Lead Quality + Primary Goal Changes?

[Staff_Sharp]

I’d switch back to the primary goal that best matches actual revenue, not just keep waiting because CPL looks stable.

What jumps out here is that the system got a stronger signal when you imported booked offline conversions, then performance deteriorated when phone calls were allowed to dominate the primary optimization set. If forms are closing better than calls, that’s a hint you taught Google to value the noisier action more heavily.

With only 21 booked offline conversions total, I’d keep the setup simple:
1. make the highest-quality downstream action primary
2. move weaker proxy actions like raw phone calls back to secondary unless you can prove they close at a similar rate
3. avoid teaching the campaign two different definitions of success at the same time

If you want to test calls as primary, I’d do it in an experiment rather than on the main campaign. At this volume, the bigger risk is muddy conversion prioritization, not that the algorithm just needs more patience.

[gptbuilder_marc]

A ROAS drop from 7x to 2x after switching offline conversions to primary is a specific signal. The question worth asking before anything else is whether the algorithm is now optimizing for a lower-quality conversion event or whether the volume of 21 imported conversions is too thin for the model to trust.

[1CommerceOfficial]

Yep — I’d revert **calls** back to secondary (or at least make them *much* stricter) and let the campaign optimize to the closest thing you have to revenue.

What likely happened is exactly what you observed: when calls became primary, Google found “easier” pockets of traffic that generate phone-call conversions (often broad/low-intent queries), even if those calls don’t book.

A practical way to structure it:
1) Keep **Booked Job (offline import)** as the only *primary / Include in “Conversions”* action for the main campaign.
2) Put calls back to secondary OR set a higher-quality call definition (e.g., 60–120s duration, qualified disposition in CallRail) before ever making them primary.
3) If you want to test calls as primary, do it as an **experiment** or separate campaign with its own conversion action set (so you don’t muddy the learning on the main one).

Also: with only 21 booked conversions total, Smart Bidding will be jumpy. You may get more stability using a proxy (form submit) *temporarily* while you build more booked-job volume—then switch back once you have a steadier stream.

If you share bidding strategy + last 30d booked volume, people can be more precise.

[stovetopmuse]

Sounds like the algo just re-optimized around the new primary, not necessarily the better signal.

With only ~20 offline conversions, I wouldn’t trust it as a primary yet. That’s pretty thin for it to learn anything useful. I’d switch back to forms as primary since they’re actually closing, and keep offline as secondary until you have more volume.

Also seen this exact thing where calls look good on paper but are low intent or junk. Your ROAS drop kinda confirms that.

[ppcwithyrv]

I’d switch the CallRail phone conversion out of primary and move primary back to the lead type that actually closes better, because Google is probably optimizing toward more calls, not more revenue. With only 21 booked offline conversions, I’d keep those as secondary for now and test them as primary later once you have enough volume to give bidding a cleaner signal.