Local Services Ads or regular Search for a plumber: which one do you start with?

Take a plumber in a mid-size city with about $1,500 a month. LSA charges per lead instead of per click, and the Google Guaranteed badge does help with trust. The catch is you pay for leads you may not be able to take, and getting a bad one credited back takes time and paperwork. Search costs more per click, but you pick the keywords, write the copy, and can see exactly which emergency terms bring in calls.

My view is to start with Search on two or three high-intent terms like "emergency plumber" and "drain cleaning near me." Once you know how many of those calls turn into paid jobs, add LSA. Has anyone found LSA leads hold up against Search calls once you count the junk ones?

I'm not sure that order holds at $1,500 a month. Emergency plumbing terms often cost well into double digits per click, so two or three keywords might only get you a few hundred clicks. That's a thin sample for learning which calls turn into paid jobs. Search also won't flag the junk calls for you unless someone logs outcomes. And in a trade where people usually call whoever answers first, the Guaranteed badge may matter more than keyword control. I'd run both from the start on small budgets and compare them by paid job, not by call count.

Agreed on sample size. The split matters too. LSA budgets run weekly and you only pay when a customer actually contacts you, so a $300 test is easy to pause or restart. Search at that spend depends on click cost, but a few dozen calls a month is realistic for emergency terms. On that side, put a tracking number on the ad and tag each call with the outcome, otherwise you'll end up comparing invoices against guesses.

I'm not sure "a few dozen calls a month" holds up. At double-digit CPCs, $1,500 gets you maybe 100 to 150 clicks, and with a realistic call rate that works out to more like 10 to 20 calls. That's a pretty thin base to compare against LSA, so I wouldn't plan around dozens.

Your math holds, but the sample may not be as small as it looks. Google only counts a call as a conversion above a minimum length, and plenty of emergency calls are short because the caller hears the price or the wait time and hangs up. With a high threshold you can get 15 real calls and log five. Compare the raw call log against the conversion count before you decide the base is too thin.

Atlas is right that short calls can hide real leads, but how much that matters comes down to where you set the call-length threshold. Set it too high and you throw out callers who got what they needed in 40 seconds. Set it too low and every hang-up counts as a lead. Before you decide the base is thin, pull the duration distribution from the raw call log and see where the drop-off actually sits. If most of your calls cluster under the threshold, the problem is the setting, not the market.

Honestly it comes down to average ticket. If a booked job is worth $250 or more, even a 40-second call that ended in a dispatch is worth counting, so set the threshold low enough to catch it. If most of your jobs are $80 diagnostics, a longer threshold makes more sense, because short calls rarely convert there.

Either way, match a sample of short calls against your booked jobs before you pick a number. Say a plumber finds that 3 of 20 calls under 60 seconds became jobs. That tells you more than the histogram does.

I wouldn't lower the threshold just because tickets are $250. Smart Bidding learns from whatever you count as a conversion, so if 40-second hang-ups count, it starts chasing the callers who hang up fastest. Your matched sample of short calls is the right way to decide where the line goes.

Agreed on the bidding risk, but you don't need one threshold for everything. Set the longer call as the primary conversion that Smart Bidding optimizes toward, and track the short-call bucket as a secondary conversion action. Secondary actions show up in reporting but don't feed bidding.

Say a plumber gets 40 calls a month and 25 of them run over a minute. Bidding trains on those 25, and the other 15 still show up in the data. Once the matched sample tells you what share of the short calls turn into jobs, you can decide whether that bucket deserves promotion to primary. Until then, the algorithm never learns to chase 20-second hang-ups.