Count what pays. Not what rings.
Call counts lie. Attribution is a judgment call. This page is how we think about measurement: what we measure, what gets credit and why, how tracking data feeds back into ad decisions, and the tracking failures we see most. Part of our Google PPC practice.
The question tracking must answer
Every tracking setup exists to answer one question: what made the phone ring, and which calls made money?
Most tracking setups answer a different question: how many calls happened. That number is trivia. A hundred calls mean nothing if eighty were wrong numbers, spam, and price shoppers from outside your service area. The number that matters is how many calls turned into qualified prospects, and how many of those turned into revenue, and which ads produced them.
So we design tracking backward from the decision. What decision will this data drive? If the data cannot change a budget, a bid, or a keyword, it is decoration. Every number we track has a job: kill this, keep this, scale this, fix this. Tracking without decisions is just a more expensive way to feel informed.
This is a methodology page, not a sales page. The technical implementation, the part where your CRM outcomes flow back into Google Ads so the bidding optimizes on revenue, is our offline conversion tracking service. What follows is the thinking that implementation is built on.
What we measure: three levels
Calls, qualified calls, revenue. Each level is harder to measure and more valuable than the last.
Level 01Calls
The raw count, tied to the click, keyword, campaign, and ad that produced it through dynamic number insertion. Necessary but misleading on its own: it counts the spam, the wrong numbers, the eight-second hangups. Level one tells you the phone works. It tells you nothing about the business.
Level 02Qualified calls
Calls filtered by duration thresholds, spam rules, and outcome tagging. A real conversation with a real prospect in your service area who needs what you sell. This is the level most businesses never reach, and it is the level where ad decisions start getting honest. A keyword that produces ten calls and one qualified prospect is worse than a keyword that produces three calls and three qualified prospects. Level one cannot see that. Level two can.
Level 03Revenue
Which calls turned into booked jobs and closed dollars. This requires the outcome recorded somewhere, a CRM or at minimum a disciplined process, matched back to the call and the click. It is the hardest level to build and the only one that ends arguments about which campaign is "working." Everything below it is a proxy. This is the thing itself.
The discipline is to never let a lower level make a decision that belongs to a higher one. Do not kill a keyword because its call count is low when its qualified-call rate is high. Do not scale a campaign because it rings a lot when none of the calls close. Each level corrects the level below it.
Our own junk-car operation measures to the revenue level: every dollar tracked to what it actually produced. That operation bought 3,000+ cars in a year and grew operations and profit 4x. And 954 Garage Door runs its marketing on the same discipline: $45,000+ per month in revenue, with every marketing dollar accountable to what it produced.
Attribution thinking: what gets credit and why
Attribution is not a setting. It is a judgment about how your customers buy.
Every attribution model is an opinion about the customer journey. Last-click says the final touch did all the work. First-click says the first touch deserves everything. Data-driven splits credit by pattern. None of them is true in any absolute sense. Each is a lens, and the right lens depends on how your customers actually behave.
Short, urgent decisions lean last-click: the burst pipe does not involve a week of research, and crediting the final search is honest. Long, considered decisions need earlier touches credited: the homeowner who researched for a month before calling was influenced by more than the last ad. Picking the model is a business judgment, not a default to accept, and we revisit it as the data grows.
Phones break attribution, and we plan for that. A customer clicks an ad on Monday, calls on Tuesday from a different device, and books on Friday. The click path is fragmented. We attribute conservatively: credit what we can prove, flag what we cannot, and never let an unattributed conversion secretly inflate a campaign's apparent performance. Honest gaps beat confident fiction.
We also separate assists from closers. Some keywords introduce customers who convert later through other channels. Killing every "assist" keyword because it never closes directly is how you strangle the top of your own funnel. Attribution thinking means knowing which keywords close, which assist, and funding both on purpose.
How tracking feeds ad decisions
Data that does not change a decision is decoration. Here is the decision loop.
