Stop using last year’s raw week total
Copying “we sold 800 last Black Friday” into this year’s PO is how you repeat stockouts. Last year’s total is censored whenever you were out of stock, and it ignores this year’s price, traffic and assortment. Build a per-SKU uplift model instead.
Step-by-step method
- Export daily units for each SKU for last year’s BF week (and the prior four baseline weeks).
- Flag stockout days (zero sales with zero fulfillable stock). Remove them from the baseline average or impute expected demand.
- Compute baseline daily demand from in-stock days.
- Compute event uplift = event-week in-stock daily rate ÷ baseline daily rate.
- Apply this year’s modifiers: price change, deal depth, ad budget, distribution, new variants.
- Forecast event-week units = this year’s baseline daily × uplift × modifiers × days.
- Translate into cover: compare to on-hand + inbound that arrives before each day of the week.
Worked example (hero vs long-tail)
Deal hero: baseline 20/day. Last BF week sold 210 across five in-stock days (42/day) with two OOS days ignored in the rate. Uplift = 42/20 = 2.1×. This year ads +20% and price −5% → modifier ≈ 1.15. Peak daily ≈ 20 × 2.1 × 1.15 ≈ 48/day. Five-day need ≈ 240 units before safety.
Long-tail: baseline 3/day, last BF uplift only 1.2×, no deal → peak daily ≈ 3.6. Do not steal FBA capacity from the hero to overstock this line.
Deal products vs the rest
Split the catalogue into deal SKUs, halo SKUs (attach/traffic) and the rest. Deal SKUs get uplift from both discount and ads; halo SKUs get traffic spillover; the rest often see mild uplift or even declines if budget shifts. Mixing them into one “shop uplift” hides the stockout on the deal ASIN.
Stockout-day correction matters
If you were OOS three days last BF, the week total is too low and a naive year-on-year “+30%” still undershoots. Prefer in-stock daily rates. When history is messy, triangulate with ad-attributed units and on-site conversion from analytics.
From forecast to PO and transfer
Convert peak daily into a short-horizon cover target (event days + short buffer). Size FBA send-ins with the send-in calculator and supplier buys with reorder point thinking — even if the PO will not arrive until after peak, you need the truth for January.
Common mistakes
- Using revenue instead of units when prices changed.
- Applying one shop-wide uplift to every SKU.
- Ignoring parent/child variation.
- Counting inbound that arrives after Cyber Monday as peak cover.
- Forgetting returns inflate post-peak on-hand.
Worked table narrative
Baseline 28-day sales 560 units (20/day). Event uplift 2.0×. Modifier 1.1. Peak daily 44. On-hand 300; inbound 200 arriving 20 Nov. Cover through 27–30 Nov works if receiving completes; if receiving slips to 2 Dec, you only have 300÷44 ≈ 6.8 days entering the weekend — act on ads or FBM.
A 10-day operating cadence
Peak weeks reward boring rhythm. Use a daily 20-minute stand-up with the same agenda: stockouts in the next 72 hours, inbounds that slipped, ad spend on thin cover, and warehouse labour. Write decisions in a shared note so night shifts and agencies see the same truth.
Assign a single owner for Amazon inbound appointments, a single owner for 3PL labour, and a single owner for ads budgets. Split ownership without a RACI chart is how two people both assume the other paused the campaign.
On Mondays, refresh forward daily rates from the last seven sellable days plus the planned promo calendar. On Wednesdays, re-check restock limits and stranded inventory. On Fridays, snapshot cover for the top 50 SKUs and archive it — you will want that file when January asks what happened.
Cash, fees and contribution profit
Black Friday is not only a units problem. Ad auctions clear higher, refund rates rise, and Amazon festive peak fulfilment fees (where they apply) sit on top of ordinary FBA fees during the published peak window. Re-run contribution profit with peak CPC and peak fees before you approve air freight.
A simple gate: emergency inbound is allowed only when expected contribution after peak fees and freight still clears your floor margin. If it does not, throttle demand instead of buying expensive volume that arrives as the spike ends.
Keep a watch list of aged FBA risk for leftovers. Units that miss the sale and sit into Q1 can attract storage pressure. Plan removals or outlet pricing in December rather than discovering the bill in February.
People, carriers and cut-off times
Inventory plans fail when nobody booked Saturday labour. Confirm 3PL cut-offs for Friday and Cyber Monday in writing. Ask which carriers still pick up late afternoon. Print a packing rules sheet for FBA cartons so you do not burn a day on labelling errors.
Customer service needs macros before the weekend: delay, partial shipment, substitute, and “back on X date”. Silence in the inbox is not calm customers — it is chargebacks later.
After the weekend
On Tuesday, reconcile what sold versus the forecast by SKU. Raise or cut January POs from that file, not from vibes. Mark which backups (FBM, transfers) you would reuse next year and which were chaos. That note becomes next October’s checklist.
How OmniRestock helps
OmniRestock keeps FBA and own warehouses as separate stock sources, forecasts per SKU, and shows stockout and order-by dates so peak decisions stay coherent. Draft POs respect MOQ and case pack; stock write and supplier emails stay off until you enable them. Alerts can go to email, Slack or Teams. If you need a prioritised list before connecting everything, start with the free stockout report.
Stockout-day correction worked example
Suppose ASIN A sold 900 units across last year’s Black Friday week, but was out of stock for two of the seven days. Naïve year-on-year uplift on 900 understates demand. Estimate daily sell-through on in-stock days only (900 ÷ 5 = 180/day) and reconstruct a full-week potential of 1,260 before you apply this year’s uplift. If you expect +20% traffic and the same conversion, plan 1,512 units of sellable cover across FBA and own warehouse combined — then subtract what is already inbound.
Deal ASINs need a separate uplift from non-deal peers. A 40% deal on a hero SKU can triple velocity while a mid-catalogue SKU without a badge stays flat. Never average them into one brand-level percentage and push that into every PO.
Channel mix inside the forecast
If Amazon was 70% of peak units last year and Shopify 30%, do not assume the same split if you are stacking TikTok creatives or a retail media push. Forecast each channel, then constrain by a shared stock pool so the sum of channel plans cannot exceed physical units. That is how you stop overselling when every channel spikes at once.
Work the maths in the sell-through calculator and keep definitions aligned with demand forecasting.
Deal versus non-deal split you can defend
Export last year’s peak week by ASIN and tag each row deal / non-deal / halo. Halo products ride traffic from a deal without their own badge — they need a mild uplift, not the deal multiplier. Build three uplift assumptions (conservative / base / aggressive) and only purchase to base until mid-November capacity and ad auctions are visible.
Correct returns and cancelled orders out of the history before you uplift. A 12% cancellation rate on a viral creative last year will reappear if you reuse the creative; planning to gross shipments without that haircut overstates replenishment need and ties up cash in the wrong carton.