Portal Lead Webhook Integration: AI Calling Architecture for Real Estate

Q: What happens when a portal lead arrives while the same phone number is already in an active AI calling nurture campaign?

The AI calling system should check the active lead pool before initiating a new call. If the number is in an active nurture sequence, the incoming lead triggers a priority escalation: the record is merged, the source updated to include the new portal, and a fresh speed-to-lead call is scheduled within 90 seconds referencing the new inquiry. The prior nurture sequence is paused. Concurrent duplicate calls to the same number generate buyer complaints and must be prevented by a phone number deduplication check at call initiation.

Q: We use Sell.Do CRM — what’s the recommended way to integrate AI calling via webhook?

Sell.Do supports outbound webhook triggers via its Automations module. Recommended setup: create an Automation triggered on ‘Lead Created’ with a portal source filter, set the action to HTTP POST to your AI calling system’s lead intake endpoint, and map Sell.Do fields (name, phone, project, source) to the webhook payload. The AI calling system receives the webhook, validates the phone number, checks for duplicates, and initiates the call. Sub-90-second response times depend on Sell.Do’s webhook latency, which is typically 10–30 seconds after lead creation.

Q: A buyer submits inquiries on MagicBricks and 99acres for the same project within 5 minutes — how should source attribution be handled?

First-portal-wins for source attribution: the CRM assigns the lead source as the portal from which the first webhook arrived. The second webhook triggers a deduplication check, detects the existing record, logs the second portal as a secondary source, and suppresses a duplicate call. The lead record shows both sources. The combined multi-portal signal is flagged as a higher-intent indicator in the lead’s qualification score.

Q: What are the technical requirements for a consistent sub-90-second response time from portal lead submission to first AI call?

Four requirements: (1) Real-time webhook intake — MagicBricks, 99acres, and Housing.com all support real-time webhooks; no batch exports. (2) No polling — the AI calling system initiates the call on webhook receipt, not on a polling interval. (3) Pre-provisioned call channels — dedicated concurrent capacity reserved for new leads, not shared with retry queues. (4) Phone number normalisation and validation under 5 seconds. End-to-end target: portal webhook to AI call initiation under 60 seconds; buyer phone rings within 90 seconds of form submission.

Q: For leads arriving between 10 PM and 7:30 AM, should the system attempt an immediate call or queue until morning?

Queue until 8:00 AM. Calling after 10 PM is outside appropriate consumer contact hours. The correct after-hours protocol: (1) Send an immediate WhatsApp acknowledging the inquiry and setting the expectation of a morning call; (2) Hold the lead in a priority overnight queue; (3) At 8:00 AM, overnight leads are released to the speed-to-lead queue before any other leads, with a 8:00–8:30 AM response target. The WhatsApp sent at inquiry time significantly reduces lead decay for overnight arrivals.

Q: How do we detect silent webhook failures — cases where the portal webhook stopped firing without error notification?

Two mechanisms: (1) Hourly volume monitoring — alert if portal lead volume drops below the historical baseline for that hour (e.g., if MagicBricks typically sends 6–10 leads per weekday business hour, a 2-hour zero-lead window triggers an alert). (2) Monthly test leads — submit a test inquiry on each portal using a dedicated test number and verify the AI call arrives within 90 seconds. Volume monitoring catches complete outages; test leads catch partial configuration failures. Document the last successful test date in your integration log.

Zappio Team

AI & Real Estate Experts · 8 February 2026 · 10 min read

Technical Architecture · Speed-to-Lead Integration

The 90-Second Response Window Begins at Portal Form Submission — Not When the Lead Arrives in Your CRM

Every brokerage operating in Gurugram's competitive real estate market tracks speed-to-lead as a KPI. Most measure it from CRM entry. The correct measurement starts at the moment the buyer clicks submit on the portal form. The gap between those two timestamps — determined by the webhook architecture connecting MagicBricks, 99acres, Housing.com, and other portals to your AI calling system — is where lead decay begins. Understanding and optimising this integration pipeline is a technical prerequisite for achieving sub-90-second response.

Data Flow Architecture

A portal lead's journey from form submission to first AI call passes through six stages, each with its own latency contribution:

Stage 1 — Buyer submits inquiry on portal (MagicBricks / 99acres / Housing.com)

↓ Portal webhook fires (0–15 seconds)

Stage 2 — Webhook POST arrives at CRM endpoint

↓ CRM creates lead record (5–30 seconds)

Stage 3 — CRM automation rule evaluates lead source + triggers outbound webhook

↓ AI calling system receives webhook (1–5 seconds)

Stage 4 — Phone normalisation, deduplication check, channel availability check

↓ Call initiated (2–8 seconds)

Stage 5 — Buyer phone rings

↓ Call connects → AI conversation begins

Stage 6 — Qualification data written back to CRM

⚡

The total latency budget for sub-90-second response is 90 seconds across all six stages. Stage 3 (CRM automation rule firing) is the most common failure point — a misconfigured rule or a CRM with a polling-based integration adds 5–15 minutes of hidden delay. Most brokerages calling at "10 minutes" are actually losing the speed-to-lead window at Stage 3, not at the portal webhook.

