How to Recover Google Ads Smart Bidding After a Tracking Outage

A conversion tracking outage on Google Ads is really two problems wearing one trench coat. The first is the broken tag, and fixing it usually takes an hour. The second is what Smart Bidding learned while the tag was down, and that one follows you around for weeks if you handle the recovery wrong. This is the playbook for problem two.

Why the damage outlives the outage

Smart Bidding strategies like Target CPA and Target ROAS steer by recent conversion data, weighted heavily toward the last few weeks. When your tag stops firing, the algorithm does not know tracking broke. From its point of view, your traffic suddenly stopped converting. Target CPA responds exactly the way you would expect from a system told "nothing converts anymore": it becomes conservative, bids down, enters fewer auctions, and in bad cases nearly stops spending. Advertisers describe this as the campaign "being strangled," and that is a fair description.

Meanwhile, your lifetime conversion history helps far less than people hope. The system leans on the recent window. Five hundred conversions across the campaign's lifetime do not offset fourteen days of apparent zero. So the recovery is not "fix the tag and wait." It is: fix the tag, cut the poisoned data out of the record, rebuild confidence, then restore your targets. In that order.

Step 1: Confirm the fix is real, not just claimed

Before anything else, verify conversions are actually recording again. Not "the ticket was closed," not "the developer said it's deployed." Fire a real test conversion and watch it arrive: check the tag with Tag Assistant, then confirm the conversion action shows "Recording conversions" and the test appears in the interface (allow for normal reporting lag). If you skip this, everything that follows builds on hope. Our guide to confirming your Google tag walks through the checks, and why conversion tracking stops working covers the usual root causes.

Also write down the outage boundaries as precisely as you can: the last day conversions look normal, and the first day they look normal again. You need those dates for the next step. If the break traces to a specific change, a deploy, a plugin update, a consent banner swap, note that too: the change date is usually the true start of the outage, even if reported conversions limped along for another day or two on lagging attribution before hitting zero.

Step 2: Create a data exclusion for the outage window

Google Ads has a purpose-built tool for exactly this situation, and most advertisers have never opened it. Data exclusions tell Smart Bidding to ignore a date range when learning. The zero-conversion days stop counting as evidence about your traffic, which is the single biggest lever you have for shortening the recovery.

1. In Google Ads, go to Tools, then Budgets and bidding (in older layouts: Bidding strategies), and find Data exclusions.
2. Create an exclusion covering the full outage window, from the last clean day to the first verified-clean day. Be generous at the edges. A partial broken day counts as broken.
3. Scope it to the affected campaigns and the affected conversion actions. If only one tag broke, do not exclude data from actions that kept working.

Two honest caveats. Data exclusions work best for outages measured in days, not months. If tracking was broken for six weeks, excluding the entire period leaves the algorithm with very little recent signal either way, and you should treat the campaign as effectively new. And exclusions are not retroactive magic: the sooner you add one after an outage, the less time Smart Bidding spends learning from garbage.

Step 3: Loosen the constraint while it relearns

With the exclusion in place, the algorithm needs room to rediscover that your traffic converts. A tight Target CPA is the opposite of room. Two workable approaches, pick one based on how much control you need:

• Keep tCPA but raise the target 20-30% for the recovery period. This keeps a guardrail up while letting the system win auctions again and collect fresh conversions.
• Switch to Maximize Conversions without a targetfor one to two weeks. This is the "let it breathe" option. It recovers volume fastest, at the cost of some expensive conversions along the way. Watch spend daily, because without a target the system will happily use your whole budget.

What matters more than which option you choose: do not flip-flop. Every bid strategy change resets learning. Choose one, give it 7 to 14 days, and judge it on conversion volume returning, not on day-three CPA.

One mechanic that trips people up while they wait: conversion delay. Google attributes conversions to the click date, not the conversion date, so the last few days in any report always look worse than they will end up. During recovery this effect is amplified, because you are anxiously refreshing exactly the window where data is least complete. Decide up front that you will evaluate trailing 7-day windows that end at least three days ago, and that no single day means anything. Recovery judged on complete windows looks like steady progress. The same recovery judged on this morning's numbers looks like chaos, and chaos is what makes people flip strategies and start the clock over.

Step 4: Restore your targets gradually

Once conversions have flowed normally for at least a week, and ideally volume is back near pre-outage levels, tighten back up. Restore Target CPA at a slightly higher number than your old target, then walk it down 10-15% at a time, waiting several days between moves. Big sudden target drops put campaigns right back into the choked state you just escaped. Slow is fast here.

