Amazon Peak Season strategy — Prime Day, Black Friday and the Q4 sprint.

"Peak Season" on Amazon isn't Black Friday. It isn't Prime Day either. It's the compound of three separate events, stacked against a Q4 gifting tail, that between them do 30–45% of a typical seller's annual revenue in about nine weeks. The sellers who win peak are the ones who plan it as one arc, not three unrelated promotions — and who solve the problem that every peak-season bidding system has on Day 1: the trailing 14-day performance window they normally bid off is completely irrelevant the moment traffic doubles.

This guide covers the core play — rolling bids back to the equivalent window of last year's peak to jumpstart Day 1, then migrating campaign by campaign onto current-season performance as live data accrues — plus the three peak windows, the inventory lead time that gates all of them, and the full bidding arc from T-14 to T+21.

The core play — jumpstart from last peak, then migrate

Every automated (and most manual) bidding workflows on Amazon bid off a trailing window: last 7 days, last 14 days, last 30 days of CVR, ACOS and CPC. That works for 47 weeks of the year. It breaks the week peak starts, because the trailing window contains none of the peak-shaped demand you're about to price into.

The consequence is predictable. Sellers who let their trailing-window logic drive bids into Prime Day, PBDD or Black Friday show up on Day 1 with bids calibrated for October baseline traffic. Everyone else — the ones who prepared — outbids them for the opening 24–48 hours, wins the placements, and banks the highest-CVR clicks of the year. By the time the trailing window catches up on Day 3, the deal is over.

The fix is a two-phase play — jumpstart, then migrate.

Phase 1 — Jumpstart from the previous season's window

Fourteen days before the event, roll your bidding reference window backwards: instead of bidding off the last 14 days, bid off the equivalent 14-day window of last year's peak. For BFCM 2026, that's November 15–28, 2025. For PBDD 2026, that's the two weeks around PBDD 2025. For July Prime Day, the same window a year prior.

Concretely, for each keyword or target you already own:

1. Pull the previous-year peak-window search-term report (SP), keyword performance report and placement report.
2. Recompute the bid as ASP × CVR_lastPeak × target_ACOS × placement_index — using last peak's CVR, not the current trailing window's.
3. Apply the previous-peak placement modifier too — top-of-search modifiers that were correct in October will be badly under-tuned for the 96-hour BFCM window.
4. Stage the change T-3 to T-1: upload the jumpstart bids the day before the event opens, not the morning of. Amazon's budget and pacing systems need a few hours to route around the new bid landscape.

Two adjustments matter. First, inflate last-peak CVR by year-over-year category growth if you have it (Brand Analytics or Marketing Stream will tell you). Second, if the ASIN has meaningfully more reviews or a better main image than it did a year ago, add 10–20% on top — a better-listed version of the same ASIN converts higher at the same traffic.

For keywords you did not own last year (new ASINs, new campaigns), use the closest analogue you have: the same keyword on a sibling ASIN, or the category median CVR from Brand Analytics for that search term. It's less precise than your own history, but it's still bidding off peak-shaped data instead of baseline-shaped data — which is the entire point.

Phase 2 — Migrate to current-season data as it lands

The jumpstart bids are a scaffold, not a plan. They are wrong for two reasons the moment the event actually starts: the current season is not last season, and the sample you need to trust current-season data lands quickly once traffic triples. The job is to migrate off the previous-peak anchor and onto live current-peak data as fast as the statistical signal allows — but not faster.

A weighted-blend migration works well. Set the effective bid for each keyword as:

bid = w · bid_livePeak + (1 − w) · bid_lastPeak

where w is the confidence weight on current-season data. Ramp it as clicks accumulate on the keyword during the event:

• 0–15 clicks in the event window: w = 0. Hold the last-peak bid. Sample is too small.
• 15–50 clicks: w = 0.3. Blend, but the previous-peak anchor still dominates.
• 50–150 clicks: w = 0.6. Current-season CVR is now the primary signal.
• 150+ clicks: w = 1.0. Fully migrated — bid purely off live peak performance. Drop the last-peak anchor.

