TL;DR
Google Ads automation only performs as well as the conversion signals it receives. If every form fill counts as a success, Smart Bidding will find more form fills, even if many never become customers.
Before changing bids, budgets, or campaign types, check what Google is optimising for. Use CRM data to distinguish qualified leads and sales from weak enquiries, import those outcomes back into Google Ads, and make sure conversion values reflect real business value.
In short: fix the signal before chasing the metric. Better data gives Google a clearer instruction and a better chance of finding customers who create revenue and profit.
The wrong debate about Google Ads automation
When a Google Ads campaign starts to miss its targets, one of the first suspects is often Smart Bidding. Teams change the target CPA, move budget, or switch campaign types. They look for a better setting because the numbers on the screen have changed.
That can be useful, but it skips an earlier question, what result has the account been told to value? Google cannot see whether a form enquiry became a sale unless that outcome is sent back from the CRM. If every form fill counts as a conversion, Smart Bidding will look for more people likely to complete that form. It has no reason to prefer the people your sales team can actually close, because it has not been given the sales outcome that tells it which enquiries were worth finding.
This is why sometimes a lower cost per lead can still leave the business worse off. The campaign may be doing its job very well. It’s following the signal it was given, even when that signal treats a poor-fit enquiry and a high-value customer as the same event.
Start with the signal. Before changing the bid strategy, check which conversions reach Google and whether the CRM shows which leads became customers.
The Google Ads system does what it is told
Google’s automation has no commercial instinct of its own. It cannot look at a lead and decide whether the customer is likely to buy, whether the sale will be profitable, or whether the inquiry will waste your team’s time. It learns from the conversion events and values you send it.
That means the specificity and clarity of your instructions matters. Tell Google that a completed form is a success, and it will find more people who complete forms. If those forms include poor-fit prospects, job seekers, or people who never answer the phone, the system still counts them as wins because that is what it has been trained to find.
Cost per lead, conversion rate, and return on ad spend are useful measures. They can show whether something has changed. On their own, though, they cannot tell you whether the campaign is helping the business.
A lower cost per lead may look like progress while sales spends more time chasing poor enquiries. A high return on ad spend can hide low-margin orders. Even a rising conversion rate can be misleading if the account is pushing people towards an action that has little commercial value.
That is why teams can spend months adjusting targets, budgets, and campaign settings without fixing the real problem. They are responding to the output on the dashboard, while Google is still learning from the same weak instruction underneath it.
The important question is not simply whether the metric improved. It is whether the conversion behind that metric represents something the business wants more of. If it does not, better-looking numbers can take the account further away from the result that matters.
Where Google Ads signals usually break down
The trouble rarely starts with one big tracking failure. It builds through gaps that are hard to spot.
A Google Tag may fail to load for some visits. Consent rules and browser blocks can stop a click or form from being tied back to the ad. The account still records some wins, so the loss can hide in plain sight. That matters. Google has less proof of who took action, and Smart Bidding has less to learn from. One small gap rarely ruins a campaign. Several gaps can leave it chasing a skewed view of the people who buy.
Lead work has a second gap. A thank-you page shows that a person sent a form, but it can’t show whether they took the call, met your needs, or bought. If each lead gets the same mark, the system will hunt for the easiest leads, including ones your sales team won’t close.
Online sales can go wrong in the same way. Sending revenue on its own treats two £500 orders as equal, even when one has a low margin, a high return risk, or no chance of a repeat sale. The need is simple: know what the account sees, where it goes blind, and which outcome comes closest to real value before you ask automation to make more choices. Feed in qualified leads or closed sales where you can. Then Google has a better reason to seek more people like them.
Fix the signal before you fix the bid
Start by opening the conversion list in Google Ads. Find out which actions are included in bidding. If a newsletter sign-up, contact form, and completed sale all count as the same kind of success, Google cannot tell which one matters most.
