What to check before you touch anything
- Is this clustering actually a problem, or just a reflection of real demand? If one segment (geo/device/audience) genuinely accounts for the bulk of paying demand in your niche, concentrating there may be correct algorithm behavior, not a distortion.
- How aggressive is the current target CPA? At a tight (low) target, the system is forced to concentrate on the cheapest, most predictable segment just to hit its CPA goal — that's expected smart-bidding behavior, not an anomaly.
- Have your audience signals or targeting themselves narrowed recently (for example, in Performance Max, or through tight audience signals)? Then the clustering reflects your settings, not an algorithmic decision.
- Is the campaign's structure broad enough to begin with (does it cover a range of keywords/audiences)? If the structure was narrow from the start, "clustering" may simply be a case of there not being much else to choose from.
- Are conversions clustering in a segment with a lot of historical signal (like remarketing/past-visitor audiences) simply because there's more data there to optimize on, rather than because other segments are genuinely worse?
- Do the segments that aren't converting have messaging and landing pages built for them specifically? If your ads and landing page are built around the "average" or dominant segment, and other audiences see the same message with no adaptation, low conversion there may reflect an offer mismatch, not a lack of demand. Worth checking this before concluding a segment is simply "bad."
- How well does your messaging for the underperforming segments match the competitive landscape specifically in those segments? Competitors who are strong in your dominant segment aren't necessarily strong everywhere — there may be an open opportunity in the "weaker" segments that's currently going unused because the message there isn't tailored.
Possible approaches
- If the clustering comes from a tight target CPA (the system is conserving budget by showing up only in the cheapest segment), you could try loosening the target slightly, or adding budget specifically to expand beyond that narrow segment, accepting a temporary CPA increase as the cost of diversifying.
- If one segment genuinely dominates on real demand quality, you could instead lean into that concentration deliberately — via bid adjustments that reinforce it — and treat the other segments as a separate initiative, rather than trying to artificially spread the budget toward weaker-performing directions.
- Google generally doesn't treat clustering as inherently a problem — the real question is whether it matches your business goals (reach vs. efficiency). If reach matters to the business too, not just efficiency, that should be built explicitly into the strategy (for example, a separate reach-focused campaign with a looser target).
- If the clustering is tied to narrow audience signals or a narrow campaign structure to begin with, broadening your targeting/keywords can give the system more room to diversify without touching the CPA target at all.
- In some cases, if it's important to the business to test new segments but the algorithm keeps reverting to the familiar narrow one, it's common to run a separate test campaign on the new segment with its own (looser) budget and target, rather than mixing the test into the main, already-working campaign.
- If it turns out underperforming segments are getting the same messaging and the same landing page as the dominant one, it's worth analyzing the competitive landscape in those segments specifically (their ads, offers, landing pages) and adapting your own message to fit. If that adaptation calls for a meaningfully different offer or page structure, it's often more effective to spin that segment out into its own campaign with its own messaging — and possibly its own (looser, to start) target — rather than waiting for the overall tCPA target to somehow "find" conversions there with an unchanged message that doesn't fit that segment.