If you haven't set up webhooks yet, start with the webhook setup guide<\/a>.<\/p>\n\n\n\n Every Tolinku webhook delivers a JSON payload with this structure:<\/p>\n\n\n\n The five event types you can subscribe to are:<\/p>\n\n\n\n Each event includes a The simplest pipeline has three components:<\/p>\n\n\n\n You can skip the queue for low-volume setups (under a few thousand events per day) and write directly from the receiver. But for anything production-grade, the queue protects you from destination outages and rate limits.<\/p>\n\n\n\n Here is a minimal Express receiver that verifies the webhook signature and pushes events to a queue. This pattern works for any downstream destination.<\/p>\n\n\n\n The key detail: use BigQuery<\/a> is a natural fit for deep link analytics because it handles append-heavy workloads well and integrates with most BI tools.<\/p>\n\n\n\n Create a table that accommodates all event types:<\/p>\n\n\n\n Partitioning by date and clustering by event type and platform keeps query costs low when you filter on those columns, which you almost always will.<\/p>\n\n\n\n Use the BigQuery Storage Write API<\/a> for real-time inserts. The older The For higher volumes, buffer events locally (or in Cloud Storage) and use a load job<\/a> instead of streaming inserts. Load jobs are free (no per-row charge) and handle large volumes efficiently. The tradeoff is latency: load jobs take seconds to minutes, while streaming inserts are near-instant.<\/p>\n\n\n\n A practical pattern: stream events to a Cloud Storage bucket as newline-delimited JSON files (one file per minute), then run a scheduled load job every 5 minutes.<\/p>\n\n\n\n Snowflake<\/a> works best with a staging approach: land the raw JSON in a stage, then transform it with a COPY command or Snowpipe for continuous loading.<\/p>\n\n\n\n The Snowpipe<\/a> monitors a cloud storage location and automatically loads new files. Your webhook receiver writes JSON files to S3 or GCS, and Snowpipe picks them up within minutes.<\/p>\n\n\n\n This approach decouples your receiver from Snowflake entirely. The receiver writes files; Snowpipe handles the rest.<\/p>\n\n\n\n Amplitude<\/a> expects events in a specific format through their HTTP V2 API<\/a>.<\/p>\n\n\n\n A few things to note:<\/p>\n\n\n\n Mixpanel's Import API<\/a> accepts events in a similar structure.<\/p>\n\n\n\n The If you need to send events to more than one destination (say, BigQuery for long-term storage and Amplitude for product analytics), the queue pattern from earlier makes this straightforward. Each destination gets its own consumer.<\/p>\n\n\n\n Use Webhooks are push-based: they deliver events as they happen. If your pipeline was down for an hour and missed events, you need a way to recover.<\/p>\n\n\n\n Tolinku's approach<\/strong>: Webhook deliveries are logged with status codes, response times, and attempt counts. Failed deliveries are retried automatically (3 retries at 1 minute, 5 minutes, and 30 minutes). Check the retry logic guide<\/a> for details.<\/p>\n\n\n\n![\"Tolinku](\"https:\/\/tolinku.com\/blog\/wp-content\/uploads\/2026\/03\/platform-webhooks.png\")
\nThe webhooks page with create form, webhook list, and delivery log.<\/em><\/p>\n\n\n\nThe Webhook Payload<\/h2>\n\n\n\n
{\n "event": "link.clicked",\n "timestamp": "2026-05-20T14:32:08.441Z",\n "data": {\n "prefix": "go",\n "token": "summer-sale",\n "hostname": "links.example.com",\n "ip": "203.0.113.42",\n "platform": "ios",\n "device_type": "mobile",\n "campaign": "instagram-story-may"\n }\n}\n<\/code><\/pre>\n\n\n\n\n
