Ask any DTC team what their Q4 change freeze looks like and you’ll get a date and a document. Code freeze on a named day, publishing pause, no theme changes, no app installs, an on-call rota, a named person who can approve an exception. It’s boring and it’s standard and every serious Shopify operation runs one.
Ask an Amazon brand the same question and you get a version of: “yeah, we try not to change much in Q4.”
That’s not a policy. That’s a shared assumption, held by two parties who have never compared notes on what it covers — and after managing hundreds of brands through peak, we can tell you exactly how it fails. Somebody ships a change in November that everybody thought was off the table, and nobody can point at the sentence that said so, because there isn’t one.
The uncomfortable part: your Amazon account has more open write paths than your Shopify store does. A flat file, a repricer, a listing tool, a catalogue contribution from another seller, an agency’s bulk upload, Amazon’s own auto-published content changes. The DTC team froze the surface where every change requires a deploy. The Amazon team left open the one where changes arrive on their own.
The freeze everyone assumes exists
Two things get confused here, so let’s clear them first.
There is an Amazon-side freeze — the system-generated listing freeze that happens when Amazon locks a detail page during a review or a data conflict. That’s something that happens to you. Not what we’re talking about.
There’s also the creative measurement deadline we’ve written about separately: the calendar working backwards from peak that determines whether a creative change can produce a clean read before traffic distorts. That’s about when a change stops being measurable.
This is different and broader. This is a written change-control policy across the whole account: what may be touched between mid-October and early January, what may not, who decides, and what happens when something genuinely urgent lands on a Saturday.
Most brands have none of it in writing. Both they and their agency believe the other party is holding the line.
A freeze without an exception path is worse than no freeze
This is the part we want brands to internalise before they write anything down, because we’ve watched it go wrong more than once.
If you announce “no changes in Q4” and stop there, urgent things still happen. Of course they do — a listing gets suppressed, a competitor undercuts you by 12%, a campaign is at 4x its normal spend by noon, an image gets flagged. The freeze doesn’t prevent the event. It just removes the legitimate route for dealing with it.
So the change gets made anyway. At 9pm. By whoever was awake and had access. Outside the process, because the process said no. With no record of what was changed or why, because logging it would mean admitting the freeze was broken.
You’ve now got the worst possible combination: changes still happening, and no visibility into them. In January, when you’re trying to work out why a hero SKU’s conversion moved in week two of December, the answer is in somebody’s memory and nowhere else.
A freeze is a routing document, not a prohibition. Its job is to make sure changes go through one door instead of five.
Most brands freeze the cheap things and leave the expensive ones open
Here’s the inversion we see constantly on takeover accounts.
The things that get frozen tend to be the things that are easy to reverse: bids, budgets, campaign settings. Somebody says “don’t touch the ads in Q4,” and the account runs November on October’s bids into inflated CPCs, with a campaign capping out at 2pm on Black Friday because nobody was allowed to raise the budget.
The things that stay open tend to be the ones that are hard or impossible to reverse: a flat-file push that overwrites forty titles, a variation restructure, a new ASIN created under the wrong brand field, an image swap on the hero, a category change. Nobody froze those because nobody thought of them as changes. They thought of them as work.
A useful test when you’re drafting: for every item, ask how long it takes to undo and what it costs while it’s wrong. A bid is thirty seconds and a few hundred dollars. A variation restructure is a case, a queue, and a week of split review counts during the highest-traffic period of your year.
Freeze by reversibility, not by department.
The one-page document
Four columns. It should fit on one page, and if it doesn’t, it won’t get read.
Column 1 — Frozen. No changes, no exceptions inside the window.
- Main image and image stack order
- Title, bullets, and backend search terms
- Variation structure (adding, removing, or re-parenting children)
- A+ content resubmission (the review queue runs longer in Q4 and restarts on rejection)
- Category, browse node, brand name field, GTIN
- Campaign structure: creating, archiving, or restructuring campaigns and ad groups; match-type reorganisation; bid strategy changes
- List price and strike-through architecture
- New tool or integration authorisations
Column 2 — Explicitly not frozen. Say this out loud or people will freeze it by default.
- Bids and budgets (these are your peak levers; freezing them is the most common self-inflicted wound we see)
- Negative keywords
- Factual corrections to specs, dimensions, compatibility statements
- Suppression, compliance, and policy fixes
- Inventory actions: removals, replenishment, reserve deployment
- Case work and IP enforcement
- Deal execution already committed
Column 3 — Requires named approval before it happens.
