Quick Answer: Smart Bidding fails most often because the campaign structure it's running on was built for manual bidding. Device bid adjustments, match-type-segmented campaigns, and RLSA audience targeting all fragment your conversion data into buckets too small for the algorithm to learn from. Consolidate the structure first, then switch to automated bidding.
The Most Common Smart Bidding Mistake
You turn on Target CPA or Maximize Conversions. Performance drops for two weeks. You revert. You tell yourself Smart Bidding doesn't work for your account.
What actually happened: the automated bidding strategy ran out of signal before it had a chance to learn.
Google's Smart Bidding learns at the campaign level. It needs a steady stream of conversion data to figure out which users convert, at what times, on what devices, from which queries. When you've built a highly segmented campaign structure, that stream gets split into a dozen thin trickles. Each campaign sees too few conversions to train the algorithm. The learning phase drags on. CPCs inflate while the system probes the auction. You cut it off before it ever had a real shot.
The bidding strategy didn't fail. The structure failed the bidding strategy.
Why Manual-CPC Structure Fights Smart Bidding
Manual CPC rewards granular control. You want device bid adjustments so you can push harder on mobile when it converts. You want separate campaigns for broad, phrase, and exact so you can control spend by match type. You want RLSA audience bid adjustments cranked up for warm audiences.
That structure makes total sense when you're setting bids by hand. You're doing what the algorithm would do if it had the data -- making educated guesses about which segments perform better and adjusting accordingly.
But Smart Bidding doesn't need pre-segmented structure. It reads device, audience, query, and time-of-day signals on its own, in real time, across every auction. When you've already split the account by those dimensions, you haven't helped it. You've removed the data it needs to learn, and you've added restrictions that limit what it can see.
Over-segmentation is the most common reason Smart Bidding underperforms for accounts that should be perfectly suited for automation.
How to Fix It Before You Switch
Step 1: Collapse device bid adjustments
Go into every campaign you're planning to run on Smart Bidding and set all device bid adjustments to 0%. Not negative. Zero. You're not saying mobile doesn't matter -- you're removing a manual override that competes with what the algorithm is already doing. Smart Bidding factors in device performance automatically. Your adjustment on top of that is noise.
Step 2: Merge match-type-segmented campaigns
If you're running separate campaigns for broad, phrase, and exact on the same keyword themes, consolidate them into one campaign before switching bidding strategies. The goal is to give the algorithm one large pool of conversion data instead of three smaller ones. A campaign earning 40 conversions a month learns faster and more accurately than four campaigns each earning 10. The same total spend, far better signal.
Step 3: Shift RLSA layers to observation
If you have RLSA audiences set to targeting (not observation), you're restricting who Smart Bidding can show ads to during the learning phase. That's backwards. Switch them all to observation before you flip the strategy on. Observation lets the algorithm see the full auction, identify which audiences actually convert in this account, and then adjust bids on its own -- which is what you wanted the manual RLSA adjustment to do in the first place.
Step 4: Set a realistic starting target
Once the structure is consolidated, set your Target CPA at or near your current actual CPA over the last 30 days. Don't start aggressive. Give the algorithm a target it can hit immediately and let it stabilize before you tighten. If you start 30% below current CPA, you're asking it to restrict spend before it's learned anything.
What Happens When You Get the Structure Right
The learning phase shortens because each conversion lands in one bucket instead of several. Google's automated bidding has a complete picture of which queries, audiences, and devices convert -- not a fragmented sample from a single segment. CPCs stabilize faster. Performance holds through the transition instead of cratering.
You're not giving up control. You're giving the algorithm the data it needs to do what you hired it to do.
The advertisers who write off Smart Bidding as broken almost always have the same thing in common: a structure built for manual bidding that was never updated before automation was turned on.
Get the Full Account Structure
If your account is over-segmented and you want a clean framework for rebuilding it around Smart Bidding, Signal Rich Structures walks through exactly this -- how to consolidate campaigns, handle audience signals, and structure your account so automated bidding actually works. Get the details at freak.marketing/signal-rich-structures.