link.clicked<\/code>: A deep link was tapped or clicked<\/li>\ndeferred_link.claimed<\/code>: A deferred deep link was resolved after app install<\/li>\ninstall.tracked<\/code>: An app install was attributed to a deep link<\/li>\nreferral.created<\/code>: A new referral was registered<\/li>\nreferral.completed<\/code>: A referred user completed the required action<\/li>\n<\/ul>\n\n\n\ndata<\/code> object with fields specific to the event type. The X-Webhook-Signature<\/code> header contains an HMAC-SHA256 signature for verification, and the X-Webhook-Event<\/code> header contains the event type string. See the webhook event types guide<\/a> for full payload documentation.<\/p>\n\n\n\nArchitecture: Webhook Receiver to Warehouse<\/h2>\n\n\n\n
\n
Tolinku Webhook \u2192 Receiver (HTTP) \u2192 Queue (SQS\/Pub\/Sub) \u2192 Loader \u2192 Warehouse\n<\/code><\/pre>\n\n\n\nReceiver: Verify and Enqueue<\/h2>\n\n\n\n
import express from 'express';\nimport crypto from 'crypto';\n\nconst app = express();\napp.use('\/webhooks', express.raw({ type: 'application\/json' }));\n\nconst WEBHOOK_SECRET = process.env.WEBHOOK_SECRET!; \/\/ whsec_...\n\napp.post('\/webhooks\/tolinku', (req, res) => {\n const signature = req.headers['x-webhook-signature'] as string;\n const expected = crypto\n .createHmac('sha256', WEBHOOK_SECRET)\n .update(req.body)\n .digest('hex');\n\n if (signature !== expected) {\n return res.status(401).send('Invalid signature');\n }\n\n \/\/ Respond immediately, then process asynchronously\n res.status(200).send('OK');\n\n const event = JSON.parse(req.body.toString());\n enqueue(event);\n});\n\nfunction enqueue(event: any) {\n \/\/ Implementation depends on your queue: SQS, Pub\/Sub, Redis, etc.\n console.log(`Enqueued ${event.event} at ${event.timestamp}`);\n}\n\napp.listen(3000);\n<\/code><\/pre>\n\n\n\nexpress.raw()<\/code> so you receive the raw body for signature verification, then parse JSON after validation. See the webhook security guide<\/a> for a detailed walkthrough of signature verification.<\/p>\n\n\n\nBigQuery Integration<\/h2>\n\n\n\n
Table Schema<\/h3>\n\n\n\n
CREATE TABLE IF NOT EXISTS `your_project.deep_links.webhook_events` (\n event_type STRING NOT NULL,\n event_timestamp TIMESTAMP NOT NULL,\n received_at TIMESTAMP NOT NULL,\n prefix STRING,\n token STRING,\n hostname STRING,\n ip STRING,\n platform STRING,\n device_type STRING,\n campaign STRING,\n raw_data JSON\n)\nPARTITION BY DATE(event_timestamp)\nCLUSTER BY event_type, platform;\n<\/code><\/pre>\n\n\n\nStreaming Insert<\/h3>\n\n\n\n
insertAll<\/code> streaming API works too but has higher per-row costs.<\/p>\n\n\n\nimport { BigQuery } from '@google-cloud\/bigquery';\n\nconst bigquery = new BigQuery();\nconst dataset = bigquery.dataset('deep_links');\nconst table = dataset.table('webhook_events');\n\nasync function loadToBigQuery(event: any) {\n const row = {\n event_type: event.event,\n event_timestamp: event.timestamp,\n received_at: new Date().toISOString(),\n prefix: event.data.prefix || null,\n token: event.data.token || null,\n hostname: event.data.hostname || null,\n ip: event.data.ip || null,\n platform: event.data.platform || null,\n device_type: event.data.device_type || null,\n campaign: event.data.campaign || null,\n raw_data: JSON.stringify(event.data),\n };\n\n await table.insert([row]);\n}\n<\/code><\/pre>\n\n\n\nraw_data<\/code> JSON column stores the full event data object. This gives you a structured schema for the most common fields while preserving the complete payload for event types with different data shapes.<\/p>\n\n\n\nBatch Loading Alternative<\/h3>\n\n\n\n
Snowflake Integration<\/h2>\n\n\n\n
