A brand I reviewed last month had 631 conversions showing in Google Ads. The campaign looked healthy, the count was climbing, and the team was ready to scale. One problem: every conversion came through with no revenue value attached. The system knew something happened 631 times. It had no idea whether each event was a $40 order or a $4,000 order. As Menachem Ani noted on LinkedIn, asking Smart Bidding to optimize for return on ad spend when there's no spend to return against is asking it to chase value it cannot see.
This is the part most teams miss about automation. The algorithm is only as good as the signal you feed it. A conversion count with no revenue is not a small reporting gap. It is the difference between a system that can optimize for profit and one that is flying blind.
The Signal-to-Noise Problem
Search Engine Journal documented a Performance Max campaign that generated 4,000 clicks and produced 37 purchases, yet the platform reported a 62% conversion rate. That math only works when roughly 90% of the conversions are button clicks, form interactions, and abandoned checkouts being treated as equal to revenue.
- Add-to-cart events counted as conversions even when the user never returned
- Checkout-start events weighted the same as completed purchases
- Button clicks logged as micro-wins, overwhelming the real signals
That is a signal-to-noise ratio of roughly 9:1 against the algorithm. The campaigns look healthy inside Google Ads, the conversions look high, and the return on ad spend appears strong. None of it matches what the business actually experiences when the finance team looks at the bank account.
The fix is not to test another bid strategy. The fix is to examine the conversion architecture. Google's documentation on primary and secondary conversion actions makes the distinction clear: primary actions are reported in the Conversions column and used for bidding, while secondary actions are for observation only. When every action is labeled as primary, you train Smart Bidding to optimize toward a vague composite of engagement instead of what matters.
The Minimum Data Threshold
Smart Bidding needs conversion data to work. Chris Chambers, Head of Paid Search at Understory, puts the threshold at 20 to 30 conversions per month for whatever goal you're optimizing toward. Most B2B SaaS accounts don't hit that at the campaign level. So teams either limp along on Maximize Conversions with four leads a month feeding it, or they abandon smart bidding entirely.
Google's own documentation for Target ROAS specifies at least 15 conversions in the past 30 days for Search and Shopping campaigns. Below that, the algorithm is guessing more than learning. The result is volatile bids and inefficient budget allocation.
Portfolio bid strategies solve this without forcing you to merge campaigns. You apply one shared bid strategy across multiple campaigns. They keep their own structure, keywords, negatives, and geo targets. But all their conversion data feeds into one pool for the algorithm to learn from. Instead of your feature-specific campaign seeing 5 leads, your DSA campaign seeing 3, and your core non-brand campaign seeing 8, Google is now looking at 16 combined. You hit that learning threshold faster.
The Attribution Window Mismatch
Cometly's analysis of attribution window problems describes a scenario most B2B marketers will recognize: conversions are climbing, ROAS looks solid, so you confidently double your budget. Two weeks later, actual revenue hasn't budged. Your sales team reports the same lead volume as before. The Google Ads interface still shows those beautiful conversion numbers.
The default click-through attribution window is 30 days. If someone clicks your ad today and converts within the next 30 days, Google Ads counts that conversion. The problem emerges when your actual sales cycle doesn't respect that arbitrary boundary. A customer who clicked your ad 45 days ago and finally converted today? That conversion might get credited to a branded search ad they clicked yesterday, or to direct traffic, or to nothing at all.
Improvado's Google Ads analytics framework recommends extending conversion windows to 90 days for B2B campaigns. The default windows were designed for a different era of digital marketing, one with shorter sales cycles and simpler customer journeys.
The Value Data Problem
Verde Media's audit findings put it bluntly: if your conversion tracking isn't set up correctly, you just handed the algorithm a blindfold and a steering wheel. It's going to drive with absolute confidence. It's just not going to drive anywhere good.
LGG Media's analysis of Target ROAS identifies a subtler failure mode. Even a target set exactly to your breakeven is wrong if Google is not seeing all of your revenue. Say your real-world breakeven ROAS is 200%, but Google only tracks 80% of the revenue you collect. Set the Google target to 200% and the algorithm optimizes to hit 200% on the 80% it can see, which pushes your real-world ROAS to roughly 250% and starves the campaign for volume the whole way up.
The fix is to measure the data-loss rate by comparing the revenue you record against what Google actually processed, then set the Google-reported target below true breakeven by exactly that gap.
The Outage Recovery Trap
Playhouse Digital documented what happens when conversion tracking breaks temporarily. Maybe the tag was paused or deleted from the checkout page by accident. Maybe there was a processing issue with Shopify. The website still works, customers can still buy, but the Google Ads Conversion Tag stopped recording.
The longer this goes on, the bigger the impact on campaign performance after the fix. The bid strategy thinks its targeting signals need adjusting. It incorrectly adjusts Max CPC bids. The result is increased costs and reduced revenues during the recovery period.
The solution is the Data Exclusions feature in Google Ads. You tell the bid strategies which days to ignore so they don't act on bad data. Find it under Tools & Settings, Shared Library, Bid Strategies, Advanced Controls, Data Exclusion tab.
The Pilot Checklist
Before you blame the bidding strategy, run this diagnostic:
- Check that every conversion is carrying its actual value, not a placeholder or zero
- Confirm primary conversion actions are limited to revenue events, not engagement signals
- Verify you have at least 15 to 30 conversions per month at the campaign or portfolio level
- Extend attribution windows to match your actual sales cycle
- Measure the gap between Google-reported revenue and actual revenue collected
- Use Data Exclusions to quarantine any tracking outage periods
Most the algorithm isn't working problems are really the data isn't there problems. The scary part, as one commenter noted, is that nothing looks broken. Conversions are showing up, dashboards are moving, and everyone assumes the data is good. The CFO sees the discrepancy first, when the forecast doesn't match the bank balance.
