Skip to content
whitead

Fix: Target CPA switched on before the campaign had enough signal to learn

finding google ads updated 2026.05.27 8 min read
// contents · 7 sections
  1. Why this matters
  2. How to verify the issue
  3. How to fix it
  4. How to confirm the fix worked
  5. Checklist — verify within 14 days of re-enabling tCPA
  6. Related rules + concepts
  7. Sources

Target CPA promises lower CPA — but only if the algorithm has enough recent signal to anchor a model. Switching tCPA on a campaign with sparse history doesn't speed learning, it stretches it. The campaign sits in Learning purgatory for weeks, CPA swings violently, and the team blames Smart Bidding for what was really a sequencing problem [1][2].

Why this matters

Google's official line on Smart Bidding is permissive: "Advertisers can start using Target CPA with no conversion history" [2]. That is technically true and operationally misleading. Google's bidding-algorithms documentation also advises waiting for a few conversion cycles before implementing a new bid strategy so the algorithm can adjust [3]. You can flip the toggle on day one, but the model has nothing to anchor against — and the cost gets paid in CPA volatility, not in a clean error message.

Three things happen when tCPA is enabled prematurely. First, the bid strategy enters Learning — Google describes it as "minor performance fluctuations as Google Ads optimizes your bids" [6], but in low-volume accounts CPA swings substantially from baseline (practitioner observation, not Google-published) as the algorithm explores with too few events to converge. Second, the Learning window extends. Google's documentation states the strategy needs "around 50 conversion events or 3 conversion cycles" to calibrate [1] — a campaign at 5 conv/week needs ~10 weeks. Third, the temptation to "fix" volatility by tweaking the target re-triggers Learning each time, and the campaign never converges.

The practitioner threshold of 30 conversions in trailing 30 days is the most cited number in the industry. Adalysis published a study of 16,825 campaigns showing that "once a campaign exceeds 30 conversions per month, it's more likely to use Target CPA than Max Conversions" [4]. Optmyzr restates it as a hard floor: "At least 30 conversions/month", because below that, "Target CPA can't gather enough conversion signals to optimize properly" [5]. This is a practitioner benchmark, not Google-published — the framing is not "Google requires 30" but "calibration depth is shallow below 30, and the cost of premature switching outweighs the marginal benefit."

See Smart Bidding and Smart Bidding learning phase for upstream mechanics.

If you just launched. This rule requires temporal data — it analyses bid-strategy history alongside conversion volume over a trailing window. If your account is new (<30 days since first conversion), rule status will be insufficient data with the message "Temporal analysis available from the next audit." Re-run after the account has accumulated 30+ days of conversion history.

How to verify the issue

  1. Pull the bid-strategy history. In Tools → Shared library → Bid strategies, or at campaign level via Campaigns → (campaign) → Settings → Change history, note the date tCPA was enabled.
  2. Count conversions in the 30 days preceding that switch. Open Reports → Predefined → Time → Day, filter to the campaign, sum Conversions for the 30 days ending the switch date. If the total is below 30, the switch was premature by the practitioner benchmark [4][5].
  3. Check current bid-strategy status. In Bid strategies, look at the Status column. If Learning has persisted >14 days, the algorithm is still calibrating — consistent with premature switching [6]. Brief Learning periods (≤7 days) are normal.
  4. Audit recent target-CPA edits. Each target change above ~20% retriggers Learning [1][6]. Frequent target edits in the first 30 days post-switch indicate the team tried to course-correct instability, extending Learning further.
  5. Confirm conversion tracking is clean. Open Tools → Conversions and verify the primary conversion action has Status: Recording with no anomalous spikes or zero-days in the 30 days before the switch. tCPA optimises against whatever fires — corrupt signal poisons the algorithm regardless of volume. See conversion tracking.

If steps 2 and 3 both confirm — fewer than 30 conv/30d at switch time and Learning >14 days — the finding holds.

How to fix it

Time estimate: 30 minutes to revert, 2-6 weeks to accumulate signal, 15 minutes to re-enable tCPA correctly.

