If you use Smart Bidding in Google Ads, you are trusting Google’s machine learning to decide how much every click is worth. That system only works as well as the data you feed it.
When conversion tracking breaks, changes, or starts firing incorrectly, Google does not know that something is wrong. It assumes the data is real. Smart Bidding then trains on that bad data and starts making decisions that can quietly destroy performance for weeks or months after the problem is fixed.
That is exactly what Data Exclusions exist to prevent.
Yet most advertisers never use them, even when they absolutely should.
This article explains what Data Exclusions are, how they work, and real world scenarios where not using them can permanently harm an account.
What are Data Exclusions in Google Ads?
Data Exclusions tell Google Ads to ignore conversion data from a specific date range for Smart Bidding training.
They do not remove conversions from your reporting. They do not change what shows up in the interface. They only prevent those conversions from being used by Google’s bidding algorithms when learning what a good user looks like.
Think of them as a “do not learn from this period” flag.
They are applied at the conversion action level, which means you can isolate which signals Google should ignore.
This is critical because Smart Bidding systems build a statistical model of who converts, what they do, what device they use, how long they take, and how much those conversions are worth. If you feed it junk, it learns junk.
Why this matters more than most advertisers realize
Smart Bidding does not just use yesterday’s data. It looks back weeks and sometimes months to understand patterns.
If you have one week of wildly incorrect data, it can pollute the training set long after that week is over. You may fix the tracking and wonder why performance never fully recovers. This is often the reason.
Once the algorithm believes a certain type of user or behavior is valuable, it will keep chasing it.
Data Exclusions are how you tell Google, “That week never happened.”
Where to find Data Exclusions
In Google Ads:
- Click Tools
- Go to Budgets & Bidding
- Click Adjustments, then Exclusions
- Fill in the details of what you want to exlude
- Hit save!
Data exclusions are needed more often than you think
Here are some real client situations where I’ve recommended data exclusions!
Client #1: Add to Cart accidentally counted as a purchase
This happens constantly.
This advertiser meant to optimize for purchases, but their marketing assistant set up Add to Cart as a primary conversion. Smart Bidding now thinks a cart is just as valuable as a sale.
For two weeks, the account shows amazing CPA and ROAS. In reality, the system has learned to target people who like to browse and abandon.
They fix the conversion setup, but now Smart Bidding has two weeks of training telling it that non buyers are gold.
Without Data Exclusions, Google keeps optimizing toward cart abandoners.
With Data Exclusions, you exclude that two week window from the Purchase conversion. Now the algorithm learns only from real buyers again.
This is one of the highest impact uses of Data Exclusions.
Client #2: A tracking bug causes duplicate purchases
A site update caused my client’s the thank you page to fire twice. Every sale became two conversions.
ROAS suddenly looks incredible. Smart Bidding aggressively raises bids because it thinks value per click doubled.
She caught the bug three days later and fixed it. But the system now believes users from those campaigns are far more valuable than they really are.
This often leads to inflated CPCs and declining profitability.
Exclude those three days from the Purchase conversion and the model returns to reality.
Client #3: A form starts firing on page load
This is SO common with GTM changes.
A lead form conversion fired as soon as someone lands on the page, even if they never submit anything. My client suddenly had hundreds of “fake” leads.
If my client let Smart Bidding train on that, it would optimize toward people who might bounce instantly.
That can permanently shift traffic quality.
Appling a Data Exclusion to the affected date range allowed the algorithm to ignore that garbage data.
Client #4: A temporary promo distorts behavior
My client ran a one day flash sale with 70 percent off. Conversion rate tripled. Average order value plummeted. Buyers behaved nothing like their normal customers.
That behavior is not representative of my client’s real business, so it should not train Smart Bidding.
Excluding that date range told Google NOT to think extreme bargain hunters our ideal audience.
OK, but what happens if you do nothing?
Most advertisers assume Smart Bidding will just recover once the issue is fixed. That is not how machine learning works.
Bad data becomes part of the training set. It shapes how bids are set going forward. In many cases when bad data is received over a long time, the account never fully recovers because the model keeps chasing ghosts.
This is why people say things like “the campaign has never been the same since that tracking issue.”
They are usually right.
When you should not use Data Exclusions
Do not use them just because performance was bad.
Data Exclusions are for broken or misleading data, not for poor results. If the data was accurate, even if it was ugly, Smart Bidding should learn from it.
Only exclude data when the conversion signals were wrong or unrepresentative of your real business.
Final Takeaway
Tracking breaks. Teams make mistakes. Websites change. Promotions happen. When those things distort conversion data, Data Exclusions are what prevent long term damage.
Most accounts that “mysteriously” fall apart after a tracking issue are not cursed. They are just training on bad data that was never excluded.
Once you start using this feature properly, you will see Smart Bidding behave far more predictably and recover far more cleanly from inevitable mistakes.
