The auction you're not in.
Every search query is its own auction. Smart Bidding doesn't just decide how much you bid. It decides whether you enter at all.
The auctions you sit out still happen. Your competitor wins them. Sometimes for a price you'd have paid gladly. You will never see this loss in any Google Ads report.
Watch it happen.
Below: a live simulation of Smart Bidding deciding whether to enter each incoming auction. Drag your target ROAS up. Watch the orange "Sat Out" column grow. The orange-tinted rows are auctions you'd have profited on — the ones Smart Bidding declined on your behalf.
The silent trade
Setting tROAS to 6x doesn't mean "every auction I enter must return 6x." It means "don't enter any auction predicted to return less than 6x." Those are very different statements. The first is a quality bar. The second is a volume cap.
When you raise tROAS, Smart Bidding's reported ROAS climbs. Cost falls. Everyone in the meeting nods. But total profit — the only number that pays salaries — quietly contracts, because the algorithm just stopped bidding on a thick layer of auctions that were profitable, just not spectacularly so.
Why it hides from you
Google reports auctions you participated in. Impressions, clicks, search impression share lost (budget), search impression share lost (rank). All measured against auctions you were entered in.
There is no column for "auctions Smart Bidding declined on your behalf." No alert when the algorithm starts ghost-mode on a query cluster that was driving 30% of your revenue last quarter. The only signal is volume softly dropping while your ROAS dashboard glows green.
This is by design. Smart Bidding optimises for the metric you told it to. It does not surface the trade-off. See also: conversion data staleness, which makes this even worse during peak periods.
The math nobody runs
Take an account with 60% gross margin. Breakeven ROAS is 1.67x. Every auction predicted to return between 1.7x and your tROAS is an auction you'd profit on but won't enter.
At a tROAS of 6x, you're declining roughly 70% of the auctions you'd have won money on. The "saved" budget doesn't recycle into better auctions. It just sits unspent. Your competitor — running a more honest tROAS — picks up the demand.
This is the structural reason most accounts hit a profit ceiling well below their potential. The bidding strategy is doing exactly what it was told. The instruction was wrong.
What to do instead
- • Set tROAS by margin, not by habit. Breakeven first, then a sensible cushion. Not "what we did last year."
- • Test downward. Drop tROAS 20%. Measure total profit (not ROAS) for 14-21 days. The answer is almost always more profit at a lower ROAS.
- • Watch volume, not just efficiency. A sudden 15% conversion drop with stable ROAS is the signature of Smart Bidding ghost-moding a query cluster.
- • Segment campaigns by margin. Different products, different tROAS. One blanket target across a mixed catalogue guarantees you over-bid on some and ghost-mode on others.
- • Audit Auction Insights weekly. A new competitor appearing on your branded terms often means you've quietly stopped bidding on tail queries that were feeding them.
Next steps
Related Reading
More on the gap between what Smart Bidding optimises and what your P&L cares about.