For a two-day event like Prime Day or PBDD, most hero-ASIN keywords will reach the 150-click threshold by end of Day 1; the long tail migrates over Day 2. For the eleven-day BFCM arc, the hero terms migrate inside 24 hours and the tail catches up by the middle weekend. For the December gifting tail, most terms never fully migrate — traffic per keyword is too diffuse — and the last-peak anchor keeps doing useful work into January.

Rebuild the trailing window carefully post-peak. Do not roll straight back to a 14-day trailing window on January 2 — that window will still be 70% peak-shaped data and will produce badly over-tuned bids for baseline January traffic. Either shorten the trailing window to 7 days for the first two weeks of January, or explicitly exclude the peak window from your baseline calculation. Coming out of peak is symmetric to going in: the trailing-window logic is wrong in exactly the same way, just in reverse.

This whole play — freeze the reference window, roll it back a year, migrate on live signal, then rebuild the trailing window post-peak — is what most peak-season "war rooms" actually spend the event doing. The reason to write it down as a system is that in the middle of BFCM, sitting in front of a bulk-operations sheet, is not the moment to be inventing the methodology.

Reading the previous season correctly

The jumpstart is only as good as the last-peak data you anchor to. Three cleanup steps before you use it:

• Strip out-of-stock days. Any day last peak where the ASIN lost the Buy Box or went to zero inventory has artificially depressed CVR. Exclude those days from the reference window entirely, don't just under-weight them.
• Normalise for last year's promo mechanic. If the ASIN ran a 20%-off Lightning Deal last BFCM and isn't running one this year, last peak's CVR is not your CVR. Discount the reference CVR by roughly the observed deal-CVR uplift for your category (typically 30–60% for Lightning Deals, 15–25% for Prime Exclusive Discounts).
• Separate branded from non-branded terms. Branded search CVR is stable across peak; non-branded CVR moves with promotion intensity. Migrate branded terms off the last-peak anchor faster (they don't need the scaffold as long); hold non-branded terms on the anchor longer.

The rest of this guide covers the event calendar, the inventory lead times that gate everything, and the bidding arc that surrounds the jumpstart-and-migrate core play.

The three peak windows

Prime Big Deal Days — October

The "second Prime Day" — a two-day Prime-exclusive event in early-to-mid October. Traffic is 60–80% of July Prime Day but conversion rates are higher: shoppers who show up in October are gift-buyers, not deal-hunters. Treat it as the opening salvo of Q4, not a summer event.

Black Friday / Cyber Monday — late November

The biggest single traffic window of the year on Amazon. Officially a five-day event (Thanksgiving through Cyber Monday), realistically an eleven-day one — Amazon has stretched "Black Friday Week" earlier and earlier, and the deal-hunter traffic now starts the Monday before Thanksgiving.

The gifting tail — December 1 to December 22

Under-planned by most sellers. Traffic drops from the BFCM peak but conversion rates stay elevated for three full weeks, and CPCs fall as competitors pull back thinking peak is over. The last-shipping-day cliff on around December 22 is where the tail ends.

Inventory lead times — the deadline nobody moves

The single most common peak-season mistake is planning the campaign calendar before the inventory calendar. Amazon Warehousing and FBA inbound windows are the real deadline; the marketing calendar is a consequence.

• Prime Big Deal Days inbound cutoff: typically early September. Miss it and your October units check in as regular inventory with no deal-eligibility guarantee.
• Q4 / BFCM inbound cutoff: typically early-to-mid November. FBA warehouses run at 120–140% capacity in the two weeks before BFCM; shipments that arrive late get parked and don't check in until December.
• Working stock buffer: plan 8–10 weeks of cover at forecast peak run-rate, not baseline. Under-cover on the hero ASIN and Amazon suppresses the Buy Box on the highest-traffic day of the year.

The bidding posture in the next section assumes you have cover. If you don't, throttle down — over-bidding into an under-covered ASIN converts a merchandising problem into an out-of-stock accelerator.

