Google uses conversions as signals for automated bidding. Each conversion tells it what to seek more of.
If you count a phone-number tap as a lead, Google may seek more taps. The same goes for repeat calls or YouTube subscriptions. The numbers can rise while sales stay flat.
Check which actions count as leads. Then see which leads become customers. Do that before paying for a more automated campaign.
Your conversion count may be answering the wrong question
Google Ads can see the actions you track. It cannot see who became a customer unless you send that result back. When a conversion is included in a campaign's bidding goal, each count becomes a signal. Google bids for more people likely to take that action. The wrong goal can steer spending toward the wrong result.
A phone-number tap can count even if nobody speaks to the caller. A form can count when the page loads, before anyone submits it. The same caller can count several times. Google may record each event as configured. The total can still overstate the number of viable enquiries.
Our September 22, 2026 review covered 19 managed accounts. In three, the counted action did not reliably represent a lead. Thirteen had at least one contact action set to count repeatedly after an ad click. In one account, two tracking routes could record the same form enquiry. We managed those accounts. Fixing the settings is our responsibility.
One YouTube campaign shows how the signal can drift. Subscriptions and follow-on views made up most of its reported conversions. Those are valid measures of video engagement. They do not show that someone asked for a quote. If they are included in the bidding goal, Google seeks more of that engagement. Added to enquiries, they also make the account's total look stronger than its lead count.
Start with the action that confirms contact. Count a submitted form or a meaningful call once. Screen out spam. Remove any tracking route that records the same enquiry twice. Then compare Google's count with the enquiries staff received. Explain the gap before calling the dashboard's cost per conversion a cost per lead.
Real enquiries are only the middle of the story
Counting genuine enquiries still leaves Google guessing about quality. One caller asks for a service the business does not offer. Another books. A competitor or job seeker may fill out a buyer's form. If each contact counts as a conversion, Google receives the same success signal. Staff learn which ones were worth pursuing.
Accounts move through three stages. First, they count actions that are not real enquiries. Next, they count real enquiries but do not send back the results. In the final stage, staff record what happened. Google matches those outcomes to ad clicks. Bidding can then favour clicks more likely to produce customers. All 19 accounts in our September review were still in the first two stages. Some graded leads. None had enough matched, graded outcomes to guide value bidding.
A useful record links each enquiry to its source, the person's request, and the outcome. A spreadsheet works if staff keep it current. Mark spam and duplicates. Record relevant enquiries, appointments or quotes, and sales. Do this while each result can still be connected to the ad click.
Only four of the 19 accounts had a way to grade leads. In the other 15, qualified lead volume was unmeasured. We had not built a reliable way to collect those grades and send them to Google. That is our responsibility. Clients cannot use a feedback process we have not provided.
An auto service shop recorded 31 qualified leads. Google reported 16.94 matched conversions. Those results were marked secondary, so they did not guide bidding. Another account uploaded 604 outcomes, but none matched. Work had been done in both accounts. Neither bidding strategy used a useful outcome signal. Reports should separate what staff recorded, what Google matched, and what bidding used.
Three clinics recorded 445 tracked calls in one month. None could be tied to an ad click. The count showed phone activity. It could not show which ads drove calls or which callers booked. Connect each ad click to its call record. Mark whether staff answered and whether the caller was relevant. Check whether staff can answer when the ads run.
Five questions to put to your agency
The useful account review fits into five questions. Each answer should come from settings and lead records, rather than a slide showing a rising conversion total.
- Are we counting real enquiries, once each? Ask what makes a form or call count, whether repeat contacts can be counted again, and whether page views or video engagement appear in the lead total.
- Where does every lead end up? Ask to see the record that joins the contact to its source and shows what happened after staff received it. A missing or split record makes reconciliation guesswork.
- Does someone mark which leads were good? Ask who labels spam, out-of-area requests, qualified enquiries, bookings and sales. If nobody does, the business knows more than the bidding system but has no way to use it.
- Is Google told which ones were good? Ask how often results are sent back, how many Google matched, and whether bidding is actually using them. Sending an update and teaching the campaign are separate steps.
- Can we tell which phone calls came from the ads? Ask for call records tied to ad clicks and for the disposition of those calls. A count of ringing phones cannot establish what the ads sold.
An agency should show these answers for the same period. Record why Google's total differs from staff's, including spam, repeat calls and failed matches. Mark tracking changes before comparing periods because they also move the trend line.
Set the lead price from your own books
Gross profit, the spending limit, and the close rate set an affordable lead price. Google's conversion dashboard cannot set it. The dashboard does not know your margin or how many enquiries become customers.
Suppose a new customer brings $2,000 in gross profit during the payback period. The owner will spend up to 30 percent of that profit to win the customer. That sets a $600 limit per customer. If one lead in five buys, the lead-to-sale rate is 20 percent. Multiply $600 by 20 percent. The maximum cost per lead is $120.
Dividing $600 by 20 percent gives $3,000 per lead. That reverses the calculation and makes an unaffordable lead look acceptable. Google cannot catch the error from its conversion count. It does not know the owner's profit target. Use the calculator in the technical playbook with your own profit, spending limit, and close rate. Calls, forms, and services may need different ceilings if they close at different rates.
The close rate must come from the business record. An agency without sales outcomes cannot fill it in responsibly. If that record cannot show which ad enquiries became customers, the missing link is the finding. Do not use an assumed close rate to declare the current cost per lead affordable.
Keep the campaign simple until the feedback works
Search reports some of the searches that triggered ads. An agency can exclude terms outside the business's services or area. Search also gives later tests a point of comparison.
Performance Max can show ads across Google's channels. AI Max expands matching and creative options within Search. Demand Gen places ads across visual surfaces. With automated bidding, each can seek more of the conversion action set as its goal. If that action is an ungraded form or a video subscription, wider reach can raise the count. That count does not show whether more customers arrived. Start with Search. Make these options a regular part of the mix when matched, graded outcomes reach bidding. Any earlier test needs a defined business outcome and a comparison with Search.
Eleven of the 19 accounts spent under $1,500 in the examined month. Split that spend among several campaign types and each may yield too little evidence to judge. A lead record, call attribution, or a working upload connection may be the better first investment. Each makes the current spend easier to evaluate.
These findings cover our managed accounts at one point in time. They do not show how often other agencies face the same problem. Google's matching figures cover only outcomes the platform recognized. For our accounts, the order is clear: verify the contact, record its outcome, send it back, then decide whether broader automation has earned its budget.
The full signal-first Google Ads playbook contains the technical checks, account evidence and calculator behind this article. Google Ads management at Choice OMG starts with the lead record; contact us to examine yours.