Seasonality Adjustments in Google Ads 2026: When They Help and When They Break Smart Bidding
A three-day flash sale, a conversion tag that silently broke over the weekend, and a Q4 demand spike are three different problems. Google Ads has a separate control for each of them, and advertisers routinely reach for the wrong one — applying seasonality adjustments where a data exclusion belongs, manually inflating budgets where a seasonal budget adjustment exists, and feeding Smart Bidding distorted signals in the most expensive quarter of the year.
This is a working breakdown of all three controls: what each one actually does to the model, the hard limits, how to calculate the percentage from your own history instead of guessing, and a calendar for getting an account ready for Q4. Plus a protocol for the worst case — tracking breaking mid-promo.
Three controls people constantly mix up
They live in different parts of the interface and solve different problems. One tells the algorithm to expect a different conversion rate. One tells it to ignore a period entirely. One temporarily hands the campaign more money.
| Control | What it does | When to use it | Where it lives |
|---|---|---|---|
| Seasonality adjustments | Tells Smart Bidding to expect a conversion rate change on specific dates | Short promotions, launches, offer changes | Tools → Shared library → Bid strategies → Advanced controls |
| Data exclusions | Removes a date range from the bidding model’s training data | Broken tracking, site outage, payment gateway failure | Same place, adjacent tab |
| Seasonal budget adjustments | Temporarily raises the daily budget and reverts it automatically | Time-boxed demand spikes where budget is the bottleneck | Campaigns → bulk actions / campaign view |
The distinction matters: a bid-side adjustment does not create budget, and a budget adjustment does not change bidding logic. Most promotions need both — and almost none need the thing people do most often, which is yanking the target CPA down the day before launch.
Seasonality adjustments: mechanics and hard limits
The control does exactly one thing. For a date range you specify, you tell the system that the expected conversion rate will be X% higher or lower than normal. The model bakes that into its forecast before it has observed any actual data from the period — which is the whole point, because otherwise the first hours of a promo are spent learning.
Where it works
- Search, Shopping and Display — only with Target CPA and Target ROAS bid strategies.
- Performance Max and App campaigns — with all bid strategies (App support arrived later and is documented as beta).
- Travel campaigns — not supported.
- Scope is manual: the whole account, specific campaigns, campaign types and devices.
Duration
Google’s own guidance: the tool is built for short events of 1–7 days and works less well beyond 14 days at a time. The reason is mechanical — over two weeks the algorithm observes the new conversion rate in real data anyway, and your adjustment starts double-counting an effect the model already sees.
The practical rule: Black Friday and Cyber Monday are two events with two different percentages, not one ten-day adjustment. And “the holiday season” as a whole is not a use case at all — that is market seasonality, which Smart Bidding handles from historical data without being told.
Calculating the percentage instead of inventing it
The number is a relative change in conversion rate, not the size of your discount. Derive it from a comparable past promotion:
Adjustment % = (CR during the last comparable promo ÷ CR over the two calm weeks before it − 1) × 100, using the same campaigns and the same geos on both sides.
Example: baseline CR is 3.2%, and the last three-day promo with the same discount mechanic ran at 4.6%. The adjustment is (4.6 ÷ 3.2 − 1) × 100 ≈ +44%. With no comparable history, take the low end of your expectation. Underestimating is cheaper than overestimating, because an inflated percentage makes the system pay click prices your conversions will not repay.
| Situation | Indicative adjustment | Note |
|---|---|---|
| 10–15% discount, landing page unchanged | 0…+15% | Often needs no adjustment at all |
| 30–50% off, on-site banner, email push | +30…+60% | Size from the last comparable event |
| Sitewide sale with free shipping and a countdown | +40…+80% | Verify against actuals on day two |
| Planned demand drop (stock-take, paused fulfilment) | −30…−70% | Negative adjustments are equally valid |
Treat those ranges as a starting point for the first calculation, not as benchmarks. The only honest source of the number is your own history — and if pulling that history is painful, that is a separate problem solved in the piece on exporting Google Ads data to BigQuery.
After the event
You do not need to apply a mirrored negative adjustment when the promo ends — the system reverts on its own once the date range closes. The one thing worth doing is leaving the target CPA or ROAS untouched for another 7–10 days so the model can digest the spike.
Data exclusions: when your own numbers lie
A data exclusion says: tracking was broken on these dates, do not learn from them. It does not delete conversions from reporting — the interface still shows the same numbers. Only the training data behind Smart Bidding changes.
Typical triggers
- A site release wiped or overwrote the conversion tag.