- Price moves beyond an agreed band (write the band as a number)
- Coupon or promotional depth beyond an agreed percentage
- New ASIN creation
- Any bulk upload touching more than a stated number of SKUs
- Anything that touches a Column 1 field
Column 4 — Who can override, and how to reach them.
One named person. One named backup, because one of them will be travelling. A phone number, not a Slack handle. Slack gets read once.
The dates, hedged as always
Work back from your own account’s notifications rather than from any blog post, including this one — Amazon stages these by account and the dates have moved earlier every year.
The shape most brands should be working with for 2026:
- BFCM inbound cutoffs land across mid-to-late October (AWD earliest, Amazon-optimised splits latest)
- BFCM deal submissions close around October 20
- Peak fulfilment fees run October 15 through January 14
- A+ review queues run longer in Q4 and restart from zero on a rejection
That puts a sensible freeze window at roughly October 15 to January 10. Note the back end: the freeze should not lift on December 26. The last two weeks of December and the first week of January are when gift returns land, when Q4 reconciliation happens, and when a well-meaning change destroys the baseline you need to grade the quarter.
Trade guidance broadly converges on locking listings four to six weeks before a peak event, on the reasoning that a change made seven days out may not be fully reflected in search results by the time traffic arrives. That’s directional rather than precise, but the direction is right: a change made too close to the event gets neither the benefit nor a clean read.
Run the escalation drill in September
The freeze document has a twin, and almost nobody writes it: what happens when something breaks during peak.
Pick three scenarios and walk them through with your team and your agency on a call this month.
The failure mode here is never that nobody knows what to do. It’s that the person who knows what to do doesn’t have the access, or the person with the access doesn’t know they’re the one who’s supposed to act. You find that out in a twenty-minute drill in September or you find it out at 6pm on a Friday in November.
The question to ask your agency this week
One email, and the reply tells you a great deal:
“What is your Q4 change freeze policy on our account, what are your coverage arrangements for the last week of November and the last two weeks of December, and who is the named escalation contact with authority to act without a meeting?”
A shop that runs this properly replies with dates, a list, and two names. A shop that replies with a paragraph about being responsive and available has told you something too — and September is a much better month to learn it than December.
Ask for their side of the freeze in the same email. Your agency has its own Column 1, and it should overlap with yours but won’t match it exactly. The overlap is where the argument happens in November, so have it now.
FAQ
Isn’t a change freeze just an excuse not to optimise during the highest-traffic quarter?
Only if you write Column 2 badly. Bids, budgets, negatives and inventory are your active levers and they should be moving daily in peak. The freeze exists to stop structural changes — the ones that take days to stabilise and weeks to reverse — from landing on the most expensive traffic of the year.
When should the freeze actually start?
Earlier than most brands want. Anything that needs to be measured should be live before the traffic distorts, which for creative means mid-September. Anything that needs Amazon to process it — A+ approval, variation changes, new ASIN indexing — needs to clear before queues lengthen. The freeze date is the date after which you stop starting things, not the date you stop working.
We’re a five-person brand. Is this overkill?
It’s the opposite. Large organisations have policy that catches some of this by accident. A five-person brand has whatever the founder set up, plus an agency, plus a VA, plus a listing tool — four write paths and no document. The one-pager takes an afternoon and it’s the cheapest Q4 risk reduction available right now, because unlike inventory or deals, there’s no queue in front of it and no dependency on anyone but you.
What if we’ve already missed the window on something we need to change?
Then decide it as a trade rather than sliding into it. A factual correction that prevents returns is a repair, not a bet, and repairs are always worth shipping. A speculative improvement to a hero image three weeks before Cyber Monday is a bet you can’t grade and can’t easily undo. Those two things should not get the same answer.
Should the freeze cover our own website and other channels?
It should at minimum name them. If you correct a spec on Amazon during the freeze because it’s driving returns, that correction needs to propagate — a factual fix that lands on one channel and not the others reproduces the problem you just solved somewhere nobody is watching.
—
If you’re looking for a team that manages every lever — creative, advertising, and operations — Velocity Sellers works with brands doing $100K+/month on Amazon. Contact us for a free account audit.
