Smart Bidding Needs Accurate Conversion Data to Work

You've probably heard that Smart Bidding uses "machine learning" to set bids. That's true, but it obscures the most important detail: what the machine is learning from.

Strategies like Target ROAS, Target CPA, and Maximize Conversions analyze hundreds of signals for every auction — device, location, time of day, audience segment, search query, even the user's browser. The algorithm uses these signals to predict whether a given click will convert and how much that conversion is worth. Then it sets a bid accordingly.

Here's where it gets interesting: Google's model is trained on your conversion data. It examines your historical conversions — which clicks led to purchases, what those purchases were worth, and the patterns connecting the two. If your conversion data is incomplete or inaccurate, the model learns the wrong patterns. It's not that the algorithm is bad; it's that you're teaching it the wrong lessons.

When tracking underreports conversions, something insidious happens. Smart Bidding looks at your campaigns and sees — let's say — 60 conversions instead of 100. It concludes your campaigns are performing 40% worse than they actually are. So it lowers bids to reduce what it perceives as wasted spend. Lower bids mean fewer impressions, fewer clicks, and fewer actual sales. Revenue drops, but Smart Bidding doesn't know it caused the decline. It just sees even fewer conversions and lowers bids further.

We've watched this spiral play out with merchants who couldn't understand why their Google Ads stopped working. The campaigns hadn't changed. The products hadn't changed. But somewhere along the way, a tracking update broke, and within 3-4 weeks, Smart Bidding had essentially throttled the account down to minimal spend.

The reverse is equally dangerous. Duplicate tracking tags that count each conversion twice make Smart Bidding think campaigns are performing better than they are. It overbids aggressively, burns through budget, and your actual cost-per-acquisition doubles while the dashboard still shows healthy numbers.

Most merchants know about missing conversions, but data quality problems go beyond just undercounting. Here are the issues we see most often:

If Smart Bidding is underperforming, resist the urge to switch bid strategies, adjust targets, or overhaul your campaign structure. Fix the data first. The best campaign structure in the world can't overcome a 30% hole in your conversion tracking.

• Audit conversion accuracy — Pull the last 30 days of Shopify orders and compare against Google Ads conversions (offset by 72 hours). If the gap exceeds 15%, your tracking needs work.
• Enable Enhanced Conversions — This single change recovers 15-20% of lost conversions by using hashed customer email data. It's the highest-impact fix for Smart Bidding accuracy.
• Remove duplicate tags — Audit every active conversion tracking method. GTM, Shopify native, third-party apps — keep exactly one source and disable the rest.
• Verify conversion values — Ensure the dollar amounts reaching Google Ads match your actual order values consistently. Check whether tax and shipping are included or excluded, and make it consistent.
• Give Smart Bidding time to recalibrate — After fixing tracking, the algorithm needs 2-3 weeks to learn from the corrected data. Don't judge performance during this period.

The first thing you'll notice is a jump in reported conversions — not because you're getting more sales, but because you're finally counting the ones that were always happening. This is a good sign. It means your tracking fixes are working.

Give Smart Bidding 2-3 weeks to absorb the new data. During this learning phase, bids may fluctuate as the algorithm recalibrates. After that stabilization period, here's what merchants typically see: