Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Google Ads & PPC › Bidding & Budgets

Bidding Through Sales and Tracking Outages

How to keep Smart Bidding on course through sales, launches and tracking outages: seasonality adjustments, data exclusions, budgets and changes to hold back.
Guideadvanced10 min readUpdated

Why unusual days mislead automated bidding

Smart Bidding sets a bid for each auction from a prediction of whether it will convert and what the conversion will be worth, and the prediction rests on what the campaign has recorded recently. The glossary's seasonality entry puts the consequence plainly: Smart Bidding learns from recent conversion rates, so a short, sharp event can catch it out. Over a long stretch the recent past is a good guide to the next auction. Over a weekend sale, a launch day or a broken tag, it is not.

Unusual periods come in two kinds, and they mislead the model in opposite ways.

PeriodWhat actually changesWhat the model seesRiskMain tool
Sale or promotionThe real CVR rises for a few daysA sudden rise it has no history forUnder-bidding at the start; afterwards, expecting the sale rate to continueSeasonality adjustment, with a raised budget
Launch with a known spikeDemand and conversion rate jump on known daysThe same as a saleAs aboveSeasonality adjustment if the spike is short and predictable
Long seasonConversion rate drifts up or down over weeksA gradual change it can learn fromSmall, if budgets keep paceA budget plan; usually no adjustment
Tracking outageNothing about demand; conversions stop being recordedA collapse in conversion rateBids cut on auctions that were convertingData exclusion where documented; hold everything else
Inflated trackingNothing about demand; conversions counted twiceA false rise in conversion rateOver-bidding on inflated signalsFix the tag; treat the window like an outage

The Smart Bidding strategies guide adds the warning that governs everything below: changing targets during a period that is already unusual is a mistake, because the model will learn the unusual period as normal.

Planned peaks: sales and launches

A planned peak has one great advantage over an outage: you know it is coming. The work happens before it starts. The sale promotion readiness checklist puts the bidding decisions on the list for the days before prices change, beside the offer, the feed and the site: budgets and pacing for the peak days, and whether Smart Bidding needs a seasonality adjustment. The checklist notes that a daily budget sized for a normal day can run out by midday on the main sale day, and that a strategy that does not know a short sale is coming may misread the jump in conversion rate.

Three questions decide the plan. How long is the event? A few days of sharply higher conversion rate is a different problem from a season that builds over weeks. How big is the change? Past events of the same kind are the evidence, read as the conversion rate during the event against the weeks around it. And does the change reverse? A sale ends on a known date, and the conversion rate falls back; a launch that adds a product to the range may not reverse at all, and a lasting change is something the model should learn rather than be told to ignore.

Long seasons need different handling from short events. A period that builds gradually, such as the weeks of rising demand before a holiday, gives the strategy time to learn from its own recent data, and in practice it needs a budget that keeps pace rather than an adjustment.

Seasonality adjustments: what they do and when to use one

Google documents seasonality adjustments for short, predictable events. The budgets and pacing guide describes what they do in outline: they tell the model to expect a change in conversion rate for the period, and they change bids, not budget, so both need setting. The Google Ads setup for online sales adds why they are the right tool for a sale: they are safer than raising targets by hand and forgetting to lower them.

Use one when the event is short, its dates are known, and you can estimate the change in conversion rate from your own history. Do not use one for a long season the strategy can learn from, for a lasting change such as a new product line, or for an event whose effect you cannot estimate, because an adjustment that expects the wrong change misleads the model as much as no adjustment at all.

The help centre sets out how an adjustment is entered, which campaigns and strategies it applies to and how long it can run, and those details change; this guide does not repeat them, and the current help page is the reference before setting one. The estimate itself is yours. Take the conversion rate during the last comparable event and the rate in the weeks around it, and expect the same direction of change, adjusted for anything different this time, such as a deeper discount or a shorter window. For a value-based campaign, check what the expected change does to return on spend with the ROAS calculator, because a sale can raise conversion rate while lowering order value.

Budgets around a peak

A seasonality adjustment leaves the budget where it was, so a campaign told to bid harder for a sale can hit its budget early on the busiest day. The budgets guide sets out the mechanics. The campaign budget is an average daily amount that Google may exceed on a strong day, up to the documented daily ceiling, and must respect across the month through a monthly limit that is recalculated whenever the daily budget changes.

The practical consequences for a peak follow from that. Raise the budget before the event, not after the limited-by-budget status appears. To spend heavily in a short window, set the daily budget for the window as the amount wanted per day, then lower it afterwards, remembering that the monthly limit recalculates with each change. Check month-to-date spend against the plan during the event, because Budget Pacing in a peak week decides how much of the month is left. Under a target strategy, a budget increase only matters if the campaign was limited by budget, which is exactly what a sale can cause. The budgets guide's seasonal peak walk-through for the budget calculator plans the total on peak costs rather than the yearly average.

Tracking outages and data exclusions

A tracking outage is the reverse of a sale: demand carries on, but conversions stop being recorded, so the model sees a collapse in conversion rate that never happened. Left alone, it bids down on auctions that were converting, and the weeks after the fix inherit a model that learned from a false signal. A tag that fires twice does the opposite, inflating conversions and pushing bids up.

The tracking outage response SOP runs the incident: triage, assess the impact on bidding, protect campaigns, fix, verify recovery, annotate and report. Its protective step names the bidding options. Hold budgets and targets steady, avoid structural edits, and, where the platform documents it, use a data exclusion for the outage window; pausing is a last resort for campaigns clearly overspending on bad signals. Google documents data exclusions for some bidding strategies as a way to keep a period of bad conversion data out of what the strategy learns. Which strategies they cover, how the window is set and how long it may be are in the help centre, not here.