Portal Webhook Integration

Each major Indian real estate portal has a different webhook implementation. Understanding the specific webhook format and trigger behaviour for each portal is required for reliable real-time integration:

Supports real-time lead webhooks via the MagicBricks lead delivery API. Webhook payload includes: lead name, phone number, email, project name, property type, budget range, and submission timestamp. Phone numbers are delivered in national format (10-digit without country code). Webhook fires within 5–15 seconds of buyer submission. Requires a registered webhook endpoint in the MagicBricks partner dashboard. MagicBricks is the highest-volume portal source for Gurugram residential leads and the most critical integration to get right.

Delivers leads via HTTP POST webhook with a similar payload structure to MagicBricks. Notable difference: 99acres includes a 'source_type' field distinguishing between paid listing leads and organic search leads — useful for source quality scoring. Webhook latency is typically 10–25 seconds. 99acres also offers a batch pull API as an alternative, but this introduces 5–15 minute delays and should not be used as the primary integration method.

Housing.com (PropTiger) supports real-time webhook delivery for developer and large brokerage accounts. Payload includes buyer contact details, property configuration preference, and budget. Housing.com leads tend to have slightly higher intent scores than broad portal aggregators due to the platform's project-first browsing model. Webhook registration requires Housing.com account manager contact.

NoBroker primarily serves the rental market but generates buyer inquiries for resale and new launch properties. NoBroker's webhook integration is less standardised than MagicBricks or 99acres — some brokerage accounts use email-to-lead parsing rather than direct webhook. If email parsing is the only available method, a dedicated parsing service with a 30–60 second email delivery target is acceptable for NoBroker leads, given their lower average intent score relative to MagicBricks portal inquiries.

Facebook and Instagram Lead Ads deliver leads via Meta's Lead Ads webhook API or via CRM integration (most CRMs support native Meta Lead Ads connection). Meta lead webhooks fire within 60 seconds of form submission. However, the 90-second response target is harder to achieve for Meta leads because the typical integration path adds more middleware. Meta leads also have lower average intent than portal inquiries, which reduces the marginal cost of a 2–3 minute response delay.

CRM Integration Layer

The CRM is the integration hub between portal webhooks and the AI calling system. Four CRM configuration failure modes account for the majority of speed-to-lead degradation in Gurugram real estate deployments:

Some CRM integrations with portals are configured as scheduled batch exports (pulling leads every 15–30 minutes via a scheduled API call) rather than real-time webhook push. A lead submitted at 10:02 AM may not appear in the CRM until 10:15 AM under this configuration — and the AI calling system cannot initiate a call on a lead that doesn't exist in the CRM. Diagnosis: check whether your portal-to-CRM lead delivery shows a consistent delay pattern (leads arriving in clusters at fixed intervals indicate batch export). Fix: switch to webhook-based push delivery in the portal partner settings.

Portal webhooks deliver phone numbers in varying formats: 10-digit national (9876543210), with country code (919876543210), with +91 prefix (+919876543210), or with STD code prefix (0-9876543210). The AI calling system must normalise all formats to E.164 (+91XXXXXXXXXX) before initiating a call. If normalisation fails, the call is not initiated and the lead is silently dropped. Typical symptom: a subset of leads from one portal consistently fail to generate calls while others succeed. Fix: add a phone normalisation step in the CRM automation rule or in the AI calling system's intake handler.

When a buyer submits inquiries on two portals within minutes, both webhooks arrive at the CRM nearly simultaneously. If the CRM's duplicate detection runs asynchronously (checking for duplicates after lead creation rather than before), both leads may be created and both may trigger separate AI calls to the same buyer within seconds of each other. The buyer receives two calls from the same company within 30 seconds — a high-complaint scenario. Fix: synchronous deduplication check on phone number before lead creation, or at minimum before the AI calling trigger fires.

CRM automation engines that evaluate triggers on a polling interval (checking for new leads every 2–5 minutes) rather than event-driven (triggering immediately on lead creation) add hidden latency invisible in most dashboards. The lead exists in the CRM, but the automation rule hasn't fired. This is common in older CRM platforms and in CRM automation rules that apply complex multi-condition filters. Diagnosis: compare lead creation timestamp to automation trigger timestamp in CRM logs. Fix: migrate trigger to event-driven hooks, or reduce the polling interval. Most modern CRMs (LeadSquared, Sell.Do, Kylas) support event-driven automation.