Choose one primary conversion that reflects a useful business outcome. For a lead-generation business, that might be a lead accepted by the sales team, rather than every form submission. Keep early actions, such as form fills and phone calls, in the account for reporting, but do not let them steer bidding if they are a poor guide to revenue.
Then make sure the CRM records what happens next. Give each lead a clear stage, such as new lead, qualified lead, booked meeting, opportunity, and sale. Each lead needs a consistent record, so you can see which enquiries became valuable and which did not. If the team cannot agree on what makes a qualified lead, solve that before asking Google to optimise for one.
The next job is to send those outcomes back to Google Ads. Capture the ad click ID and consented contact details when a lead arrives, then import qualified leads, sales, or agreed lead values from the CRM. This links the later sales outcome to the original ad click. Where sales take too long or happen too rarely, begin with the earliest stage that reliably predicts revenue, such as a qualified lead or booked meeting.
Check the data before you change the bid strategy. Compare the leads sent to Google with the CRM records: do the dates match, are duplicates removed, and do the values reflect what the business gained? Once that signal is reliable, give Smart Bidding time to learn from it. Judge the change by lead quality, sales, and profit where possible, not by whether cost per lead falls in the first week.
The fix is not a smarter setting. It is a clearer instruction: tell Google which customers create value, then give it enough reliable evidence to find more of them.
Questions to ask your marketing agency before changing the bids
Which conversions are currently steering Smart Bidding, and why?
Look for a clear list of primary conversion actions and a business reason for each one. “Form submissions” alone is not enough if many never become customers.
Does a form fill count as success before the sales team has qualified it?
Look for a distinction between an enquiry and a qualified lead. The agency should explain which stage Google is optimising towards.
Can you show which leads became meetings, opportunities, or customers?
Look for CRM evidence, not a guess based on cost per lead. They should be able to show the journey beyond the initial form.
Are qualified leads or closed sales being sent back to Google Ads from the CRM?
Look for an existing import process or a specific plan to build one, including who owns each part of the work.
Which conversion best predicts revenue when a sale takes weeks or months to close?
Look for a meaningful earlier stage, such as a qualified lead or booked meeting, backed by evidence that it usually leads to sales.
Are duplicate events, test submissions, or low-value actions included in bidding?
Look for regular conversion checks and a clear process for removing noise before it affects the account.
Do the values in the account reflect profit or lead quality, rather than revenue alone?
Look for an explanation of how values are set. A £500 order should not automatically carry the same value as another £500 order if their margins differ.
Where could tracking be missing conversions?
Look for an audit of tags, consent settings, browser restrictions, and CRM matching. A good answer identifies what the agency can measure well and where the gaps remain.
How will you check that the CRM data and Google Ads data match before changing the bid strategy?
Look for a reconciliation process: matching dates, removing duplicates, checking lead counts, and confirming that uploaded outcomes belong to the right campaigns.
Once the signal changes, how long will you let the account learn before judging the result?
Look for a defined review period and success measures that go beyond the first week’s cost per lead.
A good agency should answer these plainly and show evidence where it can. If it only talks about target CPA, campaign types, and budget changes, it may be trying to improve the output without checking the instructions Google is following. If you suspect your agency is not delivering up to their promises and using templated cookie-cutter tactics, read our guide on How to Spot a Cookie-Cutter PPC Agency vs. a Real Strategy Partner
Conclusion
Google’s automation cannot decide what a valuable customer looks like for your business. It learns from the conversions and values you send it. If the account treats every form fill as a win, Smart Bidding will find more form fills, even when most of them never turn into revenue.
Before changing targets, budgets, or campaign types, check the instruction underneath them. Make sure the conversion steering the account reflects a qualified lead, a sale, profit, or the earliest stage that reliably predicts one. Then check that your CRM and Google Ads data agree.
The system does what it is told. Give it weak signals and it will scale weak outcomes; give it reliable evidence of real value and it has a far better chance of finding more of the customers you want.
If you suspect your account is also experiencing such a drift. Download our PPC Drift Checklist with 10 simple yes/no questions and find out.