't query each user separately—use eager loading. We check all queries via Laravel Debugbar and optimize indexes. This reduces API response time from 2 seconds to 50 ms.

Step-by-Step Integration Guide

1. Analyze requirements: Define notification types, channels, and target audience.
2. Design database schema: Create tables for notifications, preferences, and push tokens.
3. Implement queue: Set up Redis or RabbitMQ and configure worker processes.
4. Integrate providers: Connect SendGrid, Twilio, FCM, and SMSC.ru with proper error handling.
5. Build user interface: Develop settings page and notification center.
6. Test and deploy: Perform load testing and deploy with monitoring.

Implementation process and timelines

What's Included in the Deliverables

• Detailed documentation of notification flows
• Access to admin panel with user management
• Codebase with commented code
• 1 month post-launch support
• Performance monitoring setup
• SLA guarantee: 99.9% uptime

Clients save on average $12,000 per year by preventing missed notifications. Typical project cost ranges from $5,000 to $15,000 depending on channels and complexity.

How to avoid duplicate notifications?

We use a unique event identifier (UUID). Before sending, we check if the event has already been processed. If yes—skip.

Contact us for a project estimate—get a consultation within 2 days. We guarantee 99.9% SLA and provide certified specialists.

Email Campaign Integration: Why Does It Often Break?

We’ve observed that a trigger email sent 10 minutes after registration converts 4–5 times better than the same email sent after 24 hours. This isn’t a marketing myth—it’s mechanics: while the user is still warm, while they remember the context. But most integrations with email services are built like this: form submits → synchronous HTTP request to API → if the API is slow, the user waits 3 seconds → the email either goes out or doesn’t, nobody knows. In one project, we saw a 30% drop in conversion simply because the email service responded with 504 and Laravel’s queue driver wasn’t configured. Lost emails often hit customers silently – no log, no alert, just a missing order confirmation.

If you’re facing lost emails or spam folder issues, order an audit of your current integration – we’ll find bottlenecks within 2 days.

Providers and Their APIs

Unisender — a Russian provider popular in the SMB segment. REST API, simple. Adding a contact: importContacts, sending a transactional email: sendEmail. Important: for transactional emails (order confirmations, password resets), Unisender Go is a separate service with a different API and separate pricing. Mixing bulk and transactional mailings in one stream is bad for domain reputation. Unisender Go handles up to 1000 requests per second.

SendPulse — provides email, SMS, web push, Viber, and Telegram bots through a unified API. Convenient for projects requiring an omnichannel approach. Automation 360 is a visual chain builder; you can trigger automation via API events. The PHP SDK (sendpulse/rest-api-php-sdk) is maintained but updated irregularly – better to use Guzzle directly.

Mailchimp — a choice for international audiences and marketing teams accustomed to the Mailchimp ecosystem. Transactional email via Mandrill (a subsidiary service). Marketing API v3 for list, tag, and campaign management. Webhooks for opens, clicks, unsubscribes, bounces.

SMS. For Russia: SMSCenter, MTS Exolve, Devino Telecom, SMS Aero. Their APIs are similar: a send method with phone, message, sender parameters (sender name must be registered separately with the operator). One nuance: the sender name must be registered through the aggregator with a contract – otherwise SMS won’t be sent on MTS/MegaFon/Beeline networks.

How to Build an Integration That Doesn’t Lose Emails?

Separate Transactional and Marketing Streams

Transactional emails (order confirmations, password resets, delivery status) go through a dedicated sender domain or subdomain tx.example.com. Marketing campaigns go through mail.example.com or news.example.com. If a marketing campaign receives many spam complaints, it should not affect the reputation of the transactional stream. According to SendGrid documentation, transactional messages should be sent through a dedicated IP pool to prevent cross-contamination.

Queue and Retry

Any call to the email API goes through a queue (Laravel Queue, Bull, Celery). If Unisender returns a 503, the job retries after 5 minutes, then 15, then 60. After 5 failed attempts, it goes to a dead letter queue with an alert. The user already received their 200 OK and knows nothing about the issue. This approach reduces bounce rate on projects to 0.5%.

Example Laravel job:

public function handle(): void
{
    try {
        $response = Http::post(config('services.unisender.email_url'), $this->params);
        if ($response->failed()) {
            $this->release(300); // retry after 5 min
        }
    } catch (\Throwable $e) {
        $this->release(300);
    }
}

Templates

We store templates in code (Blade, Twig, React Email), not in the provider’s interface. Reasons: versioning via Git, browser preview without sending, testability. For complex templates with dynamic content — react-email with export to HTML via @react-email/render.

Validation and Consent

Before adding a contact to a list — double opt-in (confirmation email). Store the confirmation timestamp in your own database. Upon unsubscription — synchronously unsubscribe both at the provider and in your database. Ignoring webhook unsubscriptions is a direct path to account suspension at the provider. All processes comply with Федеральный закон № 152-ФЗ «О персональных данных».

Deliverability Monitoring and DKIM Setup

Connect provider webhooks for events: bounce (hard and soft), spam_complaint, unsubscribe. Hard bounce — immediately mark the email as invalid in your database, stop sending. Soft bounce 3 times in a row — same. Metrics: open rate, click rate, bounce rate, unsubscribe rate — review at least once a week. Our certified engineers configure alerts in Grafana/Prometheus.

DKIM configuration steps:

1. Generate a key pair (e.g., openssl genrsa -out private.key 2048).
2. Publish the public key in DNS as a TXT record for the selector (e.g., mail._domainkey.tx.example.com).
3. Provide the selector to the provider (SendGrid, Mailgun, Unisender).
4. Verify with dig TXT mail._domainkey.tx.example.com.

SPF, DKIM, DMARC must be configured separately for each stream. We use subdomains with different DNS records.

Why Is It Important to Separate Streams?

If you send a marketing campaign from the same domain as transactional emails and receive spam complaints, you risk getting the domain blocked — and users will stop receiving even order confirmations. SPF, DKIM, DMARC (Sender Policy Framework, DomainKeys Identified Mail, Domain‑based Message Authentication, Reporting and Conformance) must be configured separately for each stream. In one project, a marketing blast with 12% spam complaints blocked the transactional domain for 48 hours — we had to re‑authenticate with Google and Yandex.

What Does the Integration Scope Include?

• Audit of current communication streams and domain reputation (SPF, DKIM, DMARC)
• Provider and schema selection: transactional vs marketing traffic
• Configuration of SPF, DKIM, DMARC DNS records
• Development of email templates (HTML + dynamic content)
• Backend integration via queues and API
• Webhook setup for deliverability and complaints
• Operations documentation and team training
• Deliverability guarantee and post‑launch support

We deliver production‑ready documentation, access to monitoring dashboards, and a handover session with your engineers. Our certified engineers provide a 30‑day post‑launch health check guarantee.

Timelines and Cost

Cost is calculated individually after audit. We provide turnkey service: from analysis to production monitoring. Contact us for a free engineer consultation — we’ll evaluate your project and give accurate timelines. Over 7 years of experience in email service integration, 50+ projects implemented. Order a free audit of your current integration and receive a report with recommendations and estimated savings.