GOOGLE ADS CONVERSION TRACKING: THE SETUP ORDER EVERYTHING ELSE RESTS ON

Almost every account problem I'm eventually called in to fix traces back to the same root: the google ads conversion tracking was set up wrong, or set up in the wrong order, and nobody noticed because the campaigns still reported numbers. Numbers that look plausible are the most dangerous thing in a google ads account, because Smart Bidding will happily optimise toward a broken signal and spend real money doing it. This is the one part of the build I refuse to rush, and it's the first thing I audit when something isn't adding up.

The angle of this article isn't "here are the buttons." Google's own interface will walk you through the buttons. What it won't teach you is the sequence, the judgement calls, and the specific setup errors that don't throw an error message but poison everything downstream. Get the tracking right on day one and every later decision, which keyword to cut, which campaign to scale, what to bid, rests on solid ground. Get it wrong and you're optimising toward a mirage.

I want to be precise about why I call this the pillar of the account rather than just another setup task. Every optimisation lever you'll ever pull, bid adjustments, budget shifts across your google ads campaigns, pausing weak keywords, promoting strong ad groups, is a decision made on top of the conversion data underneath it. If that data is inflated, missing, or attributed to the wrong action, you're not making bad decisions because you lack skill; you're making them because the ground you're standing on is tilted. That's the mental model I carry into every build: the tracking isn't reporting, it's the foundation the reporting and the bidding both stand on.

There's also a second reason it deserves this much care. The number of conversions you feed the system directly shapes how fast and how well automated bidding learns. A clean, correctly counted signal lets the algorithm find patterns among your real buyers; a noisy or doubled signal teaches it to chase the wrong people. So when I talk about setting up conversion tracking properly, I'm really talking about protecting the quality of every future decision the account will make on its own. Everything below is written from that stance.

The order I build google ads conversion tracking in

Order matters more than people expect, because each step assumes the one before it is already correct. When I set up conversion tracking in google ads from zero, I never start by creating conversion actions. I start by mapping what actually counts as a conversion for this specific business, on paper, before I touch the account. A lead-gen client and an ecommerce store need completely different definitions, and if I skip this I end up with a pile of half-relevant actions I have to untangle later.

Only then do I move into the account. Google frames the flow as choosing where conversions happen, adding a data source, then creating the actions, and their documentation on how conversion measurement works lays out that sequence clearly. I follow it, but with one rule of my own: the measurement foundation, the Google tag, goes in and gets confirmed firing before I define a single action. A conversion action pointing at a tracking code that isn't live is the most common reason a brand-new account reports zero and everyone panics.

My actual build sequence

The sequence I follow every time: define conversions on paper, install and verify the Google tag, create the primary conversion action (the one real business outcome), then layer secondary actions around it, and only after that do I let campaigns go live. Notice campaigns come last. I've seen too many accounts where spend started while tracking was still "nearly done," and those first weeks of blind data are effectively wasted budget the algorithm then has to unlearn. Whether you deploy the tag directly or through Google Tag Manager doesn't change the order, only the plumbing.

Why I install the tag before creating actions

People find this counterintuitive, because the interface nudges you to start by creating a conversion action and then generate the snippet. My reasoning is about verification. If I set up google ads tracking by placing the base tag first and confirming it loads on every page, I've established a trustworthy foundation. Then when I move to creating a conversion action, I already know the tag underneath it is healthy, so any problem I see is a problem with the action itself, not a mystery about whether the whole layer is live. Debugging one variable at a time is far faster than debugging two at once, and this ordering isolates them cleanly.

Where conversions actually get created

One practical note that trips up newcomers: conversion actions in google ads live under the Goals section, specifically the Conversions summary, not inside campaign settings and not under the general Tools menu. I've watched people hunt through campaign options for a "tracking" switch that isn't there. Creating a conversion action always starts from the Goals area, where you pick the source, website, app, phone calls, or offline, and configure how each one is counted. Getting to the right place is half the battle, because the settings that determine how conversions are tracked in google ads all hang off that one workflow. Do it there, once, deliberately, and the rest of the build has somewhere solid to attach to.

Primary vs secondary: the choice that skews everything

If there's one setting inside google ads conversion tracking that quietly wrecks more accounts than any other, it's the primary versus secondary distinction. Primary conversions are the ones that count in your "Conversions" column and that Smart Bidding optimises toward. Secondary conversions are recorded for insight but don't drive bidding. Getting this backwards, or marking everything primary, is how an account ends up bidding hard toward newsletter signups while the actual purchases barely register.

My rule is deliberately strict: one primary action per real business goal, and everything else secondary until proven otherwise. In a lead-gen account, the qualified lead is primary and the softer stuff, PDF downloads, page views, time on site, stays secondary. The temptation is always to promote a soft action because it fires more often and makes the dashboard look busy. That's exactly the trap. More conversion volume from a weaker signal doesn't help the algorithm, it misdirects it, and the bids follow the wrong outcome.

The inference I draw when I open an inherited account and see five primary actions is almost never "this business has five goals." It's that someone created actions without deciding which one mattered, and the bidding has been diluted across all of them ever since. Fixing it usually means demoting most of them to secondary and letting the strategy realign around the one outcome that pays the bills. This is the single lever that most changes what Smart Bidding actually chases, because the algorithm can only optimise toward the signals you tell it are primary.

There's a related decision about where the data comes from that shapes this same choice. Many accounts import their conversion actions from Google Analytics rather than defining a native google ads conversion, and the two paths behave differently for bidding and attribution. I'm not dogmatic about which to use, but I am strict about not mixing them for the same event, because importing a Google Analytics goal that duplicates a native action is a classic way to count the same outcome twice. Whichever source you pick for a given conversion, commit to it, and make sure the same purchase or lead isn't also being recorded through a second path in parallel.

