AI + Paid Media
Smart Bidding Data Exclusions vs. Seasonality Adjustments: Use the Right Control
Updated: September 2026
Use a data exclusion when conversion data is wrong; use a seasonality adjustment when the data is accurate but you expect conversion rate to change sharply for a short event. The two controls solve different problems. Confusing them can teach Smart Bidding the wrong lesson or encourage it to chase a temporary spike.
The decision in one table
| Situation | Best control | Why |
|---|---|---|
| Tag stopped firing, website outage, or import failure | Data exclusion | Prevents corrupted conversion signals from influencing bids |
| Planned 3-day sale with a major expected conversion-rate lift | Seasonality adjustment | Prepares bidding for a temporary, forecastable change |
| Normal holiday demand | Usually neither | Smart Bidding already models recurring seasonality |
| Offer, pricing, or landing-page change with uncertain impact | Experiment and monitor | An adjustment would encode a guess as if it were known |
What a data exclusion actually does
Google describes data exclusions as an advanced control for conversion-data problems such as tagging errors, outages, and import issues. They change the data Smart Bidding uses; they do not erase the conversions from reporting. The exclusion applies to clicks that could have produced the affected conversions, so the time window must account for conversion delay.
That distinction matters. If a lead normally converts three days after the click, excluding only the outage hours may leave contaminated click cohorts in the bidding model. Document the incident start, repair time, affected devices, affected conversion actions, and typical lag before setting scope.
When seasonality adjustments are appropriate
Seasonality adjustments tell Smart Bidding that a future event should temporarily change conversion rate. Google says they are best for major, short-lived events—typically one to seven days—and may work less well beyond 14 days. They are not a general cure for weak performance.
Estimate the adjustment from comparable events, not optimism. If last year’s promotion lifted qualified conversion rate by 20%, a 50% input because the team “expects a big weekend” creates avoidable bidding risk.
A five-step intervention workflow
- Name the signal problem. Is the measurement wrong, or is customer behavior expected to change?
- Confirm scope. Identify campaigns, devices, conversion actions, dates, time zone, and lag.
- Bound the risk. Set acceptable budgets and temporary CPA or ROAS guardrails.
- Record the intervention. Save the reason, evidence, owner, and rollback date in change control.
- Review by conversion cycle. Avoid declaring success after one day when outcomes arrive later.
Pair this process with an AI paid-media rollback plan. Also read how to use Google AI performance diagnostics without chasing a platform score.
Common mistakes
Excluding real bad performance
A weak landing page, worse offer, or lower-quality traffic is not corrupted data. Excluding it hides evidence the bidding system needs.
Using seasonality for recurring patterns
Smart Bidding is designed to account for common seasonal behavior. Use the adjustment only when the expected shift is unusually large and short.
Backfilling without a plan
Google cautions that backfilled conversion data can affect bidding. Separate reporting repair from model intervention and follow the platform’s timing guidance.
Frequently asked questions
Do data exclusions remove conversions from reports?
No. Google states that exclusions affect the data used by Smart Bidding, while the conversions remain visible in reporting.
Can I use both controls?
Yes, if a tracking incident and a planned short event truly overlap, but scope each control separately and document the rationale.
How quickly should performance stabilize?
Google says past-date exclusions may begin stabilizing after a few days; larger affected windows can take one or two conversion cycles.