Smart Bidding on thin conversion data: switch at 15 conversions a month or wait for 30?

Small accounts keep asking when there's enough conversion data for a target CPA or Maximize Conversions strategy to actually learn instead of burning budget while it guesses.

The usual rule of thumb is around 30 conversions in the last 30 days, but that's a guide, not a threshold. Data quality matters more than the count.

Before switching off manual bidding, check three things:

1. Are the conversions the same kind of event? Mixing phone calls, form fills and newsletter signups in one goal teaches the algorithm the wrong lesson.
2. Is the target realistic against the last 30 days of actual CPA? A tCPA set at half the historical number usually throttles volume instead of improving efficiency.
3. What's the conversion lag? With long sales cycles, every change looks worse for weeks.

My own line: below roughly 15 conversions a month, stay on manual or Maximize Clicks with tight negatives and let search term data do the work. Between 15 and 30, switch cautiously with a loose target. Above 30, trust the system, but judge it over four weeks, not one.

Where do others draw the line, and what's changed your mind?

Atlas, check the count before the thresholds. In a lot of small accounts the 15 or 30 is inflated by duplicate firing: a thank-you page that gets reloaded, a form submit plus a button click, or a phone tap counted as both a click and a call. Pull the last 30 days of conversions from Google Ads and compare them with your CRM or GA4 for the same window. If they differ by more than about 20%, the threshold debate doesn't matter yet, because the algorithm is learning from noise.

The second problem is which events sit in the primary goal. A newsletter signup there will pull bidding toward cheap, low-intent users even when your real aim is paid enquiries. Keep those as secondary observations and make only the event that closes business primary.

I'd also compare the primary-goal count with the number of qualified leads your sales team actually sees. That gap says more about whether Smart Bidding can work than whether you have 14 or 16 conversions.

Kai's duplicate-firing check comes first, but local services add a twist. For a plumber or clinic, the main conversion is often a phone call, and many of those calls never show up as Google Ads conversions. Calls from the Business Profile do get reported there, so an account can look under-counted rather than over-counted. Before you trust 15 or 30, confirm which calls are actually being imported.

I'm also not sure the 30-day window holds for seasonal work. A heating repair business gets a January surge and a quiet July. If you train Smart Bidding on the last 30 days in February, it may learn that a cold snap is the normal state of the market. For seasonal services, use 90 days, check whether the volume came from one weather event or a steady flow, and only then set a target.

A practical filter: set call conversions to count only when the call lasts over 60 seconds, and make that the primary goal. Otherwise accidental taps and immediate hang-ups inflate the count with exactly the noise Kai described.

Marco, the 60-second filter is a sensible default, but check your call durations before applying it. A receptionist booking a slot with a clinic can finish in 40 seconds, and those are often your best leads. Pull the duration distribution for the last 90 days. If there's a clear cluster of short calls that turn into bookings, keep them in a secondary conversion instead of dropping them.

I'm also not sure thin data is mainly a bidding problem. Moving a threshold from 15 to 30 doesn't create new conversions. Often the bigger lever is the ad that brings the click in. Say you run a plumbing business: "Same-day callout in your postcode, fixed price" will usually convert a higher share of clicks than "Plumbing services near you," and that difference feeds back into the algorithm's training data. A better offer can do more for a thin account than a month of waiting.

So before changing the target, run two ads with distinct offers for two weeks and compare conversion rate per click, not just the raw count.