How the primary set shapes everything downstream

Once the primary conversion is clean and singular, the downstream effects compound in your favour. Your reporting column now reflects the outcome that matters, so when you compare campaigns you're comparing them on business value, not vanity. Your bidding chases the right people. And your optimisation cadence, the weekly ritual of cutting and scaling, is grounded in a number that means something. This is why I'll spend an hour arguing about which single action deserves primary status: that one decision quietly governs how tens of thousands of euros get allocated over the following months.

The setup errors that don't throw an error

The dangerous mistakes in google ads conversion tracking are the silent ones, the ones that produce numbers instead of blanks. A blank report gets fixed fast because it's obviously broken. A report that's double-counting looks fine right up until you make a budget decision on it. These are the four I hunt for first in every audit, in roughly the order they do damage.

Counting "Every" vs "One"

For lead gen, a single person filling in a form twice should be one conversion, so the count setting must be "One." For ecommerce, two separate purchases are two conversions, so it's "Every." Set a lead form to "Every" and one indecisive user hitting submit three times inflates your conversions threefold. I've traced entire phantom performance spikes back to nothing more than this one dropdown.

The wrong conversion window

The conversion window is how long after a click a conversion still counts. Too short and you undercount long sales cycles, starving Smart Bidding of data it should have credited; too long and you attribute stale conversions to clicks that had nothing to do with them. I match the window to the real buying cycle, not the 30-day default, because that default is a guess about a business it knows nothing about.

Duplicate firing and double tags

When both a hard-coded tag and a tag-manager tag fire on the same page, every conversion counts twice. This is rampant in accounts that migrated to Google Tag Manager without removing the old snippet, so you end up with a conversion tracking tag deployed twice for the same event. It's also why I verify firing rather than trusting that a tag "should" be there, and it ties directly into clean ecommerce conversion setup where a duplicated purchase event corrupts revenue reporting outright.

Value tracking left empty

If you run value-based bidding but the tag sends no conversion value, every conversion is treated as identically worth the same, and a €10 order looks the same as a €1,000 one. The algorithm can't optimise toward revenue it can't see. For any account chasing ROAS, missing or hard-coded values quietly cap how well the bidding can ever perform.

Forgetting offline conversions exist

The last blind spot is the one people forget because it lives outside the website entirely. Plenty of businesses close the real sale on a phone call or after a sales team follows up a lead, which means the outcome that actually matters never happens on a page a tag can see. If you only track form fills and never import the downstream result, you're optimising toward leads instead of revenue. Bringing offline conversions back into the account, tying a closed deal to the original click, is what closes that loop, and for lead-gen especially it changes which campaigns look profitable once the true value lands. I flag its absence in almost every audit of a service business, because the gap is invisible until you go looking for it.

How I verify before a single euro is spent

Setting up tracking and confirming tracking are two different jobs, and skipping the second is how broken setups reach production. My routine is to trigger the conversion myself, complete a test lead or a test purchase, and watch it land. Google Tag Assistant tells me the tag is firing correctly; the conversion action status in the account tells me it was received and is recording. I want both green before I sign off, because a tag can fire and still be pointed at the wrong action.

Tag Assistant is the tool I lean on most here, because it lets me see the actual network calls the page makes and confirm the conversion event goes out exactly once with the parameters I expect. That "exactly once" check is what catches the duplicate-firing problem before it ever pollutes the account. If I see two identical conversion pings, I know a tag is deployed twice and I hunt it down before anything spends. Watching real traffic through that lens turns "I think it's fine" into "I've seen it fire correctly," which is the only standard I trust.

New conversion actions show a status while they wait for their first real data, and I don't consider the job done until I've seen a status move to recording conversions off a genuine event, not a hopeful assumption. I also give the account a short window and then check that the number of conversions coming in matches what I'd expect from real activity, because a setup that verifies once but then drifts, a site change breaks the tag, a consent banner blocks it, is just as damaging as one that never worked. Verification isn't a single moment; it's a habit you repeat whenever the site or the setup changes.

This whole verification pass is a fixed step in my conversion tracking testing process, and it's non-negotiable. The takeaway of the entire pillar is simple: conversion tracking in google ads isn't a task you finish, it's the foundation you keep honest, because every bid, every budget shift, and every keyword decision downstream is only as trustworthy as the signal underneath it. Build it in the right order, guard the primary/secondary line, kill the silent errors, and verify with a real event, and the rest of the account finally has something solid to stand on. That's the whole job: not clever tricks, just a foundation you can trust, maintained with the discipline it deserves.

A quick recap of the sequence

If you take one thing from this piece, take the order, because the order is what makes the rest almost automatic. Define what counts before you touch anything. Install and confirm the Google tag is live. Then move to creating a conversion action, starting with the single primary outcome that represents real revenue, and only afterward add the softer secondary actions that inform without steering the bidding. Set the count method to match the business, One for leads and Every for purchases. Match the conversion window to the actual buying cycle. Bring offline conversions in wherever the real sale closes off-site. And verify every action fires exactly once with a real test before a campaign spends against it.

Everything I've described exists to protect one thing: the trustworthiness of the conversions in google ads that your automated bidding reads every hour of every day. When people ask me why their account underperforms despite good ads and reasonable budgets, the honest answer is usually that the measurement layer was never built to be trusted, so the machine has been optimising toward a distorted picture the whole time. Fix the foundation and you don't just get cleaner reports, you get a bidding system finally pointed at the outcome you actually care about. That's the return on getting this one thing right, and it's why I always start here.