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

Meta Ads › Optimisation & Scaling

Planning Meta Ads Around a Sale or Peak

How to plan Meta Ads around a sale or seasonal peak: launch early to clear learning, ready creative and audiences, budget the window and read results after.
Guideintermediate11 min readUpdated

Why a peak behaves differently on Meta

Three things change at once in a peak, and each argues for doing the work early.

The auction gets dearer. The troubleshooting guide for performance drops lists retail seasons, major events and competitor launches as auction shifts that raise cost per thousand impressions without any change in the account, and recommends noting cost per thousand by week across the year so that a seasonal rise is recognised rather than investigated. The account's own weekly figures from last year's peak are the reference for this year's; a published average for someone else's auction is not.

Learning is less affordable. The learning phase guide, citing Meta's documentation on the learning phase, describes the period after an ad set is created or significantly edited as one of less stable delivery and usually higher cost per result, and lists the edits that restart it. A new ad set launched on the first day of a sale spends the most expensive days of the year learning.

Reporting runs behind. Meta's documentation on attribution settings, as the attribution guide explains it, reports conversions against the date of the ad interaction, so the most recent days of any period keep filling in while the window is open. During a short sale, the days being judged are always the least complete.

Seasonal peaks and planned sales share these features, which is why this guide treats them together. The decision about which promotion to run and at what discount is made before it reaches Meta, as the guide to planning promotions and discounts describes; this guide runs the Meta side of a promotion already decided, and the sale promotion readiness checklist is the channel-neutral gate that confirms the offer, the site, stock and measurement are ready. Where the whole store gears up for the peak, the peak season trading playbook runs this Meta work alongside the other teams' parts, with a daily stand-up and one trading lead deciding anything outside the plan.

Weeks before: launch early and let learning settle

Build the peak structure in the run-up and let it settle on evergreen creative. Meta does not document a lead time, and none applies to every account, because an ad set's time in learning depends on how quickly it accumulates its optimisation event. The practical rule is to launch early enough that the ad sets you will rely on during the window have exited learning and delivered stably for a while before it opens, judged from how long your ad sets usually take.

Use the run-up to check everything that a peak would expose. Confirm the purchase event is arriving from browser and server with values, and that deduplication is working; the monthly reconciliation is the quickest proof. Keep the structure consolidated: the learning phase guide's advice that fewer ad sets with more events exit learning sooner matters most when every day is expensive. Decide the attribution setting now and leave it alone, since changing it mid-flight changes both what is counted and what delivery learns from. And plan the Flighting of spend across the run-up, the window and the days after in the media plan, using the media planning framework to set each phase's job.

Avoid structural changes in the final days before the window. The learning phase guide lists changes to targeting, the optimisation event, the bid strategy and the creative of existing ads among the significant edits; any of them close to the peak puts an ad set back into learning when it can least afford it.

A product launch has the same timing problem with no sales history behind it. The page on paid social for a product launch covers the waitlist audience and the stand-in optimisation event that a peak for an existing range does not need.

Creative, audiences and the catalog ready in advance

Creative. Make the sale versions of proven concepts before the window: the same idea with the offer, the dates and the urgency added. The learning phase guide lists adding a new ad to an ad set among the documented significant edits, often mild in practice, and treats adding creative to a stable ad set as the preferred way to refresh it. Add all the sale ads to each ad set in one edit when the sale starts rather than in several, and keep the evergreen ads running beside them until the new ones have results. Check every sale claim against the readiness checklist, since price claims are where promotions get into trouble.

Audiences. Retargeting is only as good as the pools behind it, and those are built from events within a retention window, as the custom audiences guide explains. Create the website audiences the sale will use, such as recent product viewers and people who added to cart, before the run-up, so that the run-up's prospecting and teaser content fills them. Refresh customer lists used for exclusions or for a customer-only early access, following the tutorial on uploading a customer list, and keep recent purchasers excluded from prospecting so that the peak budget is not spent on people who have just bought.

Catalog. The catalog ads guide notes that the feed should carry a sale price alongside the price while a promotion runs, and that overlays showing a strikethrough price or a percentage discount only work when those underlying fields are populated. Set the sale prices, check that the feed refreshes often enough to keep prices and stock in step with the site, and build a product set for the sale items using the catalog's labels. If items vanish from delivery, the diagnosis page for catalog products not showing covers suppression for stock and price mismatches as well as feed refresh.

Budgets for the window: daily, lifetime and limits

Size the budget from a peak-period cost per purchase rather than the annual average, which the budget calculator walkthrough for seasonal peaks works through with a one-month window and a small reserve. Then choose how to hand it to Meta. The comparison of campaign and ad set budgets, citing Meta's documentation on Advantage+ campaign budget, sets out the mechanics.

OptionHow it behaves, as the site's guides state itWhen it suits a peak
Daily budgetAn average; the system may spend more on days with better opportunities and less on others, within documented weekly limitsAlways-on campaigns running through the peak, adjusted in steps
Lifetime budgetA total paced across the schedule; required for ad schedulingA dedicated sale campaign with fixed start and end dates
Advantage+ campaign budgetOne figure moved between ad sets towards the best resultsConsolidated prospecting where ad sets share the purchase goal
Ad set budgetsEach ad set guaranteed its own figureRetargeting kept on its own budget so it cannot soak up the peak spend
Campaign spending limitA cap that stops delivery when reachedA hard ceiling for the sale's total, which a budget is not

Daily budgets already lean into the peak, because the page on spending over the daily budget explains that a single day above the daily figure is documented flexibility balanced over the week, not a fault. Raising budgets is still usually needed, and the learning phase guide lists a large budget change among the edits that can restart learning, so raise in steps during the run-up rather than in one jump on the first sale day. A lifetime budget avoids the daily edits altogether, at the cost of building a separate campaign that has to be launched early enough to settle. Under either, keep retargeting on its own budget: the budget types guide warns that a shared budget lets a small pool absorb spend until frequency climbs and the pool is exhausted.

During the peak: what to touch and what to leave

The optimisation hub sets a daily check that spends nothing on changes and looks only for breakages, and a peak is when that discipline pays most.

TouchLeave
Ads that are rejected, broken or pointing at sold-out productsTargeting, optimisation event and bid strategy on working ad sets
Budget, in steps, where results are consistently strong across complete daysBudgets reacting to a single morning's figures
New creative added alongside a clearly tired adEdits to the creative of ads that are running
The end date, extended if the sale is extendedThe attribution setting
Catalog prices and stock, kept in step with the siteAd sets in learning, other than for breakages

Watch frequency on retargeting closely. The frequency and fatigue guide notes that small audiences with large budgets repeat fastest, and a peak budget on a retargeting pool is exactly that combination. Read every figure knowing the most recent days are incomplete, so a weak-looking morning is often an open window rather than a falling campaign.

When something breaks mid-peak

Breakages cost most during a peak, and the reflex edit costs nearly as much. Use the order of checks from the troubleshooting guide for performance drops, quickly:

  • Is it real? Recent days are incomplete; check that the purchase event is still arriving in Events Manager and that no one changed the attribution setting or the reporting view.
  • Was something edited? Change history for the ad sets and campaign, including edits by colleagues and automated rules.
  • Did the auction move? Cost per thousand against the same days last year; if it rose and the other links held, this is the season, not a fault.
  • Is an ad tired? Frequency rising and click-through falling against the ad's own first week, most often in retargeting.
  • Is an ad rejected, or did spend stop? Ad status, the account quality page, payment and spending limits.
  • Did the site change? Sale banners, new scripts and checkout changes are released just before peaks and break tags as well as pages. Load the page on a phone and walk the checkout.

If the site is broken, pause only the affected ads until it is fixed. If spend must fall, lower budgets rather than pausing ad sets you will need for the rest of the window, since the learning phase guide notes that an extended pause can restart learning. Fix the cause, then leave the campaigns alone while delivery recovers.

After the peak: reading results as the window fills

Do not judge the peak on the morning after. Purchases from people who clicked or viewed during the sale keep arriving for as long as the attribution window is open, and Meta credits them to the date of the ad interaction, so the sale's final days are the last to fill in. That delay between interaction and purchase is Conversion Lag, and it is longest for considered purchases. Read the sale once the window has closed on its final day, and compare with the figures read at the time to see how much they moved.

Then reconcile and judge. The tutorial on reconciling Meta results with your orders turns Meta's figures into ones the business accepts. The ROAS calculator walkthrough for discount promotions reads return against the sale's own break-even, which is higher than usual because the discount cuts the margin. And the guide to measuring a sale or promotion asks the questions Meta cannot answer: how many orders were incremental against a baseline, how much demand was pulled forward from the weeks either side, and how much margin was given away. Wind budgets down in steps afterwards, lower rather than pause what will return, and record the weekly cost per thousand for next year's plan.

An outerwear shop's winter sale (illustrative)

Illustrative example — not a real result

Frostbramble Outerwear plans and re-reads its winter sale

Frostbramble Outerwear, an invented outerwear retailer, ran a two-week winter sale at 30% off. Its normal break-even ROAS was 2.0; with the discount, the sale's own break-even was 3.5. The prospecting ad set launched three weeks before the sale on evergreen creative, budgets rose in steps through the run-up, and the sale ads were added to each ad set in one edit on the first morning. Retargeting kept its own budget.

WeekDaily budgetCost per thousandROAS read at the end of the weekROAS re-read a fortnight after the sale
Three weeks before£300£7.203.13.2
Two weeks before£420£7.603.03.1
One week before£550£8.902.62.8
Sale week one£650£12.403.03.9
Sale week two£650£11.802.73.6
Week after£400£9.102.12.5

Cost per thousand in the sale weeks was roughly two-thirds higher than in the first run-up week, close to the rise in the same weeks of the previous year, so the team treated it as the season. Read at the time, both sale weeks looked below the 3.5 break-even and a manager proposed cutting budgets on the second Monday; the plan said to hold, and the re-read a fortnight after the sale put both weeks above it once late purchases had filled in. On the sale's third day a new banner script stopped the browser purchase event for five hours; the server events continued, the developer reverted the script, and no campaign was edited. The promotions review then found about a fifth of sale orders had been pulled forward from the following weeks, which went into next year's plan alongside the weekly cost per thousand.

Common mistakes

  • Launching new ad sets on the first day of the sale. They spend the dearest days of the year in learning.
  • Restructuring in the final week before the window. Targeting, event and bid changes restart learning when it can least be afforded.
  • Adding sale ads one at a time. Batch the additions into one edit per ad set.
  • One large budget jump on day one. Raise in steps during the run-up, or use a lifetime budget on a dedicated campaign.
  • Sharing a budget between prospecting and retargeting. The small pool absorbs the peak spend and its frequency climbs.
  • Sale prices missing from the catalog. Overlays and prices in catalog ads depend on the fields being populated.
  • Judging the sale while the window is open. The final days are always the least complete.
  • Editing in response to a seasonal price rise. Compare with last year's weekly cost per thousand before calling it a problem.
  • Judging on Meta's return alone. Reconcile with orders and measure incremental orders, pull-forward and margin given away.

Frequently asked questions

When should I launch Meta ads for a sale like Black Friday?

Early enough for the ad sets to have left the learning phase and settled before the window opens. Meta's documentation on the learning phase describes new and significantly edited ad sets as delivering less stably and at a higher cost per result until enough optimisation events accumulate, and it does not give a lead time, because that depends on how quickly your ad sets gather events. Launch the structure in the run-up with evergreen creative, then add the sale ads when the sale starts, rather than building new ad sets on the first day.

Should I raise my Meta budget for a sale?

Usually, but plan the amount and the route. Size it with a peak-period cost per purchase rather than the annual average. A daily budget is an average that Meta documents can flex upwards on high-opportunity days within limits, so a modest daily figure already leans into the best days. Large budget changes can restart learning, so raise in steps during the run-up rather than in one jump on the first day, or give a fixed sale window its own campaign with a lifetime budget.

Why do my sale results look worse in Ads Manager than in my shop?

Partly timing. Meta credits conversions to the date of the ad interaction, so purchases from people who clicked during the sale keep arriving for days after it, and the last days of the window look weakest while it is still open. Partly price: auctions get dearer in busy seasons, so cost per purchase rises even when the ads work. Read the sale once the attribution window has filled, and reconcile Meta's figures with the shop's orders before judging it.

Terms used on this page

Tools to try

Frameworks

Sources

  1. About the learning phase — Meta Business Help Centre
  2. About Advantage+ campaign budget — Meta Business Help Centre
  3. About attribution settings — Meta Business Help Centre

Related