A peak season pre-flight checklist

Black Friday sits at the end of November, so the work that decides how peak week goes is happening now. Within weeks the calendar will be locked, the creative will be in production, and the only decisions left will be the ones made under pressure at eight in the morning.

The failures we get called about in peak week are almost never creative failures. A subdomain that started sending at volume on the Monday. A suppression list that stopped syncing on the Friday. A frequency cap somebody switched off "just for the promotion". A reply address nobody was monitoring.

Every one of those is an October problem with a cheap October fix. This is the order we work through them, roughly one section per week between now and the first sale send.

Freeze the sending setup now

Sender reputation is built from history, and a new configuration has none. Anything that changes how your mail is authenticated, or where it appears to come from, should be finished and stable well before volume climbs.

The freeze list we agree with clients:

• No new sending domains or subdomains introduced after this month.
• No changes to SPF, DKIM or DMARC records from the start of November until December.
• No platform migration, and no switching of the sending IP pool.
• No change to the from-name or from-address, however much the brand refresh wants one.
• No new third-party tool given permission to send as your domain.

If one of these changes genuinely has to happen, do it this week rather than next month, and give it four weeks of ordinary sending before peak. A change made in the quiet period is a change with time to recover from.

Plan the volume ramp rather than improvising it

Most peak-season deliverability damage is self-inflicted, and it comes from a volume step-change. The programme sends a steady weekly broadcast all year, then in one week it mails the lapsed file, the never-purchased file and everyone who unsubscribed from one channel but not another.

Mailbox providers read that as a different sender. Build the ramp deliberately instead.

1. Write down your ordinary weekly volume by provider, as it stands today.
2. Write down the peak-week volume the commercial plan implies.
3. If the gap is more than roughly double, stage the increase across the weeks between now and then.
4. Bring reactivation audiences back gradually, oldest-engaged first, starting now rather than in November.
5. If you are adding a new IP or pool, the warming schedule needs to be complete before the ramp starts, not running alongside it.

The reactivation point matters most. A lapsed segment mailed for the first time in a year, during the busiest inbox week of the year, will produce complaints at a rate that damages everything else you send. Mail those people in October, at low volume, and keep only the ones who respond.

Set the frequency budget before the calendar fills

Every team plans peak in the same way: each owner adds their sends to a shared calendar, and the total is whatever the sum turns out to be. Nobody agreed the total, so nobody owns it.

Invert that. Decide the maximum number of commercial contacts one customer may receive in a week, across all channels, and treat it as a budget the calendar has to fit inside. Then enforce it centrally with frequency capping rather than by editing the calendar.

Some rules worth settling in writing now:

• The cap counts email, SMS and push together. A customer does not experience them as separate programmes, and our notes on choosing the channel apply with more force when volume is high.
• Transactional and service messages sit outside the cap, and must not be repurposed as promotions to get around it.
• Triggered journeys yield to broadcasts during peak week, or the other way round, but the priority order is written down before the week starts.
• One named person can raise the cap, and only in writing.

The budget will feel too low when you set it and about right on the Tuesday of peak week.

Rehearse the send, including how it fails

A rehearsal is not a test send to the marketing team. It is the full message, built in the production environment, sent to a seed list that covers the mailbox providers your file is made of. Then it is checked on the devices your customers actually use.

Rehearse the whole path: the segment query, the personalisation fallbacks, the link tracking, the landing page under load, the unsubscribe, the reply address. Send one at an awkward hour to see what the timing logic does with it.

Then rehearse the failures, because peak week is when they happen:

• The product feed does not refresh and the recommendations render empty. What does the template show?
• A price is wrong in the email and right on site. Who can pull the send, how fast, and what goes out instead?
• The platform queues and the send lands three hours late. Is a late send better than no send, and who decides?
• Inventory sells out at midday. Which sends are cancelled automatically and which need a human?

Write the answers on one page, with names and phone numbers, and put it where somebody can find it on a Saturday.

Decide now what you switch off

Peak season is additive by default. Every triggered programme built during the year keeps running underneath the promotional calendar, and the collisions are unpleasant. A win-back offer lands on the same morning as the sale. A price-drop alert undercuts a campaign that went out an hour earlier.

Go through the automation inventory and mark each programme as run, pause or amend for the peak window. Most abandonment and browse journeys should keep running, because they carry high intent. Discount-led reactivation and lifecycle offers usually pause, since the broadcast calendar is already doing that job harder.

Amend means a specific change: a reduced offer, a different exclusion, a tighter cap. Write the change and the date it reverses, or January will arrive with a programme still in peak configuration.

What to do this week

• Publish the freeze list and get the platform and engineering owners to acknowledge it in writing.
• Pull today's weekly volume by mailbox provider, set it against the peak-week plan, and build the ramp if the gap is more than double.
• Start the lapsed-file reactivation now, at low volume and oldest-engaged first, so November mails a file that has already been tested.