Smart bidding won't fix an account that's structured wrong

Rising CPA in Google Ads is almost always a structural problem, not a bidding one. Here's how to tell the difference — and what to fix first.

‍

When Google Ads CPA starts climbing, the default response is to adjust the bid strategy. Lower the tCPA target. Switch from Maximise Conversions to Target ROAS. Give the algorithm a nudge.

‍
Sometimes this works. More often, it delays the diagnosis of the actual problem — which is almost always structural. Bid strategy adjustments rarely fix an account that's fundamentally built wrong. They sometimes mask the problem for a few weeks. But the CPA creep returns, the algorithm wobbles, and the account ends up on a cycle of constant tweaking with no real improvement.

‍
Here's what structural problems actually look like — and how to fix them in the right order.

What "structure" actually means in Google Ads

Account structure refers to how your campaigns, ad groups, and assets are organised — and specifically how that organisation aligns with your conversion goals, commercial priorities, and the signals you're giving the algorithm.

‍
A well-structured account gives smart bidding a clean, coherent signal: these are the conversions that matter, this is where the budget should focus, and here is the audience context that helps it bid intelligently.

‍
A poorly structured account sends conflicting signals — multiple conversion actions with very different values weighted equally, campaigns cannibalising each other's traffic, or ad groups so broad the algorithm has no meaningful pattern to learn from. Smart bidding is only as good as the foundation it sits on.

‍

The five structural problems that kill smart bidding performance

‍1. Soft conversions mixed with hard conversions
If your campaign optimises for both "purchase" and "newsletter signup" as equally-weighted goals, the algorithm will happily serve ads to people who sign up for newsletters but never buy. Your reported CPA might look fine. Your actual return on ad spend probably isn't.

‍
Fix: assign conversion goals by campaign. Purchase-focused campaigns optimise for purchases only. Soft conversions (scroll depth, session duration, page views) can be tracked but must not be set as primary bidding goals.
‍

2. Too many campaigns competing for the same traffic
Common in accounts that have grown over time: a brand campaign, a general search campaign, a PMax campaign, and a legacy RLSA campaign all bidding on broadly similar queries, all drawing from the same budget, all reporting partial credit for the same conversions.

‍
Overlapping campaigns dilute the conversion signal each receives, extend learning phases, and create internal auction competition. Consolidate aggressively. One campaign per job.
‍

3. Asset groups with no audience signal differentiation
A single PMax asset group covering your entire product range with an "all users" audience signal gives the algorithm nothing to work with. It learns slowly and allocates budget without commercial logic.

‍
Structure asset groups by product category or customer segment. Apply tight, commercially meaningful signals to each — customer match lists, in-market audiences aligned with your category, or remarketing lists from high-intent pages.

‍
4. Conversion volume too low for smart bidding to function
Smart bidding needs data. Google recommends a minimum of 30–50 conversions per campaign per month for tCPA or Target ROAS to function reliably. Below that, the algorithm is guessing.

‍
If campaigns are conversion-starved: consolidate to pool data, use a softer-but-still-commercial conversion action (add to cart rather than purchase) as the primary goal, or temporarily switch to Maximise Conversions without a target to let it gather signal first.
‍

5. Broad match without audience signal coverage
Broad match in 2026 interprets intent loosely and will match queries you'd never have chosen manually. It works well when paired with first-party audience signals and smart bidding. It works poorly when the audience layer is thin or absent.
‍

If CPAs have been creeping since you expanded to broad match, this combination — broad
match without meaningful audience signals — is likely a significant factor.

‍

How to tell if your problem is structural or algorithmic

A quick diagnostic:

‍
Structural problem signals

• CPA trending upward over 8+ weeks despite multiple bid adjustments
• Conversion volume inconsistent week to week with no seasonal explanation
• Different campaigns showing very different CPAs for apparently similar traffic
• Smart bidding campaigns repeatedly entering and exiting learning phase
  
  

Bidding / algorithmic problem signals

• CPA spiked after a specific change (budget increase, bid target change, audience update)
• CPA elevated on specific devices or days but not others
• Performance recovered after a bid adjustment before reverting
  ‍

If the pattern is the former: fix the structure. If it's the latter: the bidding mechanism may need
attention — but at least you're diagnosing the right thing.

‍
The right order of fixes

• Audit conversion goals — remove soft conversions from primary bidding goals first
• Consolidate overlapping campaigns — one campaign per commercial job, no cannibalisation
• Review audience signals on PMax asset groups — replace "all users" with commercial signals
• Check conversion volume per campaign — consolidate if below 30/month
• Review broad match coverage — ensure meaningful audience signals are in place before expanding

Don't try to fix everything at once. Each significant structural change triggers a fresh learning
phase. Make one change, give it 3–4 weeks, evaluate, then move to the next.

‍
What to do this week

• Open your campaign list and flag any with fewer than 30 conversions in the last 30 days — these are consolidation candidates
  ‍
• Pull your conversion action report and check whether soft conversions are set as primary goals in any campaign
  ‍
• For any PMax campaign, confirm whether asset groups have specific audience signals or are running on "all users"
  ‍
• If CPA has been rising for more than 6 weeks and bid adjustments haven't held, treat
  this as a structural problem until proven otherwise

‍

FAQ

How do I pick the right primary conversion goal?

Use the action that most directly represents revenue or qualified intent. For ecommerce: purchase. For lead gen: the form submission that has historically correlated with closed deals
in your CRM — not just any form fill. If you're unsure, look at which conversion actions have
the highest downstream close rate, not the highest volume.

‍
Is consolidation always the right move?

Not always. Some accounts legitimately need separate campaigns for different geographies,
audiences, or product lines with very different margin profiles. The question is whether separate campaigns serve a strategic purpose — or whether they're historical accumulation
that nobody has cleaned up.

‍
How long until I see improvement after structural changes?

Expect a 4–6 week recalibration period after any significant structural change. CPA may
temporarily worsen before it improves as the algorithm rebuilds its model. Budget for this —
and don't panic-revert changes in week 2.

‍