- Order value stopped being passed, so Target ROAS trained on zeros.
- The site was down or the payment gateway rejected transactions — clicks continued, conversions did not.
- A server-side container failed and part of the event stream never arrived. What breaks there and how to catch it early is covered in the server-side tagging with sGTM guide.
- Duplicate conversions from a double-fired tag — the inverse case, where the data is inflated.
How far back to exclude
The common mistake is excluding exactly the days the tag was silent. Conversions arrive with a lag — typically anywhere from a few hours to seven days depending on the sales cycle — so the hole in the data starts earlier than you noticed it: clicks that happened before the break were supposed to convert during it.
Rule: exclusion start date = the day the break happened, minus your typical conversion lag. Read the lag from the time-to-conversion report; if you do not know it, start with three days and refine later.
What exclusions do not do
- They do not recover lost conversions — that is what offline conversion import and backdated uploads are for.
- They do not fix reporting — the period still looks terrible in the interface.
- They do not help if the outage lasted a month. Cutting a third of the training window leaves the strategy with almost nothing to learn from; at that point it is more honest to accept the dip and let the model relearn.
One more thing: exclusions are an incident tool, not routine hygiene. If you apply them monthly, the problem is not Google — it is the stability of your measurement. Start with an account audit checklist and verify that the GCLID actually survives the trip into your CRM.
Seasonal budget adjustments: the control almost nobody uses
The third tool answers a different question — not “how should bids behave” but “where does the money for this week come from”. You set a temporary budget for a date range, and the campaign returns to its original amount on its own afterwards.
- Available for Search and Shopping campaigns.
- Duration is 3 to 14 days, starting today or on a future date.
- Consecutive adjustments on the same campaign require a 7-day gap between them.
- Not available for campaigns on shared budgets, drafts, campaigns with ad scheduling, flighted campaigns, or campaigns that already have an adjustment applied or scheduled.
- It affects monthly spend limits exactly like an ordinary budget change.
- If you also change the budget manually during the window, the campaign reverts to a value that accounts for that manual edit — a reliable source of surprises.
The practical point: in Q4, budget usually becomes the constraint before bids do. Before touching conversion-rate percentages, look at lost impression share due to budget — above roughly 20–30% you have a budget problem, not a bidding problem. How to read that and plan ahead is covered in impression share and auction insights and budget planning with Performance Planner.
A Q4 2026 preparation calendar
Seasonality work does not start the day before the promo. This sequence can be taken as is.
| Timing | What you do |
|---|---|
| 8 weeks out | Pull last season’s data: CR, CPA and ROAS by week and campaign. Record the conversion lag. |
| 6 weeks out | Calculate the break-even CPA against seasonal margin. Verify conversion values are set correctly. |
| 4 weeks out | Run any structural experiments now or never. Peak season is not the time to restructure. |
| 3 weeks out | Test the full measurement chain: tag, server container, value passing, duplicates. Last calm window to fix things. |
| 2 weeks out | Freeze strategy edits. Create seasonal budget adjustments and put promo dates in the calendar. |
| 3–5 days out | Create seasonality adjustments on the exact promo dates with the calculated percentage. |
| Promo day | Do not touch targets. Watch two things only: budget-limited impression share and whether conversions are actually recording. |
| After | 7–10 days of no changes, then a post-mortem: actual CR versus the percentage you assumed. |
On the freeze: any change to target CPA or ROAS during the promo blends with the effect of the promo itself, and afterwards you cannot separate the two. If you genuinely want to test something, use the proper instrument covered in the piece on Google Ads experiments.
How this interacts with the August 2026 target bidding change
Since mid-August 2026, budget-constrained campaigns on Target CPA and Target ROAS no longer systematically overshoot the target and now track closer to the number you set. For Q4 that changes tactics: the old habit of setting an aggressive target and expecting the system to “find volume anyway” no longer applies. The change and a diagnostic sequence are covered in the breakdown of target-based bid strategy changes.
The takeaway: in peak season you buy volume with budget and with a conversion-rate signal, not by cutting the target. Cutting the target during a peak is the fastest way to leave the auction exactly when traffic is worth the most. The full peak-season playbook sits in the Black Friday and promo peaks guide.
Seven expensive mistakes
- One adjustment for the whole season. The tool is built for 1–7 days; a month-long percentage double-counts what the model already learned.
- A percentage pulled from thin air. +100% because “the sale is big” rather than because last year’s data said so. The result is overheated bids and a budget gone in hours.
- Adjusting bids when the problem is budget. If the campaign is capped, a conversion-rate signal changes nothing — there is no more money.