  1. Stop the bleed. Open Campaigns → (campaign) → Settings → Bidding → Change bid strategy. Switch from Target CPA to Maximize Conversions without a target. It uses Smart Bidding's auction-time signals but does not anchor to a target the algorithm cannot defend. Manual CPC is the more conservative alternative if you want predictable spend.
  2. Set a budget that supports learning. Underspending starves the algorithm [5]. Confirm the daily budget is at least 5-10× your expected CPA, so the algorithm has room to explore.
  3. Freeze structural changes for 7-14 days. Do not edit keywords, ad groups, creatives, or budgets above ±20%. Each change above ~20% retriggers Learning [1][6] — and Maximize Conversions inherits the same constraint.
  4. Track conversion volume against the 30/30 gate. Once the campaign clears 30 conversions in trailing 30 days (practitioner benchmark, not Google-published) [4][5], it is structurally ready for tCPA. If you cannot clear that threshold in 60 days, the problem is upstream (audience, keywords, budget), not the bid strategy.
  5. Re-enable tCPA with a defensible target. Pull the campaign's average CPA from the last 30-60 days of Maximize Conversions. Set initial tCPA at or 10-20% above that historical CPA [7] — this gives the algorithm room to learn. Do not anchor tCPA 30-50% below historical; the strategy throttles delivery rather than "stretch" toward an unattainable target.
  6. Freeze again for 14 days post-switch. Google describes calibration as needing "around 50 conversion events or 3 conversion cycles" [1] — for a campaign at 30 conv/30d, that is roughly 14-21 days. Read-only window.
  7. Tighten target gradually. After 14+ clean days, tighten by no more than 10-15% per fortnight. Aggressive cuts re-trigger Learning and cost volume more than they save CPA [7].

For accounts below the 30/30 gate that cannot accumulate volume in 60 days, see Smart Bidding learning phase — the longer-term answer may be Portfolio Bid Strategies that pool signal across multiple sub-scale campaigns.

How to confirm the fix worked

Checklist — verify within 14 days of re-enabling tCPA

  • Conversions/30d ≥ 30 at re-enable moment, confirmed in Reports → Predefined → Time → Day.
  • Bid-strategy status clears Learning and shows Eligible within 7-14 days [1][6]. If still Learning past day 14, do not edit — wait another 7 days before intervention.
  • CPA stabilises within ±20% of target by day 14. Persistent CPA >50% above target signals an aggressive target or poor signal quality.
  • Conversion volume within ±20% of pre-switch baseline by day 14. Larger drops usually mean the target is set too low and the strategy is throttling delivery [7].
  • No mid-Learning edits — change-history shows zero target adjustments, no >20% budget swings, no major composition changes during the 14-day window [6].
  • Conversion tracking still recording cleanly with no zero-days in trailing 14 days. See conversion tracking.
  • No re-entry into Learning triggered by downstream changes [6].

If all seven pass, the finding closes. Most common partial-pass: CPA stabilises but volume drops sharply — target was set below historical CPA and the strategy is starving. Raise the target by 10-15% and let Learning re-converge.

Blast radius. This finding is sequencing-driven and co-occurs with campaign_learning_stuck (campaign sits in Learning >14 days — see Learning stuck) and bidding_strategy_goal_alignment (campaign on a strategy that does not match the goal — see bidding strategy alignment). Sequencing matters: if you "fix" tCPA by switching to Maximize Conversions, then discover the campaign needs Target Impression Share instead, the second switch retriggers Learning and burns another 14 days. Diagnose goal-alignment first, signal-volume second, then sequence bid-strategy moves accordingly. Accounts that fix this finding in isolation typically also fix troas_without_value_tracking on the next audit — both are signal-quality failures dressed as bidding failures.

Sources

  1. Google Ads Help — Duration of the learning period for campaigns and what affects it. https://support.google.com/google-ads/answer/13020501 (accessed 2026-05-27)
  2. Google Ads Help — About Target CPA bidding. https://support.google.com/google-ads/answer/6268632 (accessed 2026-05-27)
  3. Google Ads Help — How our bidding algorithms learn. https://support.google.com/google-ads/answer/10970825 (accessed 2026-05-27)
  4. Adalysis — When should you switch from Max to Target bidding? https://adalysis.com/blog/when-should-you-switch-max-target-bidding/ (accessed 2026-05-27)
  5. Optmyzr — When and How to Use Each Google Ads Smart Bidding Strategy. https://www.optmyzr.com/blog/smart-bidding-strategies/ (accessed 2026-05-27)
  6. Google Ads Help — About bid strategy statuses. https://support.google.com/google-ads/answer/6263057 (accessed 2026-05-27)
  7. Google Ads Help — About Smart Bidding. https://support.google.com/google-ads/answer/7065882 (accessed 2026-05-27)
// was this useful?
// anonymous · no personal data stored
← Back to Wiki