Twice the qualified calls, by the book
A pay-per-call insurance campaign had plenty of calls, but most of them weren't worth taking. We rebuilt the targeting, added automated pre-qualification and vetted every traffic source, within US federal and state rules. The qualified call rate went from 22% to 44%, and revenue per call doubled.
- Client
- Insurance lead generation company, USA
- Model
- Pay-per-call: buyers pay for each qualified inbound call
- What we did
- Targeting, phone pre-qualification, publisher vetting, compliance
- Timeline
- 12 weeks, then weekly reviews
The client asked us not to name it. The figures come from its call tracking and the buyers' own reports, and we can arrange a reference call under NDA.
Lots of calls, few worth taking
In pay-per-call, buyers only pay for calls that meet their criteria. Every other call costs money to generate and earns nothing.
The wrong callers
People outside the states the buyer's agents could serve, people already insured, and people looking for a different product.
Paying for rejected calls
Every unqualified call was paid for at the ad, then rejected by the buyer. Volume looked healthy while margins shrank.
Risky traffic
Some traffic sources used claims that didn't match what callers found on the phone. That's a quality problem and a legal one.
Where 100 calls went, before and after
Switch between before and after. Green squares are qualified calls the buyer is happy to pay for.
Illustrative split, based on the campaign's call dispositions. The qualified rate went from 22% to 44%.
Better calls, not more calls
Three changes, each aimed at the same thing: only the right callers reach a licensed agent, and every one of them arrives the right way.
Ads that run only where and when a sale is possible
We rebuilt the campaigns around three limits: the states the buyer's agents are licensed in, the hours they can answer, and the search terms of people who actually want a quote.
Illustrative, based on the campaign's call data.
A short automated screen before any agent picks up
Three questions by key press: the type of cover, the ZIP code and whether the caller is shopping now. The screen only routes the call. Prices and cover are for licensed agents only. Pick a caller to see where they go.
- Insurance typeAuto
- ZIP code43215, licensed state
- Shopping now?Yes
Only a licensed agent discusses cover or prices. The automated steps only route the call.
Every traffic source scored, every week
Each publisher was measured on call quality, and their ads and consent records were checked. Sources that sent poor calls or made misleading claims were removed, even when they sent a lot of volume.
| Source | Qualified | Ad or consent flags | Status |
|---|---|---|---|
| Publisher ASearch ads | 51% | 0 | Scaled up |
| Publisher BComparison site | 47% | 0 | Kept |
| Publisher CSearch ads | 38% | 0 | Kept |
| Publisher DSocial video | 29% | 1 | On review |
| Publisher EDisplay network | 12% | 3 | Removed |
| Publisher FIncentivised traffic | 6% | 4 | Removed |
Anonymised and illustrative. Flags include misleading ad claims, calls without a clear reason to call, and missing consent records.
Built around US calling rules
US rules on calls, recording and insurance sales are strict, and they differ by state. We mapped every step of the call against them before changing anything. Pick a stage to see the rules and what the campaign did.
A summary of how this campaign was designed, checked against the rules in force in September 2026. It is not legal advice. Anyone running a calling campaign in the US should take advice from a US lawyer.
The qualified rate, doubled
From 22% to 44% of calls qualified, and revenue per call doubled. Fewer calls were wasted on both sides: callers reached someone who could help, and buyers paid for calls they wanted.
"We used to celebrate call volume. Now we look at how many calls a buyer is glad to pay for, and whether we'd be comfortable playing every one of them to a regulator."
Compliance first, then the changes
Twelve weeks from the first recording to a fully tuned campaign.
- Weeks 1–2ListenHundreds of call recordings and every call disposition reviewed, to see why calls were rejected.
- Weeks 3–4Compliance mapEvery step of the call checked against federal and state rules, before any change went live.
- Weeks 5–7BuildThe pre-qualification flow, routing by state and licence, and the recording and consent records.
- Weeks 8–10Targeting and publishersAds rebuilt by state, hour and search term. Every traffic source scored and reviewed.
- Weeks 11–12Launch and tuneLive in stages, with the buyer confirming call quality every week.
From key presses to a voice agent
This project used key presses. Today we'd use a voice agent for the screening, on inbound calls only, with the same rules: it says it's automated, it only routes, and people sell.
Callers say what they need in their own words. It still says it's automated, and still only routes.
Each call scored while it happens, so a strong lead reaches an agent faster.
The licensed agent sees what the caller already said, so nobody has to repeat themselves.
Recording notice, consent and opt-outs checked automatically, not just in spot checks.
Outbound calls with an AI voice need prior express consent, and prior express written consent for marketing, because the FCC treats AI-generated voices as artificial under the TCPA.
Quality and compliance go together
Any business that takes calls from ads, in any market, can use the same approach: measure the calls you want, and design the rules in from the start.
Doubling the qualified rate did more for revenue than any increase in volume would have.
The sources with misleading ads were also the ones sending the worst calls. Removing them helped on both fronts.
The automated steps sort calls. Only licensed agents talk about cover and prices.
Recordings, consent and opt-outs kept for years. If a question ever comes, the answer is on file.
How many of your calls are worth taking?
We'll look at where your calls come from, how they're screened and what happens next, and show you where quality and compliance can both improve.
Start a conversation