Two habits make the response faster. Detect the outage early, by comparing recorded conversions with the business's own orders or leads every day; the diagnosis for conversions not tracking walks the chain from the site to the conversion action once a gap is confirmed. And record the window precisely, with start and end times, because every later decision, from the exclusion to the annotation in reports, depends on it.

Changes to hold back during an unusual period

The Smart Bidding guide lists the changes Google treats as triggers for a learning period: changing the strategy type, changing the conversion actions, and large changes to targets or budgets. During a sale, a launch or an outage, each of those adds a second disturbance to one the model is already absorbing, and makes the period harder to read afterwards.

ChangeDuring a planned peakDuring a tracking outage
Target CPA or ROASHold; use a seasonality adjustment insteadHold; the model is already misreading
Bid strategy switchHold until well after the eventHold
Conversion actions or goalsHold, unless fixing the outage requires itChange only what the fix needs
BudgetRaise for the window as planned, then lowerHold
Structure, match types, new campaignsHoldHold
Negatives and ad copy for the offerAllowed; they are not among the learning triggers the Smart Bidding guide listsAllowed if needed

Plan sale budgets and adjustments in advance so the only edits during the event are the ones written into the plan. If results fall after a strategy change made near an event, the results dropped after a bid strategy change diagnosis separates learning, tight targets, budget limits and conversion delay. The academy exercise on bidding through a spring sale and a tag outage for a garden building maker practises writing that plan, with the bidding, budget and frozen changes for the sale and the response to the day the purchase tag stopped firing.

Reading results afterwards

The days after an unusual period are the most misread in the year. Conversion Lag means a sale's conversions keep arriving after it ends, credited back to the clicks, so the last days of the event look weak at first and improve. Read the event only once the delayed conversions have come in. The academy checkpoint on bidding through thin data, sales, outages and scaling spend tests these readings beside the seasonality adjustments, data exclusions and changes to hold back during an unusual period.

Compare against a fair baseline: the same weekdays without a promotion, or the same event last year, and at account level rather than campaign by campaign, since budget and bids move between campaigns during a peak. Watch the weeks after a sale too, because in practice a strategy that learned the sale's conversion rate can overbid when it ends, and a sale can pull orders forward from the following weeks. Whether the sale itself paid, net of the discount and the orders it borrowed, is a separate question that the guide to measuring a sale or promotion answers; this page covers the bidding side. After an outage, read the recovered weeks against the period before it, leave the excluded window out of comparisons, and confirm that recorded conversions are back in their usual relationship to orders. The bidding hub sets out the targets those readings are judged against.

A sale weekend and a broken tag (illustrative)

Illustrative example — not a real result

An outerwear shop through a sale and an outage

Raincobble Outerwear, an invented outerwear shop, runs a three-day weekend sale with 30% off waterproofs, on a target ROAS of 450% and a £400 daily budget. In last year's equivalent sale, conversion rate rose from 2.1% to about 3.4% for the sale days and fell back on the Monday. A month later a checkout update stops the purchase tag firing for a day.

PeriodEventActionShop ordersConversions recorded
Week before saleNormal tradingSeasonality adjustment set for the three sale days; budget plan agreed410398
Sale Friday to SundayConversion rate rises to 3.5%Daily budget raised to £900 for the three days; target left unchanged520507
Week after saleRate back to 2.0%Budget back to £400; no target change360349
Outage day, a month laterPurchase tag silent after a checkout releaseData exclusion applied for the window; budgets and targets held643
Week after the fixTracking restored and verified against ordersWindow annotated in reports405394

The sale weekend spends £2,610 against a planned £2,700 and delivers a ROAS of 470% once the delayed conversions arrive. The week after the sale shows no overbidding, because the adjustment covered only the sale days. The outage is caught the next morning by the daily check of orders against recorded conversions; with the window excluded and nothing else changed, the following week's ROAS is in line with the month before.

Common mistakes

  • Raising the target for a sale by hand. It starts learning and is easily left at the sale level afterwards.
  • Setting an adjustment and not the budget. The model bids harder and the budget runs out by midday.
  • Using an adjustment for a long season. The strategy can learn gradual change from its own data.
  • Guessing the expected change. The estimate comes from past events in your own account.
  • Changing targets or strategy during an outage. Two disturbances, one unreadable result.
  • Finding an outage days later. A daily check of orders against recorded conversions catches it in hours.
  • Judging the sale on its last days. Conversion lag fills them in; wait before reading.

Frequently asked questions

Should I raise my target ROAS or target CPA for a sale instead of using a seasonality adjustment?

Usually not. A target changed by hand for a short event has to be changed back afterwards, each large target change can start a learning period, and a target left at its sale level after the sale ends holds the campaign to a standard it can no longer meet. Google documents seasonality adjustments for short, predictable changes in conversion rate, and the site's online sales setup treats them as safer than raising targets by hand and forgetting to lower them.

What should I do in Google Ads when conversion tracking breaks?

Fix the tracking first, and meanwhile stop automated bidding learning from the gap. Hold budgets and targets steady, avoid structural edits, and where Google documents a data exclusion for the campaign's strategy, apply one for the outage window. Pausing is a last resort for campaigns clearly overspending on bad signals. Record every action with its time, and annotate the window in reports once tracking is restored.

Do I need a seasonality adjustment for Black Friday or Christmas?

It depends on the length and size of the change. A sharp jump in conversion rate over a few days is the case Google documents the adjustment for. In practice, a season that builds over weeks is something the strategy can learn from its own recent data, so it usually needs a raised budget rather than an adjustment. Decide from your own history: how much conversion rate moved during the last comparable event, and for how long.

Terms used on this page

Tools to try

Where this page is used

Part of these learning paths

Also related

Sources

  1. About Smart Bidding — Google Ads Help
  2. About average daily budgets — Google Ads Help

Related