Harbolnas 2026 runs for seven days, 10 to 16 December. Trade Minister Budi Santoso announced the schedule on 27 August 2026 and said he hopes transaction value will rise about 5 to 10 percent over last year's Rp36.4 trillion. Two weeks later, Thursday 24 December is the joint Christmas leave day and Friday 25 December is Christmas itself, according to the three-minister decree.
That means the online store, the point-of-sale system, and the website you run will carry their heaviest load of the year exactly when your team is thinnest. Our view is simple: in a week like that, the change everyone calls small is the one most often released without oversight, and so it is the one that costs the most when it goes wrong.
Small is about line count, risk is about time left
Picture three lines of code on the payment page. It looks trivial. A developer ships it on Wednesday night because it was "while we're here", and nobody tests it with real data. On Thursday afternoon, at the longest queue of the week, the pay button fails for one card type. The person who understands the code is on leave.
The size of a change does not set its risk. What sets it is how much time is left to find the mistake before customers find it first. A change shipped two weeks before the peak has two weeks to be noticed and fixed. A change shipped the day before has one day, and it is the busiest day.
Large companies already work this way
Freezing changes ahead of a peak season is not a new idea. Gergely Orosz, writing in The Pragmatic Engineer on 3 January 2023, described how Meta runs a code freeze starting in November, alternating "yellow" and "red" periods, to avoid trouble around Black Friday, Christmas, and New Year. The reason quoted is blunt: minimise revenue lost due to outages, especially for advertising products.
The same piece says freezes at large tech companies typically last about one to two weeks around Thanksgiving and one to two weeks around Christmas through New Year. It also records the criticism: freezes cost productivity and do not automatically remove risk. That second point matters, and we come back to it below.
Three decisions to make today
You do not need a heavy process. You need three decisions, written down and known to everyone who touches the system.
- One last-change date. Pick a working day, not the busy day itself. With Harbolnas starting 10 December, Friday 4 December leaves a full week of watching.
- One request list. Collect every change request before that date, for example until Friday 27 November. Whatever misses the list waits for January, without a second debate.
- One person who may declare an emergency. Name them before the season starts. An "is this an emergency?" call made in a panic almost always leans toward yes.

The dates in the figure are only an example. Count back from your own peak day, whether that is Harbolnas, payday, or the school holidays.
What may still go in after the freeze
A freeze is not a total ban. The line is simple: if transactions still work, the change waits. Failing payments, wrong prices or stock shown to customers, or a security hole being actively exploited are fair reasons. Changing a button colour, tidying code, or adding a small feature "while we're here" is not.

What people forget to freeze
Most teams freeze application code and call it done. Yet plenty of outages start elsewhere: a plugin that updated itself overnight, one payment-gateway field changed, a complicated discount rule rushed in, or an SSL certificate that expires in the middle of a campaign. Put all of those inside the freeze, and renew domains, hosting, and certificates before the freeze date.

When an emergency really happens
Emergencies will still happen, and that is normal. What changes is the way you handle them. The person who declares the emergency should not be the person who deploys the fix, so two sets of eyes are involved. The way back is prepared before the fix goes out: you know exactly how to return to the previous version. Afterwards you write down what changed, why, and at what time, so nobody has to guess the next morning.

If a full freeze is not realistic
The criticism above is fair: a rigid freeze also holds back useful fixes. So start with the lightest version. Freeze only the areas that touch money: payment, pricing, stock, and login. Let everything else move under tighter watch. A freeze that everyone follows beats a perfect one that is quietly broken.
Start today
Write down the three decisions, send them to the team chat, and put the dates on a shared calendar. If you would like your point-of-sale system or website checked before the season and supported through the freeze, the XETUP team can help through xetup.id or bisnis@xetup.id.
