How to track call conversions without losing money on bad traffic

Simple. Not easy.

Call tracking sounds like a solved problem. Buy a platform, slap a number on your landing page, watch the conversions roll in. In practice, most businesses that turn it on end up doing one of two things: counting garbage calls as wins or paying real money to answer robocalls at $0.06 a minute. Both quietly eat your margin.

I've watched this play out hundreds of times while building Ringba. The businesses that make call tracking pay for itself aren't the ones with the fanciest setup; they're the ones who decided, early and in writing, what a good call actually is, because every other choice flows from that one decision.

Here's how to do it properly.

Start with the money math

Before you track anything, know what tracking costs.

Entry-level plans on CallRail, CallTrackingMetrics, WhatConverts, Invoca, and Ringba run $45 to $150 per month, with individual numbers adding roughly $2 to $5 each and usage fees around $0.04 to $0.10 per minute. None of that is scary alone. At volume, it adds up fast.

Say you're routing 2,000 calls a month averaging four minutes each. That's 8,000 minutes, or $480 at $0.06 per minute, before you've counted a single conversion. If 30% of those calls are spam, hang-ups, and wrong numbers, you just spent about $144 answering noise, and worse, you may have told your ad platform the noise was valuable.

That last part is where the real losses hide. Per-minute fees sting. Bad optimization compounds.

Define a qualified call before you track anything

People skip this section. It's the whole game.

A ringing phone isn't a conversion. A conversation is. So set a minimum call duration before any call counts. The common range is 60 to 90 seconds, and it exists for a reason: wrong numbers and instant hang-ups almost never last a full minute, while a real prospect asking about pricing almost always does. Without that filter, your dashboards inflate and your bidding decisions get built on sand.

If you're buying pay-per-call affiliate traffic, this becomes a contract term, not a setting. That traffic often includes incentivized or robocall-driven volume, so experienced buyers negotiate payable criteria up front, usually a 90 to 120 second minimum duration plus limits on geography and business hours, and because payouts run $10 to $100 or more per qualified call depending on the vertical, the definition of "qualified" is worth real money on every single call. I wrote about this economy at length in The Pay Per Call Revolution. The buyers who define their terms tightly are the ones still in business three years later.

My opinion, for what it's worth: 60 seconds is too lenient for most service businesses. Start at 90. Adjust down only if real leads are getting filtered out, which you can check by listening. More on that shortly.

Set up attribution the right way

Dynamic Number Insertion, or DNI, is the standard method. Your site swaps the displayed number based on where the visitor came from, so a Google Ads click sees one number and a Facebook click sees another. Done well, it ties a phone call back to a specific visit, keyword, and campaign.

The catch is number pools. Each concurrent visitor needs a unique number, so a low-traffic site generally needs 4 to 10 numbers, while a site pulling thousands of daily sessions often needs 20 or more. Size your pool to your traffic and no bigger. An oversized pool is a small monthly leak. An undersized one silently misattributes calls, which is worse.

One mistake I see constantly costs people rankings rather than dollars, so it goes unnoticed for months: putting a tracking number as the primary phone number on your Google Business Profile, which fragments your NAP consistency (name, address, phone) across the directory ecosystem that local rankings lean on. Keep your real number as the primary listing. Use call tracking on paid landing pages, or use a platform feature that keeps the original number as a secondary entry.

A quiet rule that saves a lot of grief: tracking numbers belong on media you control, not on the public record of your business.

How do you keep bad calls from wrecking your ad optimization?

Feed only qualified calls back to your ad platform. Google Ads lets you import call conversions with a duration requirement, so if you send Smart Bidding only calls lasting 90 seconds or more, the algorithm learns what a real customer looks like instead of chasing whatever generates short, worthless rings.

Smart Bidding is obedient, not smart. It optimizes toward whatever you call a conversion. Tell it a 12-second robocall is a win and it will happily find you thousands more, spending your budget with genuine enthusiasm the whole time, because from its perspective it's doing exactly what you asked.

The fix takes an afternoon. Set your duration threshold, connect the conversion import to Google Ads, confirm shorter calls are excluded, then give the algorithm two to four weeks to relearn. Reported conversion volume will drop. That's not a problem. That's the truth arriving.

The dashboards look worse and the bank account looks better. I'll take that trade every time.

Block the junk before it costs you minutes

Spam filtering isn't glamorous, but at scale it's hundreds of dollars a month. Block anonymous callers, lean on the built-in spam detection in platforms like CallRail or CallTrackingMetrics, and for suspect traffic turn on keypress verification, where the caller has to press 1 to connect, since robocallers rarely do. Every filtered call is per-minute fees you didn't pay and a false conversion that never touched your data.

Then verify with your ears. Review recordings or transcripts weekly, even a sample of 10 to 20 calls. Duration thresholds are a proxy for quality, not a guarantee. A weekly listening session tells you whether your 90-second calls are actually prospects, whether your filters are catching real people by mistake, and often, whether your team is answering the phone well at all. I've seen that last discovery pay for the entire stack. Honestly, it's a little humbling every time.

Twenty minutes of listening beats any report I've ever built.

FAQ

How many tracking numbers do I actually need? For session-level DNI, 4 to 10 covers most low-traffic sites, while sites with thousands of daily sessions often need 20 or more since each concurrent visitor needs a unique number. Start small. Let your platform tell you when the pool runs dry.

What duration threshold should I set? Start at 90 seconds for service businesses and listen to a sample of filtered calls for two weeks. Real prospects getting excluded? Drop toward 60. Junk still slipping through? Raise it.

Will call tracking hurt my local SEO? Only if a tracking number becomes the primary number on your Google Business Profile and directory listings. Keep your real number there and restrict tracking numbers to paid landing pages.

Is call tracking worth it for a small business? If phone calls drive revenue and you spend more than $1,000 a month on ads, yes. A $45 to $150 plan that reveals which half of your spend produces real conversations pays for itself fast.

What if my ad platform's call reporting disagrees with my tracking platform? Trust the one with duration data and recordings. Ad platforms count call interactions; your tracking platform counts conversations. The gap between them is usually your junk calls, which is exactly what you're trying to see.

If you're setting this up for the first time, do one thing this week: write down your definition of a qualified call, including minimum duration, geography, and business hours, before you buy anything. Every decision after that gets easy. I write more about call economics and performance marketing at adamyoung.com if you want to go deeper.