When a Smart Bidding campaign starts spending on junk traffic, most advertisers blame the algorithm and switch back to manual CPC. In the accounts I audit, the bidding strategy is rarely the real problem. The conversion data feeding it is. Smart Bidding does exactly what you tell it to do: if you tell it that a page scroll is a conversion, or that every purchase happened twice, it will chase those signals with your money. This post explains how bad conversion data breaks Target CPA, Target ROAS and Maximize Conversions, and how I diagnose and fix it.
Why does Smart Bidding perform badly with wrong conversion data?
Smart Bidding performs badly with wrong conversion data because it sets bids in each auction based on predicted conversions or conversion value, and those predictions are learned entirely from the conversions you report. Smart Bidding is Google Ads' family of automated bid strategies, including Maximize Conversions, Maximize Conversion Value, Target CPA (tCPA) and Target ROAS (tROAS). It cannot tell a real sale from a broken tag. It only sees what reaches the Conversions column.
If that data is inflated, the system thinks some searches and audiences convert better than they do and overbids on them. If it is incomplete, it underbids on traffic that actually works. If values are wrong, tROAS optimizes for the wrong orders. I compare the main strategies in Smart Bidding explained; this post is about the data underneath them.
What kinds of bad conversion data break Smart Bidding?
Five problems cause most of the damage I see. Each one misleads the bidding model in a different direction:
| Problem | What Smart Bidding sees | What happens to bids |
|---|---|---|
| Duplicate conversions | More conversions than really happened | Overbids on traffic that only looks profitable; CPA looks great but sales do not grow |
| Micro-conversions as primary | Easy actions (page views, clicks) counted like sales | Chases cheap, low-intent traffic that triggers the easy action |
| Wrong or missing values | Equal value for every order, or inflated values | tROAS optimizes toward the wrong products or customers |
| Missing conversions | Fewer conversions than really happened | Underbids, volume drops, campaigns stop spending |
| Lagging or delayed data | Recent days look worse than they are | Short-term pullbacks, and targets changed too early |
Duplicate conversions
Duplicates usually come from the same purchase being counted twice: once by the Google Ads tag and again by an imported GA4 key event, both set as primary. Other common causes are thank-you page reloads, a tag firing on two triggers, or a platform app and a GTM tag both sending the conversion. I list more of these in Google Tag Manager mistakes that break conversion tracking.
Micro-conversions set as primary
Primary micro-conversions are the most expensive mistake for lead generation accounts. If "contact page view" or "click on email" is primary and counts the same as a booked job, Maximize Conversions will happily buy thousands of contact page views. These actions are fine to track, but as secondary conversions for observation.
Wrong values
For tROAS and Maximize Conversion Value, the value is the target. Sending a fixed value of 1 on every purchase, sending the value in the wrong currency, or including tax and shipping inconsistently all distort the model. On Tea of Sylhet, purchase events were missing revenue values in Google Ads at the start, which made ROAS-based bidding impossible until the tracking was rebuilt to send verified revenue.
How do you diagnose a conversion data problem?
You diagnose conversion data problems by comparing what Google Ads reports with what your business actually recorded, then checking each conversion action's settings. This is my routine:
- Compare totals. Pull last month's conversions and conversion value from Google Ads (by conversion time, not click time) and compare them with real orders or qualified leads in your store or CRM. Some difference is normal; a large surplus suggests duplicates, a large shortfall suggests missing data.
- Open Goals > Summary. List every conversion action marked Primary and check whether each represents real business value.
- Check counting. Purchases should usually count Every; leads should usually count One per click.
- Check the source. Make sure the same action is not tracked by both a Google Ads tag and a GA4 import as primary.
- Check values. Review a sample of conversions with the conversion value column and confirm they match real order values and currency.
- Check tag diagnostics. Look for warnings on the conversion action, and test a real conversion with Tag Assistant.
- Check timing. Look at the conversion lag in your account so you know how long conversions take to arrive after a click.
My Google Ads audit checklist covers the wider account review, and the lost conversion calculator helps estimate what missing data might be costing you.
How do you fix the conversion data feeding Smart Bidding?
You fix it by making the primary conversions match real business outcomes, sending accurate values, and filling the gaps that browsers miss. Work through these in order.
1. Set primary and secondary goals correctly
Only actions that represent real money, or a reliable step towards it, should be primary. Everything else becomes secondary so you can still see it in the All conversions column without bidding on it.
- Ecommerce: primary = purchase. Secondary = add to cart, begin checkout.
- Lead generation: primary = qualified form submission, calls over a sensible duration, or offline-imported qualified leads. Secondary = clicks on email, page views, short calls.
Also check campaign-level goals. A campaign can use account-default goals or its own goals, and a campaign set to the wrong goal will optimize for the wrong action even if your account defaults are correct.
2. Send accurate conversion values
Pass the real order value from the data layer, in the right currency, with a consistent rule on tax and shipping. For lead businesses, assign different values to different lead types if you know their typical worth, or use conversion value rules to adjust values by location, device or audience. My break-even ROAS calculator helps you choose a sensible tROAS target once values are trustworthy.
3. Recover missing conversions with enhanced conversions
Enhanced conversions send hashed first-party customer data, such as email and phone number, with your conversion tags so Google can match more conversions to ad clicks. It is one of the simplest ways to reduce under-reporting. I explain the setup in my enhanced conversions guide, and the enhanced conversions hash generator shows how normalised hashing works.
4. Import offline conversions for lead businesses
For service businesses, the website form is not the sale. Importing qualified leads or closed deals from your CRM, using the click ID captured on the form, lets Smart Bidding optimize for customers instead of form fills. I walk through it in offline conversion tracking: send CRM sales back to Google Ads.
5. Use data exclusions for tracking outages
If your tracking broke for a few days, for example after a site update removed the tag, use data exclusions in the bid strategy tools to tell Smart Bidding to ignore conversion data from that period. Google designed data exclusions for conversion data outages, and recommends using them sparingly. Excluded conversions still appear in reporting; they are only ignored for bidding. For expected conversion rate changes, such as a short sale, use seasonality adjustments instead.
What should you do after fixing the data?
After fixing conversion data, give Smart Bidding time to adjust before judging results. Changing primary goals, values or the bid strategy can put a campaign back into a learning period, shown as "Learning" in the bid strategy status, where performance can fluctuate.
Rules I follow after a cleanup:
- Make one significant change at a time, so you know what caused the result.
- Avoid changing targets or budgets heavily during the learning period.
- Judge results over a period longer than your conversion lag, not day by day.
- If old data was heavily inflated, start with a realistic target based on the corrected numbers, not the old CPA.
- Keep comparing Google Ads conversions with real orders or leads every month.
If the old data was badly inflated, expect reported CPA to rise once duplicates are removed. That is not performance getting worse; it is the account finally seeing the truth.
Final thoughts
Smart Bidding is only as good as the conversions you feed it. Clean up primary goals, send real values, recover missing conversions with enhanced and offline conversions, and protect the model from outages with data exclusions. If your campaigns behave strangely and you suspect tracking, my Google Ads management service starts with a conversion data review, or you can book a free tracking audit.
