Weekly Meta Ad Report for Mortgage Loan Officers: Why Call Recordings Come First

A weekly ad report turns into an argument when nobody trusts the data behind it. In the first full review of one mortgage brokerage's Meta campaigns, 28% of the spend had gone to people aged 65 and over, and that spend had produced no leads.

The owner runs FHA and non-QM campaigns and wanted a report the team could read without debating it. What follows is the setup we put around that report. It applies to any group of mortgage loan officers who generate leads on Meta and work them in a CRM.

The situation

The call was the first full review of the brokerage's FHA and non-QM campaigns. We went through targeting, creative and lead flow together, and two findings stood out.

The owner also wanted the weekly report to be objective. The goal was a report that puts the data in front of the team and leaves opinions out of the room.

• 28% of spend went to the 65+ age group, and that group returned zero leads.
• The first non-QM round had run completely broad, with no investor data behind the targeting, and it produced almost no leads.

What was broken

The targeting problem was easy to see. The harder problems sat underneath it.

First, call data was missing. One loan officer was sending leads by email instead of calling them through GoHighLevel. Calls placed through the CRM are recorded and calls placed anywhere else are not, so there was no way to verify lead quality. Without that, the ad performance report could not be completed. A report that stops at lead counts cannot tell a good lead from a bad one.

Second, the data lived in three places: the ad account, the CRM pipeline and the loan origination system. Nobody could look at them side by side.

Third, the project board had redundant tasks. Claude agents read that board, and duplicate tasks made them collide with one another and spend tokens on work that was already covered.

What we set up

The work split into reporting, call discipline and cleanup. These were the agreed next steps from the call, not finished work.

• Targeting: we agreed to cap the share of spend reaching the 65+ group and to raise FHA spend, with the targeting update set for that same night.
• Weekly report playbook: a saved prompt that builds the same report every week, so the team reads objective data and not anyone's opinion.
• Connectors: API keys and MCP connectors on the team's computers, with a command-line terminal as the hub where Claude pulls from the loan system, GoHighLevel and Meta Ads.
• Call discipline: the loan officer re-calls every FHA and non-QM lead through GoHighLevel, so each lead has a recording to judge.
• Board cleanup: redundant ClickUp tasks removed so the agents stop colliding.
• Training: a one-hour GoHighLevel training day for all staff, ahead of a review that merges Meta, pipeline and loan-system data.
• Creative and audience: new scripted talking-head video ads for the FHA and DSCR campaigns, shot to feel like posts and not ads, plus investor data for the next non-QM round. Claude's analysis of the ad data pointed to talking-head video as the highest-ceiling asset.

What changed

No result was reported on this call, so there is no lead count or cost to share. What changed is how the reporting is set up to work. The targeting update was scheduled for that night. The next review was set up to merge three sources in one view: Meta ad data, the GoHighLevel pipeline and the loan system's reporting. The weekly report is meant to come from a saved prompt, so it is built the same way every week.

The dependency is call recording. Until every lead is called through the CRM, the report can show where money went but not which leads were worth it. That is the gap this setup closes first.

Check your own weekly report in ten minutes

This section is advice for your own account. Break last month's spend down by age and placement and look for any group that took spend and returned no leads. Then pick ten recent leads and check whether each one has a recorded first call. If several do not, fix that before you judge any campaign on lead quality.

Finally, if AI agents read your task board, search it for duplicate tasks and merge them before you add more automation.

Frequently asked questions

Why does call recording matter when judging Meta lead quality?

A lead count only says that a form was filled. A recording shows whether the number worked, whether the person answered and what they wanted. Here, leads sent by email instead of called through the CRM left no recordings, so quality could not be verified and the report could not be finished.

What should a weekly ad report for a mortgage team include?

At minimum, spend and leads by campaign from the ad account, pipeline movement from the CRM and loan status from the loan system. This team planned to merge those three sources in one review so each campaign can be read against what happened to its leads afterward.

Why would duplicate tasks burn tokens when AI agents read a project board?

An agent that reads the board treats a duplicate as separate work. Several agents can then try to handle the same item, collide with one another and spend tokens for no benefit. Cleaning the board before adding automation avoids that waste.

Can I limit an age group on mortgage ads?

Meta applies extra restrictions to some targeting options for credit-related ads, so check what your account allows before you rely on a cap. Even where you cannot limit an age group directly, the breakdown still shows where spend went, and that can guide your creative and audience choices.

If your weekly report cannot tell a good lead from a bad one, book a 30-minute AI follow-up and reporting audit. We look at where your ad, CRM and call data stop lining up and show you the report we would build.