Every year, between late October and mid-November, the same question comes up: "should we do something to Smart Bidding for Black Friday?". The honest answer is that it depends on how closely this year's Black Friday resembles last year's. And the question almost nobody asks, which matters more, is a different one: is the conversion data from the last few months reliable, or has the system been learning from tracking that broke at some point?
Those two questions are answered by two different tools, which sit side by side in the bid strategy section of the interface and are therefore confused all the time: seasonality adjustments and data exclusions.
Two tools, two directions in time
The most useful distinction is temporal. A seasonality adjustment looks forward: it tells Smart Bidding that in a future window the conversion rate will differ from what the system would expect from history. A data exclusion looks back: it tells the system to ignore a past interval because the data from those days is wrong.
| Seasonality adjustment | Data exclusion | |
|---|---|---|
| Direction | Future | Past (or a future window in which you already know tracking will be down) |
| When to use it | Promotion, flash sale, launch: the conversion rate will change sharply | Broken tracking, duplicated conversions, faulty offline uploads |
| What it does | Applies a multiplier to the expected conversion rate | Smart Bidding ignores the data from those dates |
| Typical length | 1-7 days | The length of the outage, plus conversion lag |
| Effect on reporting | None | None: the conversions stay visible |
In neither case does reporting change. Both act only on what Smart Bidding uses to estimate bids, which is exactly why a configuration mistake can go unnoticed for weeks.
Seasonality adjustments: what they actually do
Smart Bidding estimates the probability of conversion for each auction from recent data. That works well as long as the future looks like the recent past. When you know that for three days your conversion rate will rise by 50% because you are running 30% off the whole catalogue, the system cannot know in advance: it will notice once the event is under way and react with a lag that, on a 72-hour event, costs a lot.
A seasonality adjustment closes that lag. You set a window with start and end date and time, a scope (specific campaigns or campaign types, optionally narrowed by device) and an expected change in conversion rate. If you expect +50%, you enter an adjustment of up to +50%. Technically it is a multiplier: in the Google Ads API the permitted value runs from 0.1 to 10, where 1 means no change.
According to Google's documentation, adjustments are available for Search, Shopping and Display campaigns using Target CPA and Target ROAS, and for Performance Max with all bid strategies. Coverage evolves: before setting one up, check in the Help Centre that your campaign type is included.
The limit that matters: they are built for short events
Google is explicit: seasonality adjustments are ideal for events of 1-7 days and may work less well over longer periods, beyond 14 days. The reason is simple. Over a long window Smart Bidding adapts to the new conversion rate on its own within a few days: from then on your multiplier stacks on top of a system that has already corrected, and the result is inflated bids.
An adjustment is not a season. "December converts better" is not a use case: it is recurring seasonality, which the system sees in its history and handles by itself. An adjustment running from 1 to 24 December is the most common mistake we find in accounts in January, and it usually coincides with a December cost per acquisition that got away from everyone.
Data exclusions: for a broken past
Data exclusions do one thing: stop Smart Bidding learning from days on which the conversion data was wrong. Typical cases:
- The conversion tag disappeared after a site release and conversions dropped to zero for four days.
- A change to the cookie banner broke the consent signals and measured conversions halved while real sales did not move: a problem we covered in our piece on Consent Mode v2.
- A duplicated purchase event doubled conversions and value for a week.
- An offline conversion upload was missed, or run twice.
For the excluded interval, Smart Bidding ignores the data from those dates for the campaigns in scope. The conversions remain in reporting: the exclusion only changes what the system learns from.
Three details that make the difference
- Conversion lag. If there are three days on average between click and conversion, the outage also contaminated clicks from the days before it. Google recommends extending the exclusion backwards to cover that delay.
- Scope. An account-level exclusion touches every eligible campaign. If the fault only affected the lead conversion action and the ecommerce campaigns tracked normally, exclude only the campaigns involved.
- Recovery is not instant. After an exclusion on past dates, fluctuations usually settle within a few days; if a week or more of clicks was affected, they can last one or two conversion cycles.
Exclusions can also be set on future dates when you already know tracking will be interrupted: an ecommerce platform migration, a change of consent management platform, a risky release. It is one of the few cases where an exclusion is scheduled in advance.
