Ad schedules with Smart Bidding: still worth setting, or let the algorithm price hours?

Pausing ads at 2am and pushing bids up at lunch is an old habit from manual CPC days, and with tCPA or Maximize Conversions I'm not convinced it still helps. The bidding model already factors in hour and device when it prices each auction, so a hard ad schedule mostly just removes volume you could have bought at a fair price.

It still makes sense when the business can't actually serve the lead. A call-only offer with the phone answered 9 to 6, or a store that closes at 8, is the obvious case. Those clicks are real waste, not just cheap-looking ones.

For everything else, I'd leave the schedule open and judge it by conversion rate and cost per conversion by hour after a few weeks, not one afternoon. Has anyone seen a schedule clearly beat the algorithm on a pure lead-gen account? My guess is it's rare.

The bidder prices every auction off the conversions it can see, and it can't tell a booked patient from a 2am form fill that nobody will answer until 9. Take a dental clinic with a pile of night-time submissions. Cost per lead looks fine, cost per booked appointment is awful, and the bidder keeps paying for the night traffic because it's counting the wrong thing. Trimming hours is a blunt way around that. Sending qualified conversions back through offline import, or making a booked visit the primary action, goes after the real problem.

Schedules aren't as blunt as people make them out to be. Offline import helps, but booked-visit data usually shows up days late, and until it lands the bidder keeps learning from cheap 2am leads. A night-time bid cut is crude, sure, but it works as a stopgap while the offline signal catches up.

It comes down to how many of those 2am leads ever turn into a booked visit. If it's close to zero, the schedule is just cutting waste, which is something the bidder can't do on its own. If a real share book a day later, a hard cut throws out good patients along with the bad ones, and a bid adjustment is the gentler lever. You can pull that rate from the CRM by lead timestamp in an afternoon, and it'll tell you more than guessing how long the offline signal takes to land.

I'd push back on the afternoon pull. If you cut the CRM by lead timestamp today, the 2am leads from the last week or two haven't had time to book yet, so the night booking rate looks worse than it really is. That lag you wanted to skip is exactly what decides whether the number means anything. Use a cohort older than your usual time to book, say 30 to 60 days back, and compare hours only on that. Otherwise you could end up cutting night traffic that just hasn't finished converting.

Honestly I'm not sure the 30 to 60 day window holds up for local services. A lot of people book within a few days of filling out the form, so a two to three week cohort already catches most of the bookings. Going back two months can backfire too, because you'd be judging night hours on a different season and different prices, and that can mislead you about as much as the lag does.