Why peaks need a freeze
Peaks concentrate risk. A sale, a seasonal gifting period, a product launch or a fundraising appeal puts a large share of the year's demand into a short window. A fault that would cost little on an ordinary day costs much more when it lands in that window, and it is harder to diagnose when traffic, prices and campaigns are all changing at once.
Many of those faults are self-inflicted. A tag edited to add a new event stops the purchase tag firing. A template change breaks the layout of the sale email. An automation rule updated for a new segment catches the wrong customers. An integration upgraded the week before the peak drops orders on their way to the CRM. None of these changes was wrong in itself; it was made at the wrong time, with less room to test and less time to recover. A move between affiliate platforms is one such change, and the tutorial on moving an affiliate programme to a new platform sets its switch date outside the freeze.
Platforms add a second reason. The site's glossary entry on the learning phase describes the period after a Meta ad set is launched or significantly edited, during which results are volatile and cost per result usually runs higher, and the guide to Meta ads around sales and seasonal peaks plans launches early so ad sets settle before the window opens rather than being edited late. The guide to Google Ads bidding around sales and tracking outages covers seasonality adjustments and the changes to hold back while a period is unusual. This tutorial does not add to what those pages say; it builds the freeze around them.
Freezes already exist on the site for single events. A rebrand rollout sets a freeze around its launch date, and the website migration checklist freezes non-essential content changes before a move. This tutorial applies the same idea across the whole marketing operation for a known high-stakes period. A period nobody planned for, such as an outage or a serious incident across the business, needs a faster version of the same discipline, which the tutorial on leading marketing through a business crisis sets out.
Before you start
- Know the peak. The commercial plan, the promotion calendar and the guide to planning promotions say when the high-stakes periods fall and what they are for. For an online store's sale or seasonal peak, the peak season trading playbook places the freeze between the build and the readiness gate and makes its trading lead the freeze's decision-maker.
- Name the freeze owner and the decision-maker. One person runs the freeze; one person, often the head of marketing or the operations lead, decides exceptions. Where the team has written its governance rules down, the freeze rule and its owner belong there, and the tutorial on writing a marketing operations charter is where that owner is named.
- List the systems and their owners. The tag container, the website and its templates, the email platform and automated journeys, the ad accounts, the CRM and integrations, and the campaign taxonomy. Note who can change each, including agencies and developers.
- Read the change controls that already exist. Container governance sets the review and publishing rules for the GTMTag management system. container, and automation governance sets change control for automated journeys. The freeze tightens these for a period; it does not replace them.
- Look back at past peaks. What broke, when, and what change caused it. That history is the best guide to what the freeze must cover and when it should start.
- Check capacity, not only change. If the peak may bring more demand than the business can serve, the guide to planning marketing around capacity and stock covers that separately.
The steps
Name the period and why it matters
Write one sentence that names the period and why it is protected: the gifting peak, the spring sale, the launch of a new range, the year-end appeal. Include what is at stake, such as the share of annual revenue or the campaign the business has spent most on. A freeze people understand is a freeze people respect; one announced as a rule without a reason is worked around.
List the kinds of change the freeze covers
Sort changes with the one-way and two-way doors framework: what is costly or slow to reverse during the peak is frozen. Most marketing freezes cover tags and tracking, site templates and checkout-adjacent pages, email templates and automated journeys, ad account and campaign structure, integrations between tools, and the campaign taxonomy and naming. List each kind of change by name so there is no argument later about whether something counts. A change that would also alter how results are measured is frozen by default, because it makes the peak's own results harder to read.
Decide what stays allowed and through which route
List the routine, reversible work that carries on: social posting, scheduled sends from approved templates, content publishing in existing templates, replies and community management, and budget moves within limits agreed in advance, watched through the team's normal budget pacingHow fast a campaign spends its budget across the flight compared with the plan.. Say which route each goes through, which is usually the normal one: the fast lane and launch check described in the campaign workflow guide and the reviews the approval matrix already requires. Being explicit about what is allowed is what stops the freeze being read as "stop everything".
Set the start and end dates
Set the start early enough that changes made before it have time to be tested and to settle before the peak, and the end late enough to cover the peak and the days immediately after it, when orders, returns and reporting are still flowing. Base the dates on the team's own history of past peaks and on what the platform pages say about launching early, not on a figure borrowed from elsewhere. Put the dates on the marketing calendar so everyone can see them alongside the campaigns.
Finish, test and record changes before the freeze starts
Everything the peak needs should be built, tested and live before the freeze begins: new tags, sale templates, automated journeys for the period, new ad structures. Run the sale promotion readiness checklist where a sale is involved. Record every change made in the run-up, and date any change that touches measurement in the measurement change log, so that if numbers move during the peak, the team can see what changed and when. Anything not finished and tested by the start date waits until after the freeze.