Table and Stage Setup<\/h3>\n\n\n\n
CREATE TABLE IF NOT EXISTS deep_links.webhook_events (\n event_type VARCHAR NOT NULL,\n event_timestamp TIMESTAMP_NTZ NOT NULL,\n received_at TIMESTAMP_NTZ DEFAULT CURRENT_TIMESTAMP(),\n payload VARIANT NOT NULL\n);\n\nCREATE STAGE IF NOT EXISTS deep_links.webhook_stage\n FILE_FORMAT = (TYPE = 'JSON');\n<\/code><\/pre>\n\n\n\nVARIANT<\/code> column stores the entire webhook payload. You can extract structured fields with Snowflake's JSON path syntax: payload:data.platform::string<\/code>.<\/p>\n\n\n\nSnowpipe for Continuous Loading<\/h3>\n\n\n\n
CREATE PIPE IF NOT EXISTS deep_links.webhook_pipe\n AUTO_INGEST = TRUE\nAS\n COPY INTO deep_links.webhook_events (event_type, event_timestamp, payload)\n FROM (\n SELECT\n $1:event::string,\n $1:timestamp::timestamp_ntz,\n $1\n FROM @deep_links.webhook_stage\n );\n<\/code><\/pre>\n\n\n\nAmplitude Integration<\/h2>\n\n\n\n
async function sendToAmplitude(event: any) {\n const amplitudeEvent = {\n event_type: `deep_link.${event.event}`,\n time: new Date(event.timestamp).getTime(),\n platform: event.data.platform || 'unknown',\n event_properties: {\n prefix: event.data.prefix,\n token: event.data.token,\n hostname: event.data.hostname,\n campaign: event.data.campaign,\n device_type: event.data.device_type,\n },\n ip: event.data.ip,\n };\n\n await fetch('https:\/\/api2.amplitude.com\/2\/httpapi', {\n method: 'POST',\n headers: {\n 'Content-Type': 'application\/json',\n },\n body: JSON.stringify({\n api_key: process.env.AMPLITUDE_API_KEY,\n events: [amplitudeEvent],\n }),\n });\n}\n<\/code><\/pre>\n\n\n\n\n
deep_link.<\/code> to distinguish webhook events from client-side events in Amplitude's event taxonomy. This makes it easy to filter and segment.<\/li>\nMixpanel Integration<\/h2>\n\n\n\n
async function sendToMixpanel(event: any) {\n const mixpanelEvent = {\n event: `Deep Link ${event.event.replace('.', ' ').replace(\/\\b\\w\/g, c => c.toUpperCase())}`,\n properties: {\n time: Math.floor(new Date(event.timestamp).getTime() \/ 1000),\n $insert_id: crypto.createHash('sha256')\n .update(JSON.stringify(event))\n .digest('hex')\n .substring(0, 36),\n prefix: event.data.prefix,\n token: event.data.token,\n hostname: event.data.hostname,\n platform: event.data.platform,\n device_type: event.data.device_type,\n campaign: event.data.campaign,\n ip: event.data.ip,\n },\n };\n\n await fetch('https:\/\/api.mixpanel.com\/import?strict=1', {\n method: 'POST',\n headers: {\n 'Content-Type': 'application\/json',\n 'Authorization': `Basic ${Buffer.from(`${process.env.MIXPANEL_PROJECT_TOKEN}:`).toString('base64')}`,\n },\n body: JSON.stringify([mixpanelEvent]),\n });\n}\n<\/code><\/pre>\n\n\n\n$insert_id<\/code> is critical for idempotency<\/a>. If a webhook is retried and you send the same event twice, Mixpanel deduplicates based on $insert_id<\/code>. We generate it as a hash of the full event payload since the Tolinku webhook payload doesn't include a delivery ID.<\/p>\n\n\n\nHandling Multiple Destinations<\/h2>\n\n\n\n
async function processEvent(event: any) {\n const results = await Promise.allSettled([\n loadToBigQuery(event),\n sendToAmplitude(event),\n ]);\n\n for (const result of results) {\n if (result.status === 'rejected') {\n console.error('Destination failed:', result.reason);\n \/\/ Dead-letter queue or retry logic here\n }\n }\n}\n<\/code><\/pre>\n\n\n\nPromise.allSettled<\/code> (not Promise.all<\/code>) so one destination failing doesn't block the others. Failed deliveries go to a dead-letter queue for retry.<\/p>\n\n\n\nHandling Backfills and Replays<\/h2>\n\n\n\n