The bidding arc — pre-peak, peak, post-peak

Pre-peak (T-14 to T-3 days) — harvest

Two weeks before each event, traffic and CPCs are both up 20–40% versus baseline, but conversion rates are only marginally elevated. This is a harvesting window, not a selling one. Push auto and broad campaigns aggressively to capture the search terms shoppers are using to research; move the converters into exact-match manuals before the event fires.

Peak (T-2 to T+2 days) — bid to win the placement

In the peak window, CPCs spike 2–4x and conversion rates spike alongside them. Down-only bid strategies leave money on the table here — Amazon lowers your bids at exactly the moment placement matters most. Switch hero-ASIN campaigns to fixed or up-and-down for the 96-hour peak window, then switch back.

Raise top-of-search placement modifiers on the campaigns that already convert at Row 1 — a 100% top-of-search modifier with a fixed strategy is the closest thing Amazon gives you to a "buy Row 1 for the day" button. Combine with a coupon or a Prime Exclusive Discount to render the badge, and the combination usually beats a Lightning Deal for CTR on established ASINs.

Post-peak (T+3 to T+21) — pace, don't pull

The mistake most sellers make in the third week of December is turning campaigns off because "peak is done". It isn't. CPCs drop back near baseline, conversion rates stay high, and the ROAS window from December 5 through December 20 is often the cleanest of the year. Keep budgets funded through the last-shipping cliff — that's where the tail lives.

Budget pacing — where the money actually goes

A rough allocation for a Q4 peak plan, indexed to a month-of-October baseline PPC spend:

• Prime Big Deal Days week: 1.6× baseline.
• Weeks between PBDD and BFCM: 1.2× baseline — sustain the ranking gains from October.
• Black Friday week: 2.5–3× baseline.
• Cyber Week + first gifting week: 2× baseline.
• Gifting tail through Dec 22: 1.3× baseline.
• Dec 23 – Jan 5: 0.6× baseline. Traffic is returns-adjacent; convert rates collapse.

These are starting points, not targets. The real pacing job is watching the intra-day budget-exhaustion pattern during each peak window and shifting spend to campaigns that clear their budget by 4pm.

Deals to submit — and when

• Prime Exclusive Discounts — the peak workhorse. No fee, no submission window, renders the deal badge for Prime members. Layer on every hero ASIN that has the margin.
• Best Deals / Deal of the Day — 6–8 week submission window ahead of the event. If you didn't submit by early September for BFCM, you won't get one.
• Lightning Deals — bookable closer to the event, but the good peak slots go fast; book by mid-October for BFCM.
• Coupons — the fill-in mechanic for the in-between weeks. See the Coupons guide for the badge-rendering thresholds.

The peak-season mistakes that cost real money

• Bidding into Day 1 off a trailing 14-day window. The single most expensive peak mistake. Your bids are calibrated for October traffic and everyone who prepared is outbidding you on the highest-CVR clicks of the year. Roll the reference window back to last peak before the event opens.
• Never migrating off the last-peak anchor. The mirror mistake. Set the jumpstart, then forget to weight in live current-peak data as clicks accumulate. Last year's CVR is a starting scaffold, not a target — by the middle of Day 2, live signal should dominate.
• Rolling straight to a 14-day trailing window on Jan 2. Coming out of peak is symmetric. A trailing window that contains BFCM will over-tune bids for baseline January traffic. Shorten the window or exclude the peak days explicitly.
• Down-only bidding on hero ASINs during peak. You are asking Amazon to lower your bid on the day conversion rates triple. Switch hero campaigns to fixed or up-and-down for the 96-hour peak, then switch back.
• Turning campaigns off on December 15. The gifting tail is the cleanest ROAS window of Q4. Pace, don't pull.
• Chasing new-to-brand launches into peak. Peak is not the window to test a new ASIN — CPCs are punishing, review velocity gets buried and the launch coupon competes with everyone else's PED. Launch in September for the October harvest, or wait until January.