Seasonality adjustments in Google Ads: when they help and when they hurt

The adjustment is for events Google cannot know about

The seasonality adjustment is a signal to Smart Bidding that the conversion rate in a future time window will be markedly different from usual. According to Google it is meant for short events of one to seven days and only makes sense if you expect a large change, because Smart Bidding already accounts for seasonal events on its own. My rule of thumb follows from that: the adjustment is for what only you know. Your own 48-hour sale, a newsletter promotion to a large list, a voucher that runs once through a partner. For what sits in the calendar every year you rarely need it.

Technically, the adjustment is a multiplier on the expected conversion rate between 0.1 and 10.0, where 1.0 means no change (Google Ads API). In the interface you enter a percentage, so +30 percent instead of 1.3. Start and end apply in the account's time zone, and optionally only for certain devices. You create it under Tools, Budgets and bidding, Adjustments, Seasonality tab, for a whole campaign type or for selected campaigns, up to 2,000 per event (Google). It is available for Search, Shopping, Demand Gen and Display campaigns with target ROAS or target CPA, and for Performance Max with all bid strategies (Google, as of November 2026).

What the percentage really means

The number you enter is not a forecast for the strongest day. It applies equally to every auction in the window, so it has to match the average across the whole window. And it should only reflect the part of the uplift that Smart Bidding does not already expect from history.

Both points are easy to miss. Google's own example: if you expect a conversion rate 50 percent higher during a three-day sale, you enter up to +50 percent (Google). The "up to" is the part that counts.

Suppose a shop plans a four-day sale and, with the same number of clicks per day, expects these rises in conversion rate: day 1 +80 percent, day 2 +40 percent, day 3 +20 percent, day 4 +10 percent. The average is +37.5 percent. Anyone who takes the first day as the yardstick and enters +80 percent for all four days overbids clearly on three of the four days. If the promotion is clearly stronger on the first day, you can also split it into two windows, each with its own number.

How to estimate the uplift reliably

The usual practitioner method compares the conversion rate of last year's event with the period before it. GrowMyAds works through an example of 3.36 percent on Black Friday against 2.2 percent beforehand, so +52.7 percent, and expressly recommends not taking this number over in full. I would build the calculation like this:

A question about columns on steps 1 and 2. The standard columns in Google Ads attribute conversions to the click date, the "by conversion time" columns to the day of purchase (Google). For this estimate I consider the standard columns by click date correct, because the adjustment works on clicks in the window. By conversion time, purchases from earlier clicks count into the sale and inflate the raw value. Comparing both views is still worthwhile: if the gap is large, many customers buy with a delay, and the real effect on clicks in the window is smaller than the curve suggests.

Step 4 is the core. Smart Bidding knows recurring patterns. For Shopping, Google explicitly states that bids take seasonal trends at different times of year into account (Google). A rise that occurs in the same week every year is therefore already partly priced in. A rise from a promotion that you run for the first time this year, or on a new date, the system cannot foresee. That is where the adjustment belongs.

What an overestimate costs

An overestimate hits bids directly. The effect is the entered factor divided by the actual one. Optmyzr works it out: if you enter +50 percent and only +40 percent arrives, you bid about 7 percent above the value of the traffic (1.5 divided by 1.4).

What that means for contribution margin is shown by a worked example with round, assumed numbers. A shop has derived its target ROAS from margin and steers towards 500 percent. With 7 percent overbidding, the marginal ROAS of the most recently won clicks is about 467 percent (500 divided by 1.071). If break-even ROAS is 450 percent, that is just still profitable. If it is 480 percent, you lose money on those clicks. Anyone who enters +80 percent and gets +40 percent overbids by about 29 percent. The same shop then slips to about 389 percent.

From this I derive an asymmetry: an estimate that is too low costs you volume you would not have got without the adjustment anyway. One that is too high costs margin on every click in the window. When in doubt, I therefore set the number lower and size the haircut by how far the target ROAS sits above break-even. A shop with plenty of headroom can afford a bolder estimate than one that steers close to the limit.

Why Black Week is a borderline case

Black Week breaks both conditions at once. It often lasts longer than seven days, and Google warns that the adjustment may work less well over longer periods, such as more than 14 days in a row (Google). In addition, Smart Bidding knows the date from previous years.

An Optmyzr study of the Black Friday and Cyber Monday periods from 2022 to 2024, each with roughly 4,300 to 5,900 accounts, fits this. Accounts with an adjustment saw CPC rise about twice as much as accounts without one and lost 10 to 17 percent of ROAS, though they grew revenue more strongly. More volume, worse efficiency (Optmyzr). This is vendor data from a cohort comparison, not an experiment.

My assessment: for Black Week as a whole I leave the adjustment out and work through budget and target ROAS, as described in the article on Black Week preparation. It remains useful within Black Week for a single, own jump, such as an extra voucher for 24 hours that goes beyond what the shop did in previous years. Then with the number from step 4, that is, only the part that is new.

When to leave it alone

• The promotion runs longer than about two weeks. It is then not an event but a new level, and Smart Bidding adapts anyway.
• The uplift is small or uncertain. Google recommends the adjustment only for large expected changes.
• You have no comparable promotion in your data and would have to guess. A guessed value above the actual one costs margin.
• The problem is a measurement error, not a demand effect. For missing tags, website outages or import errors there is the data exclusion.

After the window you do not need to reverse anything. According to Google, campaigns return to their previous behaviour, and a negative counter-adjustment is not needed. Still check afterwards whether the number you entered was on target: conversion rate in the window by click date against your baseline, once the conversion delay has passed. The deviation is your best basis for the next estimate. How the bid strategy itself works and where its limits lie is covered in Smart Bidding for online shops.