Black Friday, Christmas, the sales: the Italian calendar
In 2026 Black Friday falls on Friday 27 November and Cyber Monday on Monday 30 November. But in the Italian market Black Friday is no longer a single day: many ecommerce sites turn it into a "Black Week" or a month of promotions, and that changes which tool you should use.
| Event | Typical length | Seasonality adjustment? |
|---|---|---|
| Single-day Black Friday, or the weekend through Cyber Monday | 1-4 days | Yes, if the discount is more aggressive than last year |
| Black Week | 7-10 days | Only on the peak days, not the whole week |
| "Black November" | 3-4 weeks | No: that is a season, Smart Bidding adapts |
| Last shipping days before Christmas | 2-4 days | Rarely: recurring demand, already in the history |
| Opening of the winter sales, the saldi (early January, dates set by each Region) | First 2-3 days | Yes, if you run online sales with deep catalogue-wide discounts |
| A flash sale with no precedent in previous years | 24-72 hours | Yes: this is exactly what the tool exists for |
How to estimate the percentage
The percentage should not be invented. Take it from a comparable event: same kind of discount, same depth, same product scope. If last year a 20% promotion moved conversion rate from 2% to 2.8%, the change was +40%. If this year's discount is identical, the system has already seen that peak in its history and an adjustment adds little. If the discount rises to 35%, a moderate adjustment on the peak days makes sense. Without a comparable event, err on the side of caution: an underestimated adjustment costs a few hours of responsiveness, an overestimated one costs margin across the entire window.
For events where you also want to change targets and budget, not just conversion rate, there is a more complete option: promotion mode, in beta on Search and Performance Max, schedules a ROAS tolerance and extra budget over a window and then lets them lapse on their own. Do not stack the two blindly: if you use promotion mode on an event, consider whether the seasonality adjustment is still needed.
Budget around the peaks
A seasonality adjustment makes bids more aggressive. If the campaign is budget-limited, the result is not more conversions: it is the same budget spent faster, often by mid-morning, with the campaign dark during the evening hours when Black Friday converts best. Before any adjustment, then, check impression share lost to budget.
- Raise the budget before, not during. Raising it at 11am on the day means you have already lost the first part of it. Update the budget the evening before.
- Remember the 2x headroom. The daily budget is an average: Google can spend up to twice it on a good day, within the monthly limit of 30.4 times the average daily budget. At peaks that headroom works for you, but plan for it. The details are in our Google Ads budget guide.
- Bring the budget down gradually. Do not cut everything by 70% on 1 December: delayed conversions from the event are still arriving, and the system needs a few days to return to normal.
- Think about stock. More aggressive bids on products that sell out in 48 hours are wasted budget. The feed needs the same care as the bids.
After the event: what to do (and what not to)
The week after Black Friday produces the most misleading reports of the year. Conversions from weekend clicks keep arriving for days, and the conversion rate in the first week of December often looks as if it has collapsed compared with the peak. That is normal, and in most cases the system should be left to work.
- Do not correct targets in the first 3-5 days. Wait for conversion lag to play out before judging.
- Do not use a data exclusion on the event days. That data is real: the conversion rate genuinely was higher. The seasonality adjustment exists precisely so Smart Bidding is not surprised by it; excluding those days afterwards takes correct information away from the system.
- Consider a negative adjustment only for an expected, documented dip. For instance, a hangover week after a very deep promotion that you have observed in previous years. Without a measured precedent, let the system handle it.
- Write down what happened. Percentage applied, actual conversion rate, windows. That is the data you will start from next year, and the only way to estimate the next percentage rather than guess it.
The most common mistakes
- Adjustments lasting weeks or months. A tool designed for 1-7 days stretched over all of November.
- Data exclusions for days that went "badly". A genuine drop in sales is not a measurement fault. Excluding it teaches the system a rosier reality than the real one.
- Exclusions that are too short. Covering only the days of the outage and forgetting conversion lag.
- The wrong scope. An account-level adjustment that also hits the lead generation campaigns, where the ecommerce promotion changes nothing.
- Double correction. A seasonality adjustment plus a manual tROAS cut plus a doubled budget on the same window: three pushes in the same direction and no way of knowing which one worked.
- Adjustments on budget-limited campaigns. Higher bids, same budget, campaign switched off by midday.
If you want to reach November with verified tracking, planned budgets and a reasoned decision on every event in the calendar, the free audit starts exactly there. You can see which services cover this ground, and the answers to the questions we get most often in the site’s FAQ.