The Post-Peak Review: Turning Black Friday Data Into a 2027 Plan

The weeks after Cyber Week hold some of the most useful data you’ll see all year, and many teams never look at it. By December, the next priority has arrived, and the debrief slips to “after the holidays.” Then January fills up too.

That habit is costly. Adobe forecasts a record $275.1 billion in U.S. online holiday sales this season. For most merchants, that means the heaviest load of the year on your site, your team, your warehouse, and your support queue. How all of it held up tells you where next year’s problems and opportunities are.

This post lays out a practical way to run a post-peak review. It covers two review dates instead of one, a short list of what to pull, a method for turning findings into decisions, and a reusable agenda.

Why most post-peak reviews fall short

Most reviews fail for the same few reasons, and none of them is about effort. These are our observations from working with merchants.

• They happen once, and too early. A single meeting right after Cyber Week can’t see returns, refunds, or late-season trends, so it judges the season on incomplete numbers.
• They have no owner. Everyone attends, nobody is responsible for the follow-up, and the action items fade within a week.
• The data lives in too many places. Analytics, the order system, the warehouse, support tickets, and carrier invoices each tell part of the story. Pulling them together is a chore, so it gets postponed.
• They stay vague. “Checkout felt slow” is a feeling. “Checkout error rate doubled between 8 and 10 p.m. on Black Friday” is something a team can fix.
• They produce just a list. A page of observations with no ranking, owner, or date changes nothing.

Each of these is fixable with a little planning ahead of time. The rest of this post shows how.

Use two review dates, not one

Schedule one review in early December and a second in mid-January. The reason is simple: the season doesn’t finish when Cyber Week does.

Returns are the clearest example. Adobe reported that one in eight returns in the 2024 holiday season happened between December 26 and 31. Last year, returns ran 2.5% below the prior year through December 12, then rose 4.7% year over year in the days after Christmas. Adobe also expected volumes to stay 8% to 15% above early-season levels through the first two weeks of January.

A review held in the first week of December would have missed most of that. It would have shown a calm season right before the busiest return stretch.

Here is how we would split the work. The dates are suggestions, so adjust them to your calendar.

• Review one, the week of December 7: the front half of the season. Sales, traffic, site performance, promotions, fulfillment, support, and incidents. The data is fresh, and the team still remembers what happened.
• Review two, the week of January 11: the back half. Returns, refunds, fraud and abuse, net margin after costs, and a final look at anything that surfaced late.

Book both now. Put them on the same calendar as your code freeze and your staffing plan, and name one owner for each. Once peak starts, nobody has time to find a slot that works for everyone.

What to pull from each area

For every area below, collect the numbers first, then ask what they mean and what you would change. The goal is a decision, not a data dump.

Traffic, conversion, and order value

