Conversions API for Real Estate CRMs (2026): How to Send Qualified Leads, Site Visits and Bookings Back to Meta and Google
Meta and Google optimise toward whatever you tell them a lead is. Here is how Indian developers and brokers capture IDs, define qualified, and send site visits and bookings back from the CRM.
A performance manager at a Mumbai developer opens the Meta dashboard on Monday. Cost per lead is down 18% since the last launch. The presales head, in the next room, says the leads are getting worse: more wrong numbers, more "just browsing", fewer site visits. Both are right. Meta was told that a lead is a form fill, so it got very good at finding people who fill forms. Nobody told it which of those people later visited the site or booked a flat.
The fix is not a new ad format. It is a feedback loop: your CRM sends stage changes (qualified, site visit done, booked) back to Meta and Google so their algorithms optimise toward buyers instead of form-fillers. This guide covers the mechanics, what each platform requires, and what you cannot close the loop on. If you want the metric argument first, read cost per lead vs cost per booking; here we stay on the plumbing.
What "sending conversions back" actually means
Meta calls its server-side route the Conversions API, and its CRM integration guide describes it as sending lead-stage events from your CRM so Meta can learn which leads were good. Google calls the equivalent offline conversion import. The idea is identical: when a lead changes stage in your CRM, you tell the ad platform which click or form submission it came from, and what happened next.
Two things make this possible, and both are decided on day one of a lead's life, not at upload time:
- An identifier that ties the lead to the ad. For Meta Instant Forms that is the Lead ID (a 15 to 17 digit number delivered with the lead). For Google it is the click ID captured from the landing page URL (GCLID, plus the newer variants for app and web). If the CRM does not store these, there is nothing to send back later.
- A written definition of "qualified". The platform optimises toward whatever event you send. If "qualified" means different things to three presales agents, the signal is noise.
From the team that built Sell.Do
See how Sell.Do runs your sales, pre-sales and marketing on one AI-first platform.
Step 1: Capture and store the click ID or Lead ID
This is where most Indian developer stacks leak. Leads from Meta lead ads often arrive in a spreadsheet or through a connector that drops the Lead ID. Leads from Google land on a page whose form never reads the GCLID parameter.
- For Meta lead ads, pull the Lead ID into a dedicated CRM field through the lead-retrieval integration, not into a free-text note.
- For Google, add hidden fields to every landing page form that capture the click ID from the URL, and persist it through any redirect or WhatsApp click-to-chat step.
- Capture the campaign, ad set and source tags at the same moment. This is the same discipline that makes source-to-booking attribution possible, and it is the reason a CRM that tags at capture beats an export-and-merge workflow.
Step 2: Define the stages you will send
Pick two or three events, not ten. A workable real estate ladder looks like this:
- Qualified lead: budget, location and timeline confirmed on a call, with a callable number. This is your high-volume signal.
- Site visit completed: the visit actually happened, verified, not merely scheduled.
- Booking: token or booking amount received. Low volume, highest value.
Meta's guidance for CRM optimisation is worth planning around: it works best when the target stage is reached by roughly 1% to 40% of leads within 28 days, and it needs a steady flow, around 200 or more leads a month, with a learning period of two to four weeks after you start sending. A single-project launch with 60 leads a month usually cannot feed a booking-level signal. Start with "qualified" and move down the funnel as volume grows.
Step 3: Send stage changes daily
Meta's CRM guide expects daily uploads of lead-stage updates; stale data weakens the model. On Google, offline conversions must be imported within 90 days of the click, phone numbers should be in E.164 format (for example +919820012345) when used for matching, and reporting can lag by up to 24 hours. Treat these as planning numbers and confirm against the current Meta and Google documentation before you build, because both platforms change requirements often.
In practice the job is small: a scheduled export or API call each night listing lead identifier, event name, event time, and (for value-based bidding) a booking value. A broker with 500 leads a month can run this from a clean sheet; a developer with several projects should automate it from the CRM.
One operational point ties back to presales discipline: stage data is only as accurate as the agents who update it. If calls are logged late, the platform learns late. That is why speed to lead and on-time call logging matter to ad performance, not just to conversion.
What you cannot close the loop on
Be honest with your team about the gaps, or the numbers will mislead you:
- Portal leads (99acres, MagicBricks, Housing): you do not control the ad platform, so there is nothing to send conversions to. Track them in your CRM for cost per booking, but they sit outside this loop.
- Walk-ins and referrals: no click, no identifier. Tag them at the front desk so they do not pollute paid-media attribution.
- Channel-partner leads: attributable in the CRM, not to Meta or Google.
- Anything without a stored ID: history from before you started capturing Lead IDs cannot be backfilled reliably.
Also keep consent and data-handling in view. You are passing personal data back to platforms, so make sure your forms carry proper consent language and your team is aligned with India's data-protection obligations. Ask your compliance owner to review before launch.
What results to expect, and how to test
Third-party testing summaries report that advertisers sending CRM stage data to Meta see roughly 15% lower cost per quality lead on average. Treat that as an indicative average from outside sources, not a promise: your result depends on volume, definition quality and creative. Run it as a test, not a switch:
- Split one project's Meta budget. Keep one campaign on standard lead optimisation and one on the CRM-fed signal.
- Judge on cost per site visit and cost per booking after eight weeks, not on CPL, because CPL will often rise while quality improves.
- Keep the stage definition frozen for the whole test.
For wider context on how Meta and Google spend fits together, see paid lead generation for Indian real estate.
How Sell.do fits
Sell.do captures leads from Meta, Google, portals, website, walk-ins and channel partners into one record, stores source and campaign tags at capture, and tracks stage changes from first call to booking, so the data you would send back is already structured. The CRM exposes stage changes through its APIs and webhooks, which can feed an upload job. We are not claiming a one-click native conversion connector here; confirm current integration options with the Sell.do team for your setup.
A 30-day starter checklist
- Week 1: add Lead ID and click ID fields; fix forms and connectors so every paid lead stores them.
- Week 2: write the one-page definition of qualified, site visit and booking; train presales on it.
- Week 3: build the nightly upload for qualified-lead events; run it manually first.
- Week 4: start the split test and set the eight-week review date.
Marketing teams who want ready-made templates for this (the signal map and upload spec) will find them in our Closed-Loop Marketing Playbook for 2026.
Want your ad platforms to learn from bookings, not form-fills? See how Sell.do tracks every rupee of CPL and campaign ROI from first click to booking in one place, then book a walkthrough to map the stage data you could send back to Meta and Google.
Related reading
Insights from the Sell.do real-estate CRM team.