+63% Qualified Leads | –38% Cost per Qualified Acquisition | +29% Policy Bind Rate | 120 Days
Regional insurance carrier writing home, auto, life, and bundled policies across six states. The team had steady inbound from referrals and partners, but paid media had become the only scalable lever they could control week to week.
Google Ads was producing volume at a cost. Meta was producing leads that did not consistently convert into quotes and binds. Reporting stopped at “lead submitted,” which meant both platforms were optimizing on incomplete information.
Budget was concentrated in broad, high-volume terms that reliably trigger comparison behavior and national-carrier auctions.
Examples observed in the query mix:
CPC was inflated by category pressure. A large share of users were evaluating options, collecting quotes, and moving on. The bidding system was trained on form submits, so it kept finding more users who submit forms cheaply, not users who bind policies.
Meta was running generic lead-gen campaigns with broad targeting and general savings creative. Cost per lead looked attractive. Downstream outcomes did not.
The gap was visible in the sales data:
Meta had no way to distinguish a lead that binds from a lead that fills a form and disappears.
There was no closed-loop signal. Leads, quotes, and bound policies were not being sent back into either platform in a structured way. That left both algorithms blind and pushed decision-making back onto gut feel.
This engagement was not “run ads better.” It was a rebuild of the acquisition system so the platforms could optimize toward the same definition of success the business uses.
We split the account by product line and intent so budget could be controlled and performance could be read without guessing.
This removed the blend of traffic types that was making the account look “fine” at a blended CPA while draining budget on weak-intent clicks.
Common patterns:
We moved budget toward terms that indicate active quote intent, with state localization where applicable:
Auto:
Home:
Bundle:
Life:
Research traffic stayed in the account, but it was separated, capped, and used to support remarketing pools:
We tightened query control to reduce waste and stabilize auction behavior. The negative list was built around three buckets:
Competitor terms were handled deliberately. If competitor traffic was not part of a defined strategy, it was blocked. If it was part of a strategy, it lived in its own isolated campaign with its own bid ceilings and landing path.
The account was previously optimizing toward form submissions. We replaced that with staged offline outcomes.
We implemented offline conversion uploads for:
Conversion values were weighted to match business reality:
Once those events were feeding back into Google, the bidding system started making different decisions. That showed up first in query mix quality, then in CPA, then in bind rate.
Meta was rebuilt to do two jobs:
We separated campaigns by policy line so creative, audiences, and optimization signals were not mixed.
Geo was limited to licensed states. Audience construction leaned on broad and lookalike approaches, but the seed sources were tied to outcomes, not leads.
Lookalike sources used:
We excluded:
Creative was rebuilt so each policy line had its own message and reason to act. Generic “save today” creative was removed.
Auto creative examples:
Home creative examples:
Bundle creative examples:
Life creative examples:
Formats used:
Retargeting was split by intent depth.
The retargeting creative referenced continuation and completion. It did not restart the conversation from scratch.
This is where performance separated from “lead gen.”
Both platforms received offline outcomes, not just leads. Lead quality was defined in a way that sales agreed with.
Those events were uploaded to the ad platforms on a consistent schedule so learning was stable.
The quote funnel was reducing conversion quality and wasting sales time.
We rebuilt the flow into a multi-step structure:
Step 1: minimal qualification and product selection
Step 2: policy-specific details
Step 3: contact and underwriting completion
Conditional logic filtered out non-target scenarios earlier. Commercial requests were routed correctly instead of polluting personal lines.
Mobile speed was improved. Form completion rate increased. Sales saw fewer dead-end conversations.
The paid system became readable. Budget could be moved with confidence because intent tiers were separated, outcome signals were fed back into the platforms, and funnel behavior matched underwriting reality.
