Conversion tracking setup before you spend a dollar

This is the step people skip because it is technical, unglamorous, and produces nothing on the day you do it. It is also the step that determines whether the next ninety days of spend teach you anything.

The cost of skipping it is not just bad reporting. Google's automated bidding needs conversion signals to work. Without them it optimizes for clicks, which is the metric you least want maximized.

What counts as a conversion for a local business?

Anything that means a real person is trying to hire you. Usually four things.

Form submissions, fired on the confirmation, not on the button click, so you count completions rather than attempts.

Phone calls from the website, tracked with a dynamic number so you can attribute the call to the ad that produced it. Set a minimum call length that matches a real conversation, or you will count wrong numbers as wins.

Clicks to call from the ad itself, which is a separate conversion from a website call and often the larger one on mobile.

Booking or scheduling completions, if you use a scheduler.

The phone deserves more weight than most tracking builds give it. Patient Prism's Dental Patient Access Report, covering 11,552,668 patient calls across 8,280 dental locations in 2025, found that 90% of attributable patient calls came from the Google ecosystem, with the Google Business Profile alone accounting for 54% and organic plus Google Ads another 36%. Meta and print combined came to under 3%. That is dental specific, and Patient Prism is a call intelligence vendor whose customers skew toward more sophisticated practices, so say both things when you use it. The structural point still holds for most local categories: the call is the conversion, and a tracking build that only watches the form is watching the smaller half.

Do not count newsletter signups, PDF downloads, or time on page as conversions in a local ads account. They are activity, not intent, and feeding them to the bidding algorithm teaches it to find people who browse.

Mark one conversion as primary and let everything else be secondary. If three different actions are all primary, the bidding cannot tell which one you actually want.

What does tracking let me see that spend alone does not?

Run the arithmetic on a real month and the answer is obvious.

Start with the account view you get without tracking. You spend $2,000. LocaliQ's 2026 search advertising benchmarks, published under the WordStream name, put the average cost per click across all industries at $5.42, so that is roughly 369 clicks (derived). You know the spend and the clicks, and nothing else. LocaliQ sells advertising management and its published averages are medians across a large campaign sample, so treat those figures as an estimate rather than your price.

Now add conversion counting. At LocaliQ's 8.18% all-industry conversion rate those clicks produce about 30 leads, at a cost per lead of $67 derived, which lands within a dollar of the published $66.69 benchmark. Reassuring, and still not actionable.

Now split it by campaign, which is the entire point. Suppose campaign A took 200 clicks and produced 22 leads, a cost per lead of $49 derived. Campaign B took the remaining 169 clicks and produced 8 leads, a cost per lead of $114 derived. The account average of $67 describes neither campaign. One of them is funding the other, and without tracking you would have moved the budget in whichever direction felt right.

That is a two-minute calculation once the tags fire and an impossible one before they do. Rerun it with your own spend, your own click count and your own lead count per campaign.

What breaks most often, and how do I know?

Four failures, and they all look identical from the outside, which is an account reporting fewer conversions than the phone suggests.

Tags that fire on the wrong page, usually the form page rather than the confirmation, which inflates your numbers with people who never submitted.

Double counting, where the same lead is recorded through two systems and your cost per lead looks half what it is.

Tags dropped in a website update. This is the most common one and the least noticed, because nothing visibly breaks. The site looks fine and the account goes blind on a Tuesday.

Call tracking that replaces your number in some places and not others, so a portion of calls arrive unattributed.

There is a fifth that sits outside your site entirely. Sterling Sky's 2026 field data, using Jepto tracking across 179 Google Business Profiles for 34 US law firms over two years, found clicks to call from Google Business Profiles declining even where rankings held steady, while website clicks held up better. Those profile calls never touch your website, so no website tag will ever see them. If your call volume is falling and your tracked conversions are flat, look at the profile before you look at the tags.

The check is simple and nobody does it. Submit your own form. Call your own number. Then open the account tomorrow and confirm both appear. Do that after every website change, every time. It takes four minutes.

What can I not do without it?

Almost everything that makes an account worth managing.

You cannot pause anything with confidence. Deciding when to pause a campaign instead of optimizing it requires knowing which campaigns produced customers, and click data will not tell you.

