Google Ads secondary conversions: do micro-steps belong in bidding or just in reports?

A common setup mistake is marking every micro-step as a primary conversion. Say a store gets around 40 purchases a month and also tracks add-to-cart and begin checkout. If all three are primary, Smart Bidding ends up fed mostly the cheap signals. The CPA target can look healthy while real purchases fall, because the bidder is chasing the easiest actions. Demoting the micro-steps to secondary keeps them visible in reporting without letting them steer bids, so my default is secondary for anything below a purchase or qualified lead. The counterargument is that at low purchase volume, add-to-cart might be the only thing giving the model enough signal to learn at all. Where do you draw the line, and has it mattered at your volume?

The line I'd draw is less about how many actions are primary and more about whether they carry honest values. With 40 purchases a month, add-to-cart is probably the only thing giving the bidder enough data, so I wouldn't demote it outright. The trouble starts when a cart add and a purchase each count as one conversion, because then the bidder treats a cheap signal as equal to revenue.

If you keep add-to-cart primary, give it a fixed value that matches its real odds. Say your average order is $80 and about 1 in 10 carts converts. A cart add is worth roughly $8, so set it there and leave purchases at the actual order value. Check the value column in your conversion actions too, and make sure the numbers aren't all defaults.

Honestly, 40 purchases is thin for Smart Bidding either way. A target CPA set from that volume will swing around, so judge the micro-step decision over a few months of data, not one week.

A fixed cart value is still a guess, and once it gets summed in with real orders, blended ROAS can end up looking better than what sales actually sees. If you're going to count carts at all, use your real cart-to-purchase rate to discount them. What is that number for you?

Same logic applies to local services, where the micro-step that matters is usually the phone call. Say a plumber gets 60 calls a month. Some are wrong numbers, some are hang-ups, and some are people asking a price who never book. If every ring counts as a conversion, the bidder learns to chase whatever produces a call, not a booked job.

So set a minimum call duration on the call conversion action so very short calls don't count. Then value the calls that remain as booked-job rate times average job value, the same way you'd discount a cart. The local equivalent of cart-to-purchase rate is call-to-booking rate, and you should pull that from your own call logs or CRM before putting any number in the value column. The duration threshold is the setting to check first, because it quietly inflates volume and nobody notices.

Marco, the duration threshold is a smart move, but I'm not sure one blended call-to-booking rate holds up. It depends on whether the rate is roughly the same across your campaigns. Someone calling after a brand search already knows the company and usually books at a much higher rate than someone who typed "emergency plumber near me", so a single averaged value overpays for the weaker source. If the rates are close, one value is fine. If they're far apart, split calls by campaign and value each group from its own booking data.

Priya's right, and the split only matters if your call mix shifts over time. If brand and emergency searches stay a steady share of call volume, one blended value is roughly fine even when the rates differ. The trouble starts when a new campaign or a seasonal push tilts the mix toward weaker sources. Then the blended value quietly overpays and nothing in the numbers tells you. So check the source split each month before deciding one value is still good enough.

Kai's mix shift point is the real test, but whether you can actually see it depends on call volume. At 60 calls a month, a few brand calls more or less can swing the blend, so checking monthly will mostly show you noise. Once you're at a few hundred calls, the source split is worth checking every month.

Fair, but I think what decides whether a monthly check tells you anything is how far apart your source booking rates are. If brand calls book at around 50% and emergency searches at 15%, shifting ten calls from one to the other moves the blended value a lot, so even 60 calls a month can show a change worth acting on. If the two rates sit close together, a few hundred calls could reshuffle the mix and the blend would still hold up.

The catch is that you can't see the spread from the call log alone. You need booked-job data per source, and for a plumber that often takes a few weeks to fill in, since a Monday call may not book until the following week. Until you have that, a quarterly split check will tell you more than a monthly one.