First, confirm the fall against the right baseline
The most common false alarm after a switch is an unfair comparison. Recent days look weak because their conversions have not all arrived yet: Google Ads credits conversions to the day of the click, and a customer who clicks today and buys next week will not appear in today's figures until then. Comparing the last few days with a closed earlier period always flatters the past.
Build the baseline carefully. Use a pre-switch period with the same weekdays and no promotion or holiday in it. Exclude the most recent days whose conversion window is still open, or read them by conversion time instead. Compare the totals that matter for the strategy: CPACost of each conversion. and conversion volume for target CPA, conversion value and ROASRevenue generated per unit of ad spend. for target ROAS. Note the switch date from change history and check whether anything else changed that week; the change history audit tutorial sets out how to classify every edit in the window and test which one explains the fall. If conversions are not merely fewer but absent, the conversions not tracking diagnosis comes first, because no strategy can bid well on a broken signal.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Volatile results; bid strategy status reads learning | Bid strategy status and its hover text; switch date | Learning period after the switch | Wait without further edits; judge after the status settles |
| Volume falls steadily; spend below budget | Target versus actual CPA or ROAS before the switch | Target set tighter than recent actuals | Set the target at recent actuals, then tighten in steps |
| Status shows limited by budget; ads stop early | Lost impression share (budget); budget versus the new strategy's spend | Budget too small for the strategy to bid freely | Raise the budget or move to a strategy without a target until data builds |
| Status mentions limited data; erratic bids | Conversions per month in the campaign or portfolio | Too few conversions for the strategy to learn from | Pool campaigns, add a broader upper-funnel action, or delay the switch |
| Recent days look poor, then recover in later reports | Conversions by conversion time; typical lag between click and conversion | Conversion delay distorting recent results | Exclude open windows from judgement |
| Conversions changed shape on the switch date | Conversion actions edited, added or reprioritised the same week | Conversion actions changed at the same time | Restore the previous action set, or let the strategy relearn on the new one alone |
Fix, verify and prevent each cause
Still learning
The help centre documents that performance can fluctuate while a newly created or significantly changed strategy adjusts, and that the status column shows the learning state. Volatility in that period is expected, not a verdict. What the Smart Bidding guide says about each status applies here; this page only asks whether the status is still active.
Fix: do nothing to the campaign except keep the budget steady and avoid new edits.
Verify: the status moves to eligible and results become steadier across comparable weeks.
Prevent: schedule switches for a quiet period with no promotion, so learning happens on representative traffic.
Target tighter than reality
A common pattern is to set the new target at the CPA the business would like rather than the CPA the campaign achieved under the old strategy. The system then bids for only the cheapest conversions and volume collapses. The same applies to a target ROAS set above what the campaign has returned.
Fix: reset the target to the recent actual figure, measured over a period with closed conversion windows, and tighten it later in steps. The CPA calculator and the ROAS calculator help derive a target from the economics rather than from hope.
Verify: spend and conversions recover toward the pre-switch level at a similar acquisition cost.
Prevent: write down the pre-switch actuals before every change and start the target there. If volume fell but costs did not move, the underspending campaign diagnosis covers the other ways a strategy can fail to use its budget.
Budget too small for the new strategy
A conversion-maximising strategy tends to spend the full budget, and one with a target needs room to bid higher in promising auctions. A budget that barely covered manual bidding may cap the new strategy before it can find its conversions, and the status reports the limit.
Fix: give the campaign enough budget for the strategy's learning, or keep a strategy without a target until data builds and the budget can grow. The interaction is explained in budgets and pacing, in the section on budget changes under automated bidding.
Verify: lost share to budget falls and the status no longer reports a budget limit.
Prevent: check the budget against the strategy's needs before switching, not after.
Too few conversions to learn from
Automated bidding learns from conversions, and a campaign that records only a handful gives it very little to go on. The status may report a data limitation, and bids swing as each new conversion moves the model.
Fix: pool similar campaigns into a portfolio strategy, add a meaningful upper-funnel action as a primary conversion where it predicts sales, or stay on a simpler strategy until volume grows. The choice of which actions to count belongs to the measurement hub.
Verify: conversion volume in the pooled or broadened setup is enough for the status to settle.
Prevent: check conversion volume as part of the decision to automate.
Conversion delay misread as decline
If customers usually convert days after clicking, the most recent days will always look worse than the period before them, and every early look at a new strategy is skewed by it.
Fix: compare conversions by conversion time, or exclude the latest open window from the comparison. The causes of gaps between click date and conversion date, and how they differ from GA4, are covered in the Google Ads and GA4 numbers guide.
Verify: the same period, re-read later, shows the gap closing.
Prevent: add the typical conversion lag to the reporting template so nobody judges a strategy on unfinished days. If the lag regularly runs beyond the action's click window, the guide to conversion windows and attribution settings resets the window from that lag, as a separate change made before the next strategy switch.
Conversion actions changed at the same time
Changing which actions are primary, adding a new action, or editing values in the same week as a strategy switch gives the system a new definition of success to learn on top of the new strategy. Results look broken because the target is now measured against different numbers.
Fix: decide which change you meant; keep one and revert the other, or accept a longer learning period. Action settings are covered by the conversion tracking setup tutorial.
Verify: the conversion column's definition is stable across the comparison period.
Prevent: never change the strategy and the conversion actions together.
Prevention: switching strategies without losing a month
A switch can be made with little lost performance if it is treated as a planned change rather than an edit. Before it, record the baseline with closed conversion windows, confirm tracking with a test conversion, check the conversion volume the campaign records, and size the budget for the new strategy. Choose a quiet period. Where the campaign is large enough, run the new strategy as a campaign experiment against the old one, so the comparison happens at the same time on split traffic rather than across two different months. Start the target at recent actuals. After the switch, change nothing while the status shows learning, and read results only on days whose conversions have arrived. Keep a note of the expected CVRShare of visits or clicks that convert. and acquisition cost so the review has something concrete to compare with.
A training provider that panicked after moving to target CPA
A training provider moves its main search campaign from manual bidding to target CPA on 2 June, with a target of 45; under manual bidding its actual CPA over the previous eight weeks was 63. In the same week it adds a "brochure download" action as primary. Ten days later, course bookings have fallen from about 30 a week to 12, and the team considers switching back. Instead it reverts the download action to secondary, resets the target to 62, and reads results by conversion time, since bookings typically follow a click by five days. Four weeks after the reset, bookings run at 31 a week at a CPA of 58, and the target is lowered to 55.
If the change was adding a target CPA to a campaign that ran on maximise conversions, the comparison of maximise conversions and target CPA explains how the constraint shifts from the budget to the target.
Common mistakes
Judging the new strategy on days whose conversions have not arrived. Setting the target from the business's wish rather than the campaign's record. Editing targets every few days during learning. Switching strategy and conversion actions in the same week.