- Data exclusions as routine. Frequent exclusions eat the training window. If tracking breaks often, fix tracking.
- Excluding too narrow a window. Ignore the conversion lag and the contaminated days end up in training anyway.
- Manual edits on top of a seasonal budget adjustment. The campaign reverts to a value that includes your manual change, not to the original.
- Restructuring during the peak. New campaigns start with no history and learn on the most expensive traffic of the year. Do structural work before the window, and do bulk edits through Google Ads Editor.
Protocol: conversions break mid-promo
- Hour 0. Confirm it. Compare conversions in the interface against orders in the CRM for the last six hours. A gap above 20% is an incident, not noise.
- Hour 1. If the break is on your side and fixable quickly, fix it and leave campaigns alone. Smart Bidding survives a few hours.
- If the fix runs past 24 hours. Move affected campaigns to a strategy that does not depend on conversions (for example, a click-cost cap) and note the switch date.
- Once fixed. Apply a data exclusion covering the outage plus your conversion lag before it.
- Recover what you can. Upload the missing conversions via offline import if the click identifiers survived.
- Return. Go back to a target-based strategy only after 7–14 days of clean data, and step the target back in 10–15% increments.
What to verify before the season starts
- Every key conversion action carries a real value rather than a zero — otherwise Target ROAS optimises toward nothing. Covered in conversion value and value rules.
- You know the typical conversion lag for each primary action.
- Budget-limited impression share is measured and written down before the season, so there is a baseline to compare against.
- Last season’s data is exported somewhere usable, not “we’ll look it up in the interface”.
- Access rights are settled: who is allowed to change budgets during the peak. That sits in the guide to running a Google Ads manager account.
If you are going into the season as a team across several accounts, rehearse the process first: who decides on target changes, who owns measurement, who holds budget access. PPC Rebels covers this on real account structures in its Google Ads services and training, and the underlying setup patterns are collected in the agency account guide.
The core idea: a seasonality adjustment is not a throttle. It is a way to tell the algorithm something it cannot know in advance. Anything it can already observe in the data does not need to be announced — and trying to “help” where no help is needed is what usually breaks a season.
FAQ: seasonality adjustments, data exclusions and budgets
Do I need seasonality adjustments for normal holiday seasonality?
No. Smart Bidding handles market seasonality from historical data. The tool exists for events the system cannot know about: your own promotion, an offer change, a product launch, a one-off partnership.
What is the maximum duration for a seasonality adjustment?
You can technically set a longer range, but Google’s guidance is that the tool is designed for 1–7 day events and works less well beyond 14 days at a time. Split long promotions into separate events with separate percentages.
Do adjustments work with Performance Max?
Yes — Performance Max supports them across all bid strategies. Search, Shopping and Display support them only on Target CPA and Target ROAS. Travel campaigns are not supported.
What happens if I set +100% “just in case”?
The system assumes double the conversion rate and bids accordingly. If the actual lift is smaller, you simply overpay for clicks on the most expensive days of the year. Overstating is riskier than understating.
Should I apply a negative adjustment after the promo?
No. The system reverts automatically when the date range ends. A mirrored negative adjustment afterwards is a common and harmful improvisation.
Will a data exclusion remove conversions from my reports?
No. Reporting is unchanged — the numbers for the period stay visible. Only the use of that period in bid strategy training is affected.
How far back should the exclusion window start?
By your typical conversion lag. If conversions usually land within three days of the click, start the exclusion three days before the break. Read the lag from the time-to-conversion report.
Can I exclude a whole month of downtime?
Technically yes, practically no. You would remove a large share of the training window and leave the strategy with almost no data. For long outages it is better to step off target-based bidding temporarily and return once clean data accumulates.
Why use a seasonal budget adjustment instead of raising the budget manually?
Because it reverts on its own when the window closes. That removes the classic failure mode of a forgotten inflated budget burning money for weeks after the sale ended.
Why is the seasonal budget adjustment unavailable on my campaign?
Usually because of a shared budget, ad scheduling, draft status, or an adjustment already applied or scheduled. The feature is also limited to Search and Shopping.
Can I combine bid and budget adjustments?
Yes, and for most promotions that is the correct pairing: the budget adjustment removes the money constraint while the seasonality adjustment tells the algorithm what to expect from the auction. Just do not change the target at the same time, or the result becomes uninterpretable.
How do I know whether the adjustment worked?
Compare the actual conversion rate during the promo against the percentage you assumed, and check whether CPA stayed inside your break-even threshold. If actuals diverged sharply in either direction, record the real figure and use it as the baseline next season.