A worked recovery timeline

To make the sequence concrete, here is what a clean recovery looks like for a campaign that normally does 10 conversions a day and lost its tag for a week:

• Day 0: outage discovered. Tag fixed and verified with a real test conversion the same day. Outage window written down: last clean day through today.
• Day 0, same sitting: data exclusion created covering the entire window, scoped to the affected campaigns and conversion actions. Target CPA raised 25%, or switched to Maximize Conversions if delivery already collapsed to near-zero spend.
• Days 1-3: spend recovers first, conversions lag a day or two behind due to conversion delay. Do not react to the scary-looking CPA of these days; the denominator is still catching up.
• Days 4-10: daily conversion volume climbs back toward normal. Resist every urge to adjust. Log what you see, touch nothing.
• Day 10-14: volume within striking distance of pre-outage levels for several consecutive days. Restore Target CPA about 10% above the old target.
• Weeks 3-4: walk the target down in 10-15% steps, several days apart, until you are back at your real number. Total cost of the episode: roughly three weeks of sub-optimal delivery. Without the exclusion and the staged restore, the same outage routinely costs six.

If you can backfill, here is how to do it right

Sometimes the conversions were not lost, only unreported: your backend knows exactly who bought during the outage. Google Ads accepts offline conversion uploads keyed on the click ID (GCLID) or on enhanced conversion identifiers, and a clean backfill genuinely shortens recovery because it replaces the false zeros with truth. The requirements that decide whether yours will be clean: you were capturing GCLIDs on your forms or orders during the outage (if the click IDs were never stored, this door is closed), the conversions fall within the action's click-through window, and you can upload the complete set, not a fragment. Upload half the missing conversions and you teach the algorithm your traffic converts at half its real rate, which is a subtler version of the same poison. Complete backfill plus a shortened exclusion window is the gold standard. No stored click IDs means exclusion only, and a note to start capturing them going forward.

What not to do

• Do not rebuild the campaign. Duplicating into a fresh campaign throws away all your clean history along with the poisoned window. The exclusion tool exists so you can keep the good and cut the bad.
• Do not import the missing conversions blindly. If you can reconstruct outage-period conversions (from your backend, with click IDs) and upload them, that can genuinely help. But partial backfills create their own distortion, and late-arriving conversions land outside the windows bidding cares most about. If you cannot backfill cleanly and completely, prefer the exclusion.
• Do not judge the account during recovery. Reporting for the outage and recovery weeks is contaminated in both directions. Mark the period in your notes, compare year-over-year or pre-outage instead, and resist drawing conclusions about creative or keywords from a window when the bidder was flying blind.

Prevention: make the next outage a non-event

While the memory is fresh, harden the account so the sequel is boring:

• Capture and store click IDs on every conversion path.A GCLID column in your orders or CRM is what makes clean backfills possible. It costs nothing to store and it converts a future outage from "lost data" to "delayed data."
• Run a redundant measurement thread. Enhanced conversions, or a server-side conversion upload alongside the tag, means one failure does not zero your signal. Redundancy is why some outages read as a 30% dip instead of a flatline, and a dip does not strangle Smart Bidding.
• Keep a change log.Most tag deaths trace to a deploy, a plugin update, or a consent change. When the dates of "what changed" are written down, diagnosis takes minutes instead of a week of archaeology.
• Alert on the leading indicator, which is the next section.

The real lesson: detection lag is the expensive part

Run the math on a typical version of this story. The tag breaks on day one. Nobody notices until day seven, because spend continued and the dashboard just looked "slow." Recovery, done well, takes another two weeks. The fix itself took an hour. The not-noticing consumed three weeks of budget, momentum, and trust. That ratio, one hour of fixing to three weeks of consequences, is almost entirely a detection problem.

The cheap insurance is an alert on the leading indicator: a campaign that normally records conversions daily recording zero for a full day. You can approximate this with a Google Ads automated rule that emails you on zero-conversion days, which is free and takes five minutes to set up. It will false-alarm on naturally slow days, and it only watches the number, not the tag behind it. Purpose-built monitoring, like Taglert, watches the conversion actions and the tracking status themselves across your accounts and emails you the day something stops recording, which turns this entire article into a one-day inconvenience instead of a three-week recovery. Either way: make sure something is watching, because the outage you catch on day one barely needs a playbook.