Open the account and it looks like a good month. Forty-seven conversions. Cost per conversion down from last month. The line on the chart goes up and to the right, which is the direction lines are supposed to go.
Then you think about the actual month. You had four real conversations. Two of them went nowhere. One person wanted a job. Somebody kept calling about a service you stopped offering two years ago.
Forty-seven and four. Both numbers are true. That is the strange part, and it is worth sitting with for a second, because the instinct at this point is to assume the platform is inflating its own homework. It usually is not. Google Ads is not lying to you. It is being literal. It counted forty-seven of the things you told it to count, and it will spend next month's budget going out to find you more of them.
Google Doesn't Know What a Customer Is. It Knows What You Counted.
There is one sentence underneath almost every wasted dollar in a small Google Ads account. Every automated system in the platform, from Smart Bidding to Performance Max, optimizes toward the conversions it can see.
That is the whole mechanism. It looks at which clicks turned into conversions, finds the pattern, and buys more clicks that look like those. It has no opinion about whether the conversion was worth anything. It cannot tell a booked job from a wrong number, because nothing in the account ever told it the difference. We have written before about why accurate tracking is the thing to fix before you hand anything to automation, and that piece makes the case that broken tracking makes a capable machine confidently wrong.
What it does not do is open the hood and show you what broken looks like. That is this article. Because in most of the accounts we audit, tracking is not switched off and it is not obviously failing. It is quietly counting the wrong things, and it has been doing it so consistently that the numbers look stable.
Four Things in Your Conversions Column That Were Never Sales
Go and look at your conversion actions. Not the summary number on the dashboard, the actual list. Here is what tends to be hiding in there.
The phone tap nobody answered. The most common call conversion on a small account fires when somebody taps the phone number on your website. That is the moment of tapping. Not connecting, not talking, not a conversation that lasted long enough to be a conversation. Somebody who tapped by accident on a bumpy bus counts. Somebody who reached your voicemail at 9pm and hung up counts. If you are tracking taps rather than calls of a real duration, a meaningful slice of your call conversions never involved a human voice.
The thank-you page counted every time it loads. If your conversion fires on someone arriving at a confirmation page, ask what happens when that person refreshes. Or hits the back button and forward again. Or bookmarks it, which people genuinely do. Google gives you a setting for exactly this, counting every conversion versus counting one, and lead generation businesses very often have it set to every. One form fill, three conversions, and a cost per conversion that looks a third better than reality.
Every form treated as equal. The quote request is a conversion. So is the newsletter signup, the job application, the supplier who wants to sell you something, and the contact form somebody used to complain. They all land in the same column with the same weight, and the algorithm reads them all as the same instruction: find more people like this.
The same lead counted twice. This one is sneaky because it appears in healthy-looking accounts. A conversion imported from Analytics and a conversion firing from the Google tag on the same form. Two systems, one event, two conversions. Jyll Saskin Gales, who spent six years at Google before going independent, describes exactly this happening to a client mid-upgrade, along with a related discovery that should make everyone check their own site: a phone number that had never been swapped over to the tracking number, quietly not being measured at all.
A Form Is Not a Lead. A Lead Is Someone Who Could Buy.
Tom Bowen, a UK Google Ads specialist who works on exactly this problem, puts it in six words: a form is not a lead.
It sounds obvious written down and it is not obvious in practice, because every reporting tool you own is built on the opposite assumption. Picture two form fills arriving an hour apart. The first is from somebody in your service area, describing the job, with a real phone number. The second has a typo'd email that bounces the moment you reply. Traditional conversion tracking swears these two events are identical. It has no field for the difference.
Ask any owner how their lead quality is doing and you will get a feeling rather than a number. It felt slow this month. The last few weeks have not been great. That feeling is usually accurate, which is the frustrating part, and it is also completely unusable, because you cannot optimize a campaign against a mood.
Primary and Secondary Is the Setting Nobody Touches
Before anything else, there is a setting in your account that costs nothing and takes about ten minutes.
Every conversion action can be marked primary or secondary. Google's own documentation is blunt about the difference: primary actions are the ones used for bidding, and secondary actions are there for observation. They get recorded so you can look at them, and they are not what the bidding goes hunting for.
Most small accounts have everything marked primary, usually because everything was primary by default when it was set up and nobody has been back since. So the newsletter signup is steering your budget with the same authority as the quote request.
Marking the low-intent actions secondary is the single highest-value change available to an owner who does not want to buy anything or hire anyone. Keep the actions that represent somebody genuinely trying to do business with you as primary. Everything else drops to secondary, where you can still see it. Then set the confirmation-page action to count one conversion per click instead of every conversion, and go and check that every phone number on your site is the tracked one.
That is an afternoon. It is not the whole answer, and it is most of the way to an honest number.
The Fix That Looks Like a Downgrade
Here is the part nobody warns you about. When you start counting honestly, the number goes down.
Gales describes a business-to-business client who was reporting sixty to seventy conversions a month, most of them spam or hobbyists buying a single part, none of them the kind of order the business needed. After moving to counting only qualified leads, they landed at about twenty a month. Her comment on that drop is the whole point of this article: yes, it went down, and that is a good thing.
If somebody is running your account and this change happens, the monthly report is going to look worse. Conversions down two thirds, cost per conversion up. Every instinct says something broke.