Speed-to-Lead Verification Protocol

After initial integration setup, and monthly thereafter, run this verification protocol to confirm the end-to-end pipeline is operating within the 90-second target:

Submit a test inquiry on each portal using a dedicated test phone number (not a live number) at a known time. Record the exact submission timestamp.
Note when the AI calling system initiates the call to the test number. The elapsed time is your actual end-to-end latency.
If the call arrives after 90 seconds, identify which stage added the delay: portal webhook (check portal delivery logs), CRM lead creation (check CRM intake timestamp), CRM automation trigger (check automation logs), or AI calling initiation (check AI system logs).
Verify the call content: does the AI reference the correct portal, project name, and buyer name from the test submission?
Document the test result with the portal name, submission time, call arrival time, and latency in seconds. Maintain a log of monthly test results to detect degradation over time.

🔍

Run verification tests during peak hours (10 AM–1 PM weekday), not overnight or on weekends. Pipeline latency is often higher under load — a system that achieves 45-second response at 2 AM may deliver 3-minute response at 11 AM when concurrent lead volume is high and call channels are under pressure. Peak-hour verification is the only test that reflects production conditions.

Complete Integration Stack

Reference architecture for a Gurugram residential brokerage receiving 200–800 portal leads per month:

Layer	Component	Requirement
Portal intake	MagicBricks, 99acres, Housing.com	Real-time webhook push; no batch export
CRM	LeadSquared / Sell.Do / Kylas	Event-driven automation; webhook outbound trigger
Phone normalisation	Inline CRM rule or AI system handler	E.164 format; <5 second processing
Deduplication	Phone number match on lead creation	Synchronous check; suppress duplicate triggers
Call channel pool	AI calling system	Dedicated new-lead priority pool; pre-provisioned capacity
After-hours queue	AI calling system scheduler	Hold overnight leads; release at 8:00 AM with priority
WhatsApp fallback	WhatsApp Business API	Immediate send on unanswered call + after-hours arrival
Monitoring	Volume alerts + monthly test leads	Alert on below-baseline volume; monthly latency verification

Frequently Asked Questions

The AI calling system should check the active lead pool before initiating a new call. If the number is in an active nurture sequence, the incoming lead triggers a priority escalation: the record is merged, the source updated to include the new portal, and a fresh speed-to-lead call is scheduled within 90 seconds referencing the new inquiry. The prior nurture sequence is paused. Concurrent duplicate calls to the same number generate buyer complaints and must be prevented by a phone number deduplication check at call initiation.

Sell.Do supports outbound webhook triggers via its Automations module. Recommended setup: create an Automation triggered on 'Lead Created' with a portal source filter, set the action to HTTP POST to your AI calling system's lead intake endpoint, and map Sell.Do fields (name, phone, project, source) to the webhook payload. The AI calling system receives the webhook, validates the phone number, checks for duplicates, and initiates the call. Sub-90-second response times depend on Sell.Do's webhook latency, which is typically 10–30 seconds after lead creation.

First-portal-wins for source attribution: the CRM assigns the lead source as the portal from which the first webhook arrived. The second webhook triggers a deduplication check, detects the existing record, logs the second portal as a secondary source, and suppresses a duplicate call. The lead record shows both sources. The combined multi-portal signal is flagged as a higher-intent indicator in the lead's qualification score.

Four requirements: (1) Real-time webhook intake — MagicBricks, 99acres, and Housing.com all support real-time webhooks; no batch exports. (2) No polling — the AI calling system initiates the call on webhook receipt, not on a polling interval. (3) Pre-provisioned call channels — dedicated concurrent capacity reserved for new leads, not shared with retry queues. (4) Phone number normalisation and validation under 5 seconds. End-to-end target: portal webhook to AI call initiation under 60 seconds; buyer phone rings within 90 seconds of form submission.

Queue until 8:00 AM. Calling after 10 PM is outside appropriate consumer contact hours. The correct after-hours protocol: (1) Send an immediate WhatsApp acknowledging the inquiry and setting the expectation of a morning call; (2) Hold the lead in a priority overnight queue; (3) At 8:00 AM, overnight leads are released to the speed-to-lead queue before any other leads, with an 8:00–8:30 AM response target. The WhatsApp sent at inquiry time significantly reduces lead decay for overnight arrivals.

Two mechanisms: (1) Hourly volume monitoring — alert if portal lead volume drops below the historical baseline for that hour (e.g., if MagicBricks typically sends 6–10 leads per weekday business hour, a 2-hour zero-lead window triggers an alert). (2) Monthly test leads — submit a test inquiry on each portal using a dedicated test number and verify the AI call arrives within 90 seconds. Volume monitoring catches complete outages; test leads catch partial configuration failures. Document the last successful test date in your integration log.

Portal webhook latency figures, CRM automation behaviour descriptions, and speed-to-lead benchmarks in this article are based on observed integration deployments for Gurugram residential real estate brokerages through 2026. Specific latency depends on portal account type, CRM configuration, and AI calling system architecture. Portal webhook APIs and CRM integration capabilities are subject to change. Verify current webhook specifications with each portal's partner support documentation.

Loading article...

Ready to stop losing leads?

Join 200+ real-estate consultants using Zappio. Go live in 2 hours.