Call conversions for local services: what call length counts as a lead?

Call tracking starts more arguments with local service clients than almost anything else. Say a plumber gets 40 calls a week from Google Ads and half of them last under 20 seconds. A lot of those are wrong numbers, price shoppers, or people who hung up during the hold music. If every call counts as a conversion, Smart Bidding happily chases the cheapest junk. The CPA looks great on paper while booked jobs stay flat.

My take is to set the call conversion at around 60 seconds, then pull a sample of those calls and check them against booked jobs in the CRM before you trust the threshold. Some real leads will hang up early, sure, but a cleaner bidding signal usually matters more than the few calls you lose. What cutoff are you using, and do you let the call count or the booked job drive the bids?

I'm not sure a single cutoff like 60 seconds is the cleaner signal, because call length is a weak proxy for a lead. A plumber's emergency caller who gets booked in 45 seconds is worth more than a 3-minute price shopper who never books. For local services I'd split by call type instead: count calls to a specific number or from a click-to-call ad as the conversion, and feed the CRM "booked job" status back in as an offline conversion. Length can work as a secondary filter to reject spam, but it shouldn't be what defines a lead. It does depend on how clean your CRM data is, though.

Agreed on the split. Count the call as the conversion and let the booked job be the signal that teaches the bidding. The part that trips people up is the matching. Google can only tie a booked job back to the ad click if the call carries the click ID, and that usually depends on the number being a Google forwarding number with call tracking on. If your ads show a static business number, the call may arrive with no click attached, so your CRM status has nothing to match against and the offline upload quietly does nothing.

So pull a week of calls before you build the loop and see what share of them show a click or call ID. If it's low, fix the tracking first, because a clean CRM won't help if the records can't be linked.

Your duration filter can stay as a spam screen. Just keep it low, around 10 to 15 seconds, so emergency callers who book fast still count.

Agreed on the click ID check. I'd also treat the duration filter as a data question instead of picking a round number. Export a month of calls, bucket them by length in five-second steps, and mark which ones became booked jobs in the CRM. For most local services you'll see a floor where almost nothing shorter than some length ever books. For a plumber it might be 8 seconds. For a clinic that reads out hours and insurance options before anyone can book, it might be 25. Set the spam screen just under that floor and you lose very little real demand.

Picking 10 or 15 seconds because it sounds reasonable can either let junk through or throw out the fast emergency bookings, and you won't know which without the table. So the rule of thumb is to put the spam cutoff wherever bookings stop showing up in the buckets. Rerun the check every few months, since call routing and the mix of ads change how callers behave.

Agreed on the table, but be careful how you join it. A lot of bookings don't come from the call the ad actually got. Someone calls, hits voicemail or gets a quote, then calls back two days later and books. If you match CRM status only to the first call, the short calls look like they never convert, and you end up setting the floor too high.

Match on caller number over a window, say seven days, and credit a booking to any earlier call from that number. Then put the floor you pick into the call conversion action's minimum duration setting. That way you have one rule instead of a spam filter plus a separate bidding signal.

I'd push back on folding the floor into the minimum duration setting. That setting decides what counts for bidding, but it also hides those short calls from the conversion column, so you lose sight of how many you get and whether the spam share is moving. Keep the duration cutoff as a number you can still see, and let the CRM booking upload drive the bids.

The seven-day caller-number match worries me too. A repeat caller could be an existing customer booking something unrelated, so I'd only credit bookings where the call carries a click or call ID and keep the rest in their own bucket.

Keep the duration cutoff visible in reporting. Watching the short-call share move over time says more about spam than the bidding signal does. For the repeat-caller match, the thing that helps most is a CRM field marking whether a booking is a new job or an existing customer. By phone number alone, a regular client calling about a fresh leak looks exactly like a one-off caller who got a quote last week. Filter on that flag before crediting any booking to an earlier call.

That flag matters for bidding too, not just reporting. If a repeat customer books inside the seven-day window and you credit it to an earlier ad call, Smart Bidding learns the ad produced work it didn't. Over a few months that pushes bids toward people who already know the business, and your new-customer cost ends up looking lower than it really is.

Blank values are the other risk. Many CRMs only fill in new versus existing after the job is logged, and quick jobs sometimes never get it at all. I'd treat a blank as unknown and keep those bookings out of the window instead of crediting them by default. Does your team set that flag at the moment of booking, or is it filled in later?