Nothing broke. The number stopped describing form submissions and started describing the business. Twenty real conversations is more revenue than seventy mostly-junk ones, and it always was, and the only thing that changed is that the report finally agrees with the bank account.
Worth saying plainly, because it protects you in the other direction too: an agency whose reported conversions are climbing every month while the business feels flat is not necessarily doing anything wrong, but it is a conversation worth having. Ask what is being counted. If nobody can tell you which of last month's conversions turned into work, the number is decoration.
Telling Google Which Leads Were Actually Real
Everything so far is about not counting rubbish. The upgrade is telling Google which leads were good, after the fact.
The industry name for this is offline conversion tracking, which sounds like a project. The idea is simple. When somebody fills in your form, Google records the click but is told nothing yet. Later, once you have spoken to that person and know whether they were a real prospect, you send the answer back. Google then connects that answer to the original click, the keyword, the campaign, the time of day.
Now the algorithm is not optimizing toward form submissions. It is optimizing toward the kind of click that historically turned into a real customer, which is what you wanted from the beginning.
Calls can be qualified the same way. A remodeling company that installs replacement windows but does not repair them can put a short menu on the tracked line: press one for a new installation, press two for a repair. The installation calls register as conversions. The repair calls, which that business cannot serve and never wants to pay for again, never reach Google at all. One agency working with a commercial cleaning company found the same trick separating out the real problem: most of their so-called conversions were people asking about franchise opportunities and jobs.
The last rung is value. Not every good lead is worth the same, and a large commercial inquiry is not a small residential one. Once you send values back rather than a simple yes, the bidding can chase the expensive ones.
Bowen makes an argument here that is worth borrowing because it cuts against the obvious. Do not wait to report the closed sale. Report the value of the quote. Two reasons, and both are practical. A sale can take weeks or months to land, by which time the data is far too stale to teach a bidding algorithm anything useful. And closing the quote is the sales team's job. Judging the ads by whether somebody else closed the deal measures the wrong department.
The Bid Adjustment That Makes Bad Leads Worse
This one deserves its own section because it is common, it feels like a fix, and it does the opposite of what almost everyone thinks.
An owner decides mobile traffic is where the junk comes from. Desktop inquiries seem more considered. So they go into the device settings and set mobile to minus fifty percent, believing they have just told Google to spend half as much money on mobile.
That is not what that setting does. It does not reduce your mobile spend by half. It reduces your mobile bid by half. You are still buying mobile traffic, you are simply refusing to pay the going rate for it, which means you no longer compete for the good mobile auctions and you win the cheap leftovers instead.
Read that back with the lead-quality problem in mind. You suspected mobile was bringing you weaker inquiries, and you responded by buying the weakest mobile positions available. It gets worse rather than better, and it does so in a way that looks like you took action.
If a device genuinely does not work for your business, minus one hundred percent removes it. Anything in between is a bid, not a budget. Honestly though, once the conversion data feeding your account is clean, this usually stops being the problem you thought it was. Device is a favorite scapegoat for what is really a counting failure.
It's Usually One Campaign
Something useful happens once you can see which leads were real. You stop guessing about sources.
Bowen describes a commercial flooring installer where roughly one in ten inquiries was residential, which they could not serve. With qualified-lead data attached to campaigns, the pattern was immediate: almost all of the bad inquiries were coming from one campaign. It was switched off, and the budget moved. That is not a clever optimization. That is a thing you can only do once you can see it.
This also explains the most expensive failure mode in automated campaigns. Tell Performance Max you want lead forms and it will find you lead forms. If it discovers a placement that reliably produces fifty of them, it will keep buying that placement forever, and the fact that all fifty are junk is not information it has been given. It is doing exactly what you asked. The instruction was wrong.
What This Costs You Past the Ad Budget
The wasted spend is the visible half. The other half never shows up in any ad report.
Somebody in your business works those leads. When most of the pipeline is unqualified, that person spends their week on calls that were never going anywhere. Follow-up on the good inquiries slows down because there is a queue. The genuinely interested prospect who filled in a form on Tuesday gets called on Thursday, by which point they have hired somebody else. Everyone starts treating new leads as probably junk, which is a reasonable conclusion from the evidence and is fatal to conversion rates.
Then the report shows conversions up, the leads feel worse than ever, and the recommendation on the table is to increase the budget. More of exactly this.
Where to Start
In order of effort, cheapest first.
Open your conversion actions and read the list. Anything that is not somebody trying to do business with you goes to secondary. Set your form confirmation to count one conversion per click. Check that the phone numbers on your site are the tracked ones, including the one in the footer that nobody has looked at since the site was built. Set your call conversion to require a real duration, thirty seconds is the usual starting point, rather than counting taps.
Then look at last month's leads, however you have them, even if that is an inbox and a notepad. Mark which ones were real. If more than a handful were not, you now know the number on your dashboard and the number in your business are different, and by roughly how much.
After that, sending qualified leads back to Google is the step that changes what the algorithm chases. It takes a bit of setup and it is the difference between an account that buys clicks and an account that buys customers.
None of this requires you to become a Google Ads specialist. It is closer to checking the oil. You do not need to be a mechanic to own a car, but you should know how to look. If you would rather hand the whole thing to somebody who does this all day, that is what we do. And if you would rather learn it yourself, genuinely, good. The owners who understand what their account is counting are the ones who never get talked into a bigger budget to fix a measurement problem.