You cannot use automated bidding well, since the algorithms optimize toward conversions and there are none to see.

You cannot allocate budget honestly. Spend flows toward whatever looks busiest rather than whatever works, and that is often how a month runs dry early. Budget pacing problems are worse in an untracked account because you have no basis for deciding what deserved the money.

You cannot hold a manager accountable. If you are paying a monthly fee, the reporting is most of what you are buying, and what management should cost depends on what the fee covers, tracking included or not.

And you cannot compare channels. If you run both search ads and Local Services, deciding between Local Services Ads and Google Ads requires a common measure of what a lead cost in each. Without tracking you are comparing a number to a feeling.

Is tracking enough on its own?

No, and this is where owners of small local businesses get oversold. Tracking has two hard limits and both are worth knowing before you buy a reporting package built on top of it.

The first limit is volume. Tracking gives you counts. It does not give you statistical certainty, and at local volume it never will. The sample a valid test needs is set by the metric's variance and by how small a change you want to catch, and it is worth deriving once so you never have to take anybody's word for it. At 95% confidence and 80% power, the two-proportion formula carries a constant of 2 × (1.96 + 0.84)², which is 15.68, call it 16. Divide that constant by the square of the relative improvement you want to detect and you have the conversions needed per variant. Catching a 40% lift needs 100 per variant (derived). Catching 20% needs 400 derived. Catching 10% needs 1,600 derived. At 300 to 800 monthly visits and a 3% conversion rate you produce nine to twenty-four leads a month, so even the crudest of those is years away, which is why A/B testing does not work for most local businesses.

Notice what fell out of that arithmetic. The "you need 100 conversions per variant" rule somebody quoted at you is the answer to one specific question, and the question is whether you can detect a 40% improvement. Nothing you do to an ad produces 40%. Kohavi, Deng, Longbotham and Xu, writing in the KDD 2014 proceedings from thousands of controlled experiments at Microsoft and LinkedIn, put successful changes at Bing at 0.1% to 1.0% once diluted to overall impact. Its companion rule, the 10,000 monthly visitors, is not a research finding at all. It enters their paper as reference [29], where it resolves to a Neil Patel blog post of 14 January 2013, and the four authors qualify it in the same breath: "the guidance should be refined to the metrics of interest."

The second limit is attribution, and it is getting worse rather than better. Google confirms that traffic from AI features in Search, including AI Overviews and AI Mode, is included in overall search traffic under the ordinary Web search type. There is no referral parameter, no separate channel, and no way to distinguish those visits from any other organic click in your analytics. Search Console added Search Generative AI performance reports in June 2026, and they publish impressions, pages, countries, devices and dates. They explicitly do not publish clicks, click-through rate, position, or the prompt the user typed, and they do not separate AI Mode from AI Overviews.

So a growing share of your unpaid traffic is arriving in a way nobody can label. That is not an argument against tracking. It is an argument for being precise about what tracking proves. Your ads account can tell you what a paid lead cost. Nothing can currently tell you what an AI answer sent you. Any vendor selling you a number for the second one is selling an estimate with a confident font.

Use tracking for the decisions it can support. Which campaigns produce leads. Which keywords produce none. Whether the total is going up or down over a quarter. Those are directional questions and the data answers them well.

Do not use it to declare a winner between two headlines after nine leads. That is where numbers start doing harm.

What to do this week

Before any campaign goes live, set up four conversion actions in Google Ads. Form completion on the confirmation page, website call with a minimum duration, click to call from the ad, and any booking completion. Mark one primary.

Then test each one yourself. Submit the form. Call the number from your mobile. Tap the call button in a preview. Check the account the next day and confirm each shows up.

Write today's date and what you installed in a note somewhere you will find it. When the numbers change in four months, the first question is always what changed and when.

Put a recurring reminder to retest after every website change. That single habit prevents the most common failure in this entire category.

Questions about whether your tracking is honest?

Email me at eric@seod.com with your website URL and tell me what you currently count as a conversion. I will check what is actually firing on your site from the outside and tell you what is missing, duplicated, or counting the wrong thing.

Takes me about ten minutes and you get the findings in plain language, not a tag audit spreadsheet. If it is already clean, that is a good thing to know before you spend.

Or keep reading more on paid search for local businesses.