Kill, keep, scale
Every keyword, ad group, and campaign gets one of three verdicts, on a schedule. Kill: qualified-call rate too low or cost per qualified call too high, cut it and reallocate. Keep: performing within targets, leave it alone and let it learn. Scale: strong qualified volume at good economics, feed it more budget in steps. The verdicts come from level-two and level-three data, never from raw call counts.
Keyword-level call outcomes
When calls are tagged by outcome and tied to keywords, patterns emerge that click data hides. The keyword with the cheapest clicks produces the worst callers. The "expensive" keyword produces the buyers. Bidding follows the outcomes, not the click prices. This is where most accounts leave their biggest gains on the table: they bid on click economics instead of customer economics.
Geography and schedule insights
Call tracking shows which cities and which hours produce qualified calls, not just clicks. The market that clicks a lot and calls a little gets its budget cut. The hours where callers actually convert get protected. Over time the account reshapes itself around reality: money follows the calls that close, and everything else starves. That reshaping is the whole point of measuring.
Common tracking failures
We see the same five failures in almost every audit. Check yours against this list.
1. No dynamic number insertion
One static number on the site means every call is anonymous. You know the phone rang. You do not know which campaign, keyword, or ad caused it. Every budget decision becomes a guess. Dynamic number insertion is the foundation everything else stands on, and its absence is the most common failure we find.
2. Counting every ring as a conversion
Spam, wrong numbers, eight-second hangups: all counted as conversions, all fed to Google's bidding as success. The algorithm then optimizes toward producing more of them. A tracking setup that cannot distinguish a buyer from a robocall is actively training your account to waste money.
3. Tracking numbers in the wrong places
Tracking numbers belong on the website for visitors. They do not belong on the Google Business Profile, in directories, or in citations, where they break NAP consistency and confuse local rankings. Done right, dynamic insertion swaps numbers only for website visitors while the real number stays everywhere else.
4. Calls measured, outcomes ignored
The business tracks that calls happened but never records which ones became customers. Marketing and sales then argue about lead quality forever, because there is no shared record of what a good lead was. The outcome has to be written down somewhere, or the loop stays open and the data stays trivia.
5. Call extensions counted as conversions
Someone taps the call button on the ad and hangs up three seconds later. Google counts it. The dashboard looks healthy. The calendar is empty. Call extensions are useful, but their taps are not conversions until duration and outcome rules say so.
The methodology above is how we think. The build that makes it run automatically, connecting your CRM outcomes back into Google Ads so bidding optimizes on revenue, is offline conversion tracking.
Questions
Straight answers, before you apply.
Measure where the money is. If forms drive revenue, measure forms deeply. But most local service businesses discover that a large share of their real conversions happen on the phone, and those were invisible before call tracking. You do not know the split until you measure both.
The one that matches how your customers actually buy. Short, urgent decisions lean last-click. Long, considered decisions need to credit the earlier touches. The model is a judgment call, not a default setting, and we revisit it as the data grows.
Not when done right. Dynamic number insertion swaps numbers only for website visitors, while your real number stays on your Google Business Profile, citations, and directories. NAP consistency is preserved.
Duration thresholds, spam filtering, and outcome tagging separate real prospects from noise before anything gets counted as a conversion. A tracking setup that counts every ring as a lead is worse than no tracking, because it teaches the bidding to chase noise.
For call tracking itself, yes: calls, recordings, and outcomes can be tagged inside the call tracking platform. Connecting calls to closed revenue needs the outcome recorded somewhere, and a CRM is the best place. A disciplined process beats a fancy tool, but you need the process.
This page is the methodology: how we think about measurement. Offline conversion tracking is the implementation: the technical build that connects your CRM outcomes back into Google Ads so the bidding optimizes on revenue. The thinking comes first; the build makes it run automatically.
Measure what pays.
Decide from truth.
Apply below. We will set up tracking that answers the only question that matters: what made the phone ring, and which calls made money.