Every few months someone asks me some version of the same question: "Should I just let Google handle the bidding?" And every few months my answer is the same, slightly annoying one: it depends - but it depends on things you can actually check.

Smart Bidding is Google's umbrella term for its automated, conversion-based bid strategies: Target CPA, Target ROAS, Maximise Conversions and Maximise Conversion Value. Manual CPC (and its half-way house, Enhanced CPC) is you setting the bids yourself. Both approaches work. Both approaches fail. The difference is almost never the strategy itself - it's whether the account around it is giving the strategy what it needs.

This article walks through how Smart Bidding actually makes decisions, the conditions where it genuinely outperforms a human, the conditions where it quietly wastes money, and how to move between the two without torching your performance in the process.

What Smart Bidding is actually doing

At auction time - every single time your ad is eligible to show - Google's system estimates the probability that this specific click will convert, and roughly what it'll be worth. It then sets a bid that reflects that estimate against your target. It's doing this using signals you can't bid on manually: device, location, time of day, browser, the user's search history, the exact phrasing of the query, whether they're on your remarketing list, and a long tail of other things.

That's the genuine advantage. A human can set a bid adjustment for "mobile" or "weekends". A human cannot set a different bid for "someone on an iPhone in Glasgow at 9pm who searched a four-word query and has visited the site twice before". Smart Bidding can, and does.

But here's the catch that gets glossed over in every Google rep's slide deck: the system is only as good as the conversion data you feed it. It is optimising towards whatever you've told it a conversion is. If that signal is thin, noisy, delayed or just plain wrong, the machine will optimise confidently in the wrong direction - and it will not tell you.

When Smart Bidding genuinely wins

In my experience Smart Bidding tends to outperform manual control when most of the following are true:

  • Conversion tracking is accurate and fires on the real business outcome - a purchase with its actual value, a qualified lead, a booking - not a page view or a button click
  • The campaign is producing meaningful conversion volume - as a rough guide, 30 or more conversions a month per campaign, and ideally more
  • Conversion values vary and matter - e-commerce with a wide range of basket sizes is the classic case where Target ROAS earns its keep
  • Your targets are realistic and based on the account's actual history, not on what you'd like to be true
  • You're not making big structural changes every week - the algorithm needs stability to learn

In that environment, Smart Bidding is usually the right call. It processes more signals than you ever could, it reacts in real time, and it frees you up to spend your time on the things that move the needle more than bids do: search terms, creative, landing pages and structure.

When it quietly costs you money

The failure modes are less dramatic than people expect. Smart Bidding rarely falls off a cliff. More often it drifts - spend creeps up, quality slides, and because the reported numbers still look fine, nobody notices for a quarter.

Thin or broken conversion data

If a campaign gets eight conversions a month, the algorithm is working from a sample too small to learn from. It will either bid cautiously and under-deliver, or lock onto a pattern that was really just noise. Low-volume lead gen accounts suffer here constantly.

Optimising to the wrong conversion

This is the most common and most expensive problem I come across in audits. The account imports every GA4 event as a primary conversion, so "page view of contact page" counts the same as "submitted enquiry form". Maximise Conversions does exactly what it says: it maximises the cheap, easy events. You get lots of conversions and no customers. The strategy isn't broken - it's doing precisely what it was told.

Unrealistic targets

Set a Target ROAS of 800% on an account that's historically delivered 400%, and the system will simply refuse to bid competitively. Impressions collapse, spend drops, and the account "performs" at the target because it's barely spending. The reverse is also true: a Target CPA that's far too generous gives the algorithm permission to overspend, and it will.

Lead generation where lead quality varies

If 40% of your enquiries are spam, students, or people who can't afford you, and all of them count equally as a conversion, Smart Bidding will happily chase the cheap ones. Until you can pass something back about quality - offline conversion imports, lead scoring, or at least a separate "qualified lead" event - manual or value-based control often beats pure automation here.

The middle ground most accounts should start from

This isn't a binary choice. There's a sensible progression that most accounts should follow rather than jumping straight to the most automated option:

  • Fix tracking first. Nothing else matters until the conversion signal is right. This is usually a tracking and analytics job before it's a bidding job.
  • Start new campaigns on Maximise Conversions or Maximise Conversion Value with no target, and a sensible daily budget. Let the system gather data for a few weeks.
  • Once there's a stable baseline, layer in a Target CPA or Target ROAS that's close to the actual historical number - not an aspirational one. Tighten it gradually.
  • Keep genuinely low-volume or experimental campaigns on manual or Enhanced CPC, where you can control spend directly until they've earned automation.
  • Use portfolio strategies to pool conversion data across similar campaigns when individual ones are too small to learn on their own.

Switching without wrecking performance

Changing bid strategy puts the campaign back into a learning period. Expect a week or two of volatility - sometimes more - and plan around it. A few rules that save a lot of pain:

  • Don't switch strategy, restructure campaigns and rewrite ads in the same week. Change one thing, let it settle, then change the next.
  • Don't move the target more than 10-15% at a time. Big jumps reset learning and cause erratic bidding.
  • Don't judge the change on three days of data. Give it at least two full weeks - and look at conversions by the date of the click, not the date they were reported, so late-arriving conversions don't make the early days look worse than they were.
  • Keep a log of what you changed and when. When performance moves, you want to know whether it was you or the market.

How to tell which situation you're in right now

If you want a quick gut-check on your own account, ask these questions:

  • Do I know exactly which conversion actions are marked as primary, and is every one of them a real business outcome?
  • Are my campaigns each getting enough conversions per month for the system to learn from?
  • Was my Target CPA or ROAS set based on actual performance, or on a number someone hoped for?
  • Has anyone changed the strategy or target in the last two weeks - and if so, am I judging it too early?
  • If I'm in lead gen, does the system know anything about lead quality, or just lead count?

If you answered "not sure" to more than one of those, the problem probably isn't manual versus automated. It's that the account hasn't been set up to make either approach work properly. That's fixable - and it's usually the first thing I look at in a free PPC audit.

Related Reading