Black Friday ecommerce checklist for 2026: what to automate, freeze and test with three weeks left
A Black Friday ecommerce checklist for the last three weeks: what to repair, automate, freeze and test before 27 November, and what not to start.
By Rokibul Hasan8 min read
- E-commerce
- Automation

It is 4 November 2026. Black Friday falls on 27 November and Cyber Monday on 30 November, so you have about three weeks. This Black Friday ecommerce checklist leaves offers and email to others and covers the automations, integrations and feeds that have to keep working when orders arrive faster than usual.
It is for owners and operations people on Shopify, Amazon or both, with a handful of apps and scripts holding the business together. It gives a dated order of work, from repairs to what to leave until December.
I build automation for Shopify stores and Amazon sellers. I do not run a store, so there is nothing here about discounts or ad spend. The freeze dates are my own recommendations, not platform rules. Platform facts were checked on 4 November.
The Black Friday ecommerce checklist by date
| When | Job |
|---|---|
| 4 to 8 November | Repair what is running: API versions, webhooks, fee figures, reports |
| 9 to 13 November | Add small automations you can undo, then freeze integration code |
| 14 to 24 November | Run the test list, load sale schedules, write the runbook |
| 25 November to 1 December | Full freeze: fixes for live faults only |
| 1 December | Write down what failed while you still remember it |
If you only have time for one row, do the first. New automation cannot help if an old one has stopped and nobody has noticed.
Size the weekend from your own numbers
Adobe's 2026 forecast, reported by Digital Commerce 360 on 29 September, expects US online sales of $275.1 billion between 1 November and 31 December, up 6.7%, with Black Friday at $12.9 billion, Cyber Monday at $15.1 billion and AI traffic to US retail sites up 130%. Those are whole-market forecasts. They say nothing about how busy your store will be.
Your load comes from your own figures. Take last year's busiest hour and count what one order sets off: a stock update to every other channel, a fulfilment message, an accounting entry. Then look for load you create yourself, which is usually worse.
An illustration with invented figures: 300 orders in the busiest hour, three lines per order, three other channels. That is 2,700 stock pushes an hour, under one a second. The same store then changes 5,000 variant prices at midnight. If each update costs 10 points on Shopify's GraphQL Admin API, that is 50,000 points. Shopify restores 100 points a second on Standard plans, 200 on Advanced and 1,000 on Plus, so the job takes over eight minutes on Standard, with some products still showing the old price. Your real cost per call is in each response's extensions.cost field, so measure it.
What has already closed
Amazon's holiday 2026 post on the Seller Forums closed deal submissions for Black Friday Week and Cyber Monday on 20 October. Its arrival dates for FBA stock were 21 October for shipments with minimal splits and 28 October for Amazon-optimised splits. If you missed those, plan around the stock you have.
The same post says holiday peak fulfilment fees run from 15 October 2026 to 14 January 2027 for FBA, Remote Fulfillment, Multi-Channel Fulfillment and Buy with Prime, with the 3.5% fuel and logistics surcharge on top. Anything that holds a fee figure may still use off-peak numbers: repricer floors, margin dashboards, or a rule that decides whether to ship a Shopify order through Multi-Channel Fulfillment. Update those this week.
Repair first: 4 to 8 November
API versions. Shopify's versioning page lists 2026-01 as the oldest accessible version; 2025-10 went on 16 October 2026. A call pinned to something older does not fail. Shopify "falls forward" and answers from the oldest accessible version, so behaviour can change with no error in your logs. Note the version each custom app and script asks for.
Webhooks. Shopify retries a failed webhook up to eight times in four hours, treats no reply within five seconds as a failure, and says that "if failures persist, the subscription is removed". A handler that does its work before replying can pass all year and start timing out at peak. Reply first, queue the work, and list your subscriptions to confirm they still exist.
Tokens. If an integration refreshes a 24-hour Dev Dashboard token, confirm a failed refresh alerts someone. The pattern is in my guide to Shopify legacy custom apps.
Amazon reports. Amazon removes two older settlement report types on 11 November. If a tool of yours uses them, see the November news roundup.
Reconciliation. Shopify's documentation says webhook delivery "isn't always guaranteed". If no scheduled job compares stock across channels, this is the one new thing worth adding at this date.
What is still safe to automate
Three weeks is enough for changes that are small, can be undone in minutes and replace nothing.
| Automate | Why it is safe this late |
|---|---|
| Alerts on failed syncs and stuck queues | Reads only, and can be removed at once |
| Stock buffers on the last few units per channel | One number per channel, easy to set back to zero |
| A daily report of orders, refunds and channel differences | Read-only |
| Timed sale changes through Rollouts | Replaces a midnight manual job and reverts by itself |
Shopify's Rollouts, under Markets > Rollouts in the admin, combine theme, discount and catalogue changes into one timed release. An Event rolls back when it ends. A Launch is permanent, so pick with care. Shopify says you can schedule or run up to five at once, which limits a run of daily flash sales.
There is one trap for integrations. Shopify's changelog of 1 October says: "When a discount is part of a Rollout, its native dates alone don't describe its availability." Anything reading startsAt and endsAt, such as a feed, an email tool or a custom app, can report a sale that differs from what shoppers get. If an integration depends on those dates, keep those discounts out of Rollouts this year, because discounts outside Rollouts behave as before.
Freeze dates
| Freeze from | What | Why |
|---|---|---|
| Friday 13 November | Integration code, app upgrades, new scopes | Two weeks of ordinary trading to expose a bad release |
| Wednesday 18 November | Theme, checkout code, new apps | Later visual changes go through a scheduled rollout only |
| Friday 20 November | SKUs, variant structure, channel mappings, feed rules | A renamed SKU can break sync without any error |
| Tuesday 24 November | Prices, discounts and schedules, loaded and previewed | Nobody edits a live sale from memory |
| Wednesday 25 November to Tuesday 1 December | Everything except fixes for live faults | Thanksgiving is 26 November |
A freeze needs an exception rule. Name one person who can approve a fix, and record every change made during it.
What to test, in this order
Use real orders where you can, then refund them.
- One order per channel, end to end, through fulfilment, tracking and the accounting entry.
- The last unit. Sell it on one channel and time how long the others take to show zero.
- A partial refund and an exchange. A quiet month rarely exercises these paths.
- A discount at its boundary. Preview the rollout, then order just after the start time.
- A dead consumer. Stop your webhook handler for ten minutes, restart it and confirm reconciliation recovers what it missed.
- Throttling. Force rate-limit errors in a test store. The queue should wait and retry without losing work.
- The stop switch. Pause every automation that writes, then confirm nothing writes. On Amazon, the Agent Policy requires tools to stop when asked.
- Orders from Google. On Shopify, direct checkout in AI Mode and Gemini is on by default for eligible stores, and Shopify says Google Analytics and custom pixels do not fire on those orders. Decide whether to keep it, using my guide to Google Universal Cart.
What not to start before December
A change needs quiet days to show its faults, and few are left.
- A new order management, warehouse or ERP connection. Run the old process through the peak and migrate in January.
- A platform move. Three weeks is not enough.
- A first custom app or a new sales channel. If you must have something now, an App Store app with a clear off switch beats custom work started this late.
- AI agents with write access to prices, listings or refunds. Drafts and reports are fine.
- Your own UCP endpoints or WebMCP tools. Stay on what your platform provides this season.
- REST to GraphQL rewrites and new theme architectures. Worthwhile, and not urgent this month.
The one-page runbook
Write this by 24 November and print it: who is on call and when, how to pause each channel, how to stop a sale early, where the logs are, and the status page for each app and platform.
On 1 December, write down what failed or nearly failed. That list is the best brief you will ever have. If it points at custom work, my marketplace and e-commerce plans start from $2,500. That is a starting price, and every project gets a fixed quote after a short brief.
FAQ
Is it too late to automate anything before Black Friday 2026?
No, as long as the change is small and easy to undo. With about three weeks left you can still add failure alerts, stock buffers, timed sale launches and read-only reports. Leave anything that replaces a core system, or that changes prices, listings or refunds without a person checking, until December. A change needs some quiet trading days to show its faults, and after mid-November there are few left.
When should an online store freeze changes before Cyber Week?
No platform sets this, so choose your own dates. A staged freeze works well: integration code about two weeks before Black Friday, theme and checkout code about nine days before, product structure and feed rules a week before, and everything except live fixes from the Wednesday before Thanksgiving until the Tuesday after Cyber Monday. Name one person who can approve exceptions, and record every change made during the freeze.
Should I switch off my repricer or other automation during Cyber Week?
Usually not, because switching off a tool you rely on during your busiest days swaps a known risk for an unknown one. Check its limits instead. Confirm price floors include peak fees, test the stop switch and make sure someone is watching its log. Turn a tool off only if it misbehaves, which is why you test the stop switch first.
If a fix on this Black Friday ecommerce checklist is beyond your team this month, or your December notes point at something bigger, send me a short brief: the channels you sell on, what connects them today, what failed or nearly failed, and when you need it working. You can start from the marketplace MVP plan, see the e-commerce automation work I do, or read how to prevent overselling with multichannel inventory sync.


