Call conversions: what minimum call length should count for a local service lead?

Call duration threshold is one of those settings people either leave on the default or set to 60 seconds because it sounds reasonable. The trouble is that both extremes can hurt you. Say a plumber gets a pile of 30-second calls that are really "do you cover my zip code?" questions. Those inflate the count, and Smart Bidding starts chasing people who only wanted a price. But a 60-second cutoff can also throw out a real emergency caller who gave the address, got booked and hung up fast.

My view is to start low, around 30 seconds, then check what share of the short calls actually turn into booked jobs. If the short ones never book, tighten it later.

What do others use for clinics, where calls are often quick but still valuable?

I'd push back on starting at 30 seconds and tightening later. Smart Bidding doesn't forget what it has already learned from, so every week at the low threshold trains it on zip code questions, and raising the bar afterward doesn't undo that. It also assumes you can tell which short calls actually booked, and plenty of local advisors can't, because the booking usually happens in a CRM or phone system that never talks to Google Ads. I'd count only the calls you already know are real leads, and track the short ones as a separate observation action that stays out of bidding until you've matched them to actual jobs.

Here's a rule that doesn't need guesswork: set the cutoff at the shortest call length you've actually seen turn into a booking in your CRM, then round up a bit. If you can't pull that data, just keep short calls out of bidding altogether. Microsoft Ads makes this harder, because local advertisers there tend to get far fewer calls, so a few weeks of zip code questions can swamp a small sample. In that case the cleaner signal is worth more than the extra volume.

Going by the shortest booked call is fragile, because a single 25-second reschedule drags the bar way too low. I'd set it at roughly the 10th percentile of booked call lengths instead, so about nine in ten real jobs still clear it.

The sample size is the weak spot. With 30 booked jobs, your 10th percentile is just the third-shortest call, and that can swing a lot from month to month, so pool a few months of data before you set it. Also check whether some bookings happen after the call, through a web form or a text back. Those won't show up as a booked call at all, so match them to the caller's number within a day or so. Otherwise the short calls that led to them will look like dead weight.

Matching on the caller's number is the fragile part. Calls from a front desk line, a spouse booking for someone else, or a different mobile than the one that called all break it. Where you can, carry the gclid through instead. Add a hidden field to the web form that captures the click ID, and when a text thread turns into a booking, log that gclid against the lead in your CRM too. Then you can upload the booking as an offline conversion tied to the original click, and Google accepts those uploads up to 90 days after the click. That links the booking even when the number doesn't match, so your short-call analysis ends up looking at the same leads Smart Bidding sees.

Agreed on the gclid, and the part people miss is when you capture it. The click ID only exists in the landing URL. If someone clicks the ad, browses two service pages, and fills the form on the third, the parameter is gone unless you saved it on the first page load, either in a cookie or a hidden field filled by script. Otherwise the form submits blank and the booking can't be matched, no matter how clean the CRM is. Easy way to test it: click your own ad, navigate a couple of pages, submit the form, and check that the value shows up in the CRM record.

Timing is only half of it. If the form lives on a different domain from the landing page, like a separate booking tool, the cookie won't follow the user there, so the script has to run on both sides. Is your form on the same domain as the ads landing page?

Cross-domain forms are the real trap here. If the form isn't on the landing domain, pass the gclid in the link itself. Append it as a URL parameter when the visitor clicks through to the booking tool, because that survives where cookies don't. The cookie is really just the fallback for same-domain visitors who browse around before they convert.