Your Target ROAS campaign keeps stalling out. It's starving, not failing.

Here is a pattern we see a lot on small, focused accounts. The Google Ads campaign spends for a while, then quietly stops spending. The owner lowers the Target ROAS to coax it back, delivery recovers for a bit, then it flatlines again. On a chart it looks like a heart monitor: spikes, then a flat line, then another spike.

The instinct is to blame the budget, the keywords, or the algorithm. Usually it's none of those. The campaign isn't failing. It's starving.

What Target ROAS is actually doing

Target ROAS is value-based Smart Bidding (Google now labels it simply "Target ROAS", formerly "Maximize conversion value with a Target ROAS"). You hand Google a return on ad spend target, and it only bids into auctions where it predicts it can hit that return. To make that prediction it needs a steady stream of conversions with values attached to learn from.

Google's guidance is to judge and run value-based bidding over periods with meaningful conversion volume, commonly cited as roughly 50 conversions in 30 days. A niche, bottom-of-funnel campaign, the kind selling into a small, specific audience, often produces a fraction of that. There simply isn't enough signal for the model to work with.

The death spiral

Now add a normal quiet stretch. Even a healthy low-volume campaign will have a week or two with no conversions, just noise. To Target ROAS, that stretch reads as "this campaign is no longer hitting its return." So Google does exactly what you told it to: it protects the target by lowering bids and pulling back delivery.

Less delivery means fewer impressions and clicks. Fewer clicks means fewer conversions. Fewer conversions makes the campaign look worse, so Google pulls back further. That is the spiral, and each loop down is harder to climb out of than the last.

This got sharper recently. As of Google's August 17, 2026 update, campaigns with limited budgets are optimized more consistently toward their specified target, which Google notes can have a noticeable impact on cost and conversion volume. In other words, a strict target on a thin, budget-constrained campaign now bites harder than it used to.

Why lowering the target makes it worse

The common fix, nudging the ROAS target down whenever it stalls, is the one that keeps you stuck. Changing the target is a significant edit, which restarts the learning period. So every time you touch it, the campaign re-enters learning with the same thin data it had before. You are treating the symptom (the output number) while the cause (too little conversion volume) never changes.

The fix, on the same budget

You don't need more money. You need to give the algorithm something to learn from.

• Optimize to a more frequent signal. Don't point the campaign only at the rare bottom-funnel event. Count an earlier, higher-frequency action too, a sign-up or a product activation, so there's a steady flow of conversions to learn from. Assign values (a purchase worth more than an activation, an activation more than a sign-up) so the important actions still carry the most weight. This is the single biggest lever.
• Match the strategy to your volume. If you genuinely can't feed a value target, don't run one. Maximize conversions asks for no target at all, which is exactly why it's the safer home for a low-volume campaign. Chase conversions first; add a target later, once volume lets you set that number honestly.
• Then stop touching the target. Set it and leave it long enough for learning to complete. Every change restarts the clock.

Do that and the shape changes. Instead of a flatline punctuated by panic edits, the campaign gets a consistent enough stream of signal to hold delivery steady on the budget it already had.