Write the exceptions route and who decides
Some changes will be needed during the freeze: a broken checkout tag, a journey sending the wrong offer, a pricing error on a landing page. Write a short route: the person raising it describes what is broken, the proposed fix, how it will be tested and how it will be reversed if it makes things worse. The named decision-maker approves or declines it. Approved changes are made by whoever normally makes them, tested, recorded and, if they touch measurement, dated in the change log. Name a deputy decision-maker for absences, so the route works when the decision-maker is unavailable.
Tell agencies, developers and other teams
Write a freeze notice and send it to everyone who can change the covered systems: agencies, freelancers, developers, the web team, the CRM team, regional marketers and anyone else with publishing rights. The next section lists what it says. Ask each recipient to confirm they have read it. Agencies and developers often have their own release schedules, and a notice that reaches them late is the commonest reason a freeze is broken by someone who did not know it existed.
Hold requested changes in a queue during the freeze
Requests for frozen changes still arrive during the peak. Record each in a queue with the request, the requester, the reason and the date, rather than refusing it or quietly making it. The queue shows requesters their change has been heard, gives the team a ready list for after the freeze, and reveals patterns: several requests for the same fix may mean something is genuinely broken and belongs in the exceptions route.
Lift the freeze in order and review what happened
Do not release the whole queue on the first day after the freeze. Release held changes one at a time or in small groups, most important first, testing each before the next, so that if something breaks, the cause is clear. Then review the freeze: what broke during the peak and whether a change caused it, how many exceptions were raised and whether they were genuine, what the queue held, and whether the dates and the list of covered changes were right. Feed the answers into the plan for the next peak.
What a freeze notice says
The notice is short enough to read in a minute and specific enough that nobody has to ask what it means.
| Section | What it says |
|---|---|
| Period and reason | The name of the peak, the start and end dates and one sentence on why it matters |
| What is frozen | The kinds of change covered, listed by name: tags, templates, journeys, account structure, integrations, taxonomy |
| What carries on | The routine work that continues and the route it goes through |
| Exceptions | How to raise one, what to include, and who decides, with a deputy |
| The queue | Where to log requests for frozen changes, and when they will be looked at |
| Measurement | That any approved change touching tracking is dated in the measurement change log |
| Contacts | The freeze owner and the decision-maker, with how to reach them out of hours |
| After the freeze | That held changes are released in order, not all at once |
An online toy retailer freezes before its gifting peak (illustrative)
Rizzambo Toys freezes before its gifting peak
Rizzambo Toys, an invented online toy retailer, takes about 40% of its annual revenue in the 7 weeks before the end-of-year holidays. The previous year, a tag change made by its agency in early December had stopped the purchase event firing for 2 days, and a template update the same week broke product images in the gift-guide email.
The operations lead planned a freeze from mid-November to the first week of January. It covered tag and tracking changes, site and email templates, automated journeys, ad account structure and the integration between the shop and the email platform. Social posting, scheduled emails from the approved gift templates, content in existing templates and budget moves within limits the head of marketing had set all carried on through the normal route.
The new gift-guide templates, the holiday journeys and the peak ad structure were built, tested and live a week before the freeze began, and every tracking change was dated in the measurement change log. The freeze notice went to the agency, the freelance developer and the customer service team, and all 3 confirmed it.
During the freeze, the queue collected 11 requests, from a new pop-up to a reworded footer. One exception was raised: in the second week of December the checkout's purchase tag stopped firing after the payment provider changed its confirmation page. The analytics lead described the fix and how to reverse it, the head of marketing approved it within the hour, and the fix was tested, published and dated in the change log the same afternoon.
In January the team released the queued changes 2 or 3 a day, testing each. The review found no breaks caused by the team's own changes during the peak, and moved the freeze start a week later for the next year, because everything had been ready earlier than needed.
Common mistakes
- Freezing everything. A freeze that stops posting and scheduled sends is ignored within days. Freeze what is hard to reverse and let routine work continue.
- Freezing nothing in particular. "No big changes during the peak" invites argument about what counts. List the kinds of change by name.
- Starting the freeze before the work is finished. Changes rushed in on the last day before the freeze are the ones most likely to break during it.
- No exceptions route. Without one, real fixes are either blocked or made quietly, and both are worse than a recorded exception.
- Forgetting agencies and developers. The freeze is broken most often by someone who never heard about it.
- Untracked exceptions. An approved change that is not dated in the measurement change log makes the peak's results impossible to explain later.
- Releasing the whole queue at once. If something breaks, nobody can tell which change caused it. Release in order and test each.
- No review afterwards. The next freeze repeats the same dates and gaps. The governance hub shows where freeze rules sit beside the approval matrix and the team's other decision rights.
