Google Ads and Meta are essentially giant machine-learning engines. They want to find you 'conversions' as cheaply as possible. But if you only tell them when a call *happens*, they will find you lots of people who like to call but never buy. To lower your Customer Acquisition Cost (CAC), you must only report 'High-Value' signals.
Most marketers get conversion tracking exactly half right. They wire up the pixel, they fire a conversion on every lead, they move on. That setup tells the algorithm one thing: 'find more people who produce events that look like this one.' If 'this one' is every call regardless of quality, the algorithm faithfully finds you more callers—and many of those callers are noise. The fix isn't more tracking. It's better tracking.
Selective Signal Reporting
The 'Hijack' works by delaying the conversion signal. Instead of reporting the call at the start, CallMatrix waits for the disposition. If the call was a 'No-Interest' or 'Wrong-Number,' we suppress the signal. If the call was a 'Closed-Sale,' we report it to Google with a high conversion value. This forces the algorithm to narrow its focus only on the subset of users who actually close.
The delayed-signal approach works because Google Ads (and Meta) support offline conversion imports that match by click ID back to the original ad event. You get the same attribution credit you would have gotten from an instant pixel fire, but now the conversion is tagged with 'Closed-Won at 4,500' instead of 'call placed.' Smart Bidding sees a real-value signal and optimizes for clicks that look demographically and behaviorally similar to that 4,500-value cluster.
Value-Based Bidding vs. Conversion-Based Bidding
There are two bidding configurations the algorithm supports. Conversion-based (Target CPA) treats every uploaded conversion as equal and tries to minimize cost per event. Value-based (Target ROAS) treats each conversion with its associated dollar value and tries to maximize total value per dollar of ad spend. For call marketplaces with variable deal sizes, value-based bidding is almost always the right choice—you're explicitly telling the algorithm that a 5,000 legal retainer is worth 50x more than a 100 lead fee.
The trap is that value-based bidding needs volume to work. Google recommends at least 30 conversions per campaign per month before vCPA / tROAS starts producing stable bids. Under that threshold, the algorithm has too few samples to learn from and the bids oscillate wildly. If you're running a low-volume campaign, stick to Target CPA with quality filtering and let value-based roll out once you cross the volume threshold.
Meta Conversion API (CAPI) Integration
Meta's browser pixel is increasingly unreliable in a post-ATT world. The answer is the Conversion API (CAPI)—server-side event delivery that bypasses the browser entirely. CallMatrix pushes call-disposition events to Meta CAPI with the same value signals: `lead` events for qualified calls, `purchase` events for closed-won deals, with the original `fbclid` from the DNI session attached for attribution. This lifts measured Meta conversions by typically 15 to 25% because you're now seeing the iOS cohort that the browser pixel never captured.
Testing the Hijack: A Controlled Rollout
Don't enable selective signal reporting across every campaign at once. Pick one campaign with a stable baseline (30+ days of consistent performance), switch it to the new conversion action, and hold the others as control. Run for 21 days minimum—the algorithm needs 3 weekly cycles to stabilize bids against a new signal. The treatment campaign should show declining CAC and improving quality by week 2. If it doesn't, your disposition data is too sparse or too noisy; fix the upstream signal before scaling the test.
Track three metrics during the test: CPA (expect to rise in week 1 as the algorithm repriced), close rate (expect to rise by week 2), and total revenue (expect to rise by week 3). If you see CPA fall but revenue also fall, you're optimizing for the wrong signal. If CPA rises steadily without revenue catching up, your signal volume is too low.
Ad algorithms are a market. You pay for the demand they create. The operators who pay least are the ones who train the algorithm sharply on their highest-value segment. Everyone else is subsidizing a model that hasn't learned to distinguish their good leads from their bad ones. Do the work of tuning your signal and Google will happily find you cheaper traffic—because that's exactly what the algorithm was built to do.