Legacy pmax campaign tanking

[ppcwithyrv]

How any conversions have you made? Looks like it went cold turkey and needs to warm up (relearn). A month is too long to expect it to keep its place at the auction.

[caramello-koala]

If you just turned it back on recently then it will enter a relearning phase, give it time before making changes or looking too deeply into the data.

Since you’re measuring CPA I assume a conversion is a lead for you? If so, do you know what percent of your leads that came via display actually converted into sales? I ask because display in pmax tends to get leads at a lower CPA, but they’re usually less qualified than search.

[Signalbridgedata]

Even though campaigns retain historical data, the auction environment doesn’t. If it was paused for a month, competitors, budgets, and bidding pressure likely changed. When it came back, the model may have recalibrated toward Search inventory because that’s where it could find faster signals.

Also, PMax has evolved over time. Older campaigns that leaned heavily on Display sometimes shift more toward Search/YouTube when reactivated. It doesn’t necessarily mean it “reset,” but it probably had to re-learn in a different competitive landscape.

If Display historically worked, I’d test asset groups that strongly bias visual placements (strong images, audience signals aligned with historical Display converters). But it may not behave exactly like 2023 again.

[TTFV]

Most of the conversion data has washed out at this point so it’ll need to gather fresh conversions to relearn what works most efficiently.

I would sit on in for a full three weeks and see if it’s recovering close to where it was before. If it’s still miles away you might try adding search themes, optimizing creatives, and adjusting settings some… many new features and changes have happened since 2023.

Even when a campaign works really well for a long time without touching it, it can usually work even better through regular optimization.

[crawlpatterns]

I’ve seen long running PMax campaigns act weird after a pause. Even if Google says it doesn’t “reset,” it definitely seems to re-enter a learning phase and rebalance inventory. If most historical volume was Display, the ecosystem shift alone could explain the higher CPA now.

[QuantumWolf99]

Yeah the month pause almost certainly nuked it… pausing and reactivating is a confirmed learning phase reset trigger. But there’s a bigger issue here. A 2023 PMax that was crushing on Display with minimal Search was essentially running on an older behavior model that Google has been phasing out.

So you got hit twice simultaneously –> full learning reset PLUS the campaign re-initialized against current 2026 PMax behavior, not the 2023 version it had spent years optimizing on. That Display-heavy performance is basically gone forever.

For my larger ecom clients I’d never pause a high-performing PMax for more than a few days… if a pause is absolutely necessary I drop budget to the absolute minimum instead to keep the signal alive. The historical learning is far too valuable to risk.

Your best path now is accepting the new reality, let it relearn for a full 6 weeks without touching it, and see where it settles. The Display separate campaign won’t replicate it because that muscle memory lived inside PMax’s cross-channel optimization, not in the audience lists themselves.

[AccomplishedTart9015]

the pause doesn’t preserve old pmax behavior. when it comes back, it’s basically operating under whatever today’s pmax system is, and it has to re find its pockets again. a month off is long enough that auctions, inventory mix, and feed signals can shift too, so that cheap cpa pocket might just be gone.

the move into search is also common when it’s trying to regain conversions fast. it leans on higher intent inventory even if it costs more.

i usually don’t try to recreate old display performance with a separate display campaign, it rarely works. i either rebuild a fresh pmax with clean conversion goals and updated assets, or i split out proper search so pmax stops eating everything and the account has more control.

for a quick diagnosis, check if conversion actions changed, attribution windows changed, or brand started getting counted differently. those 3 explain a lot of the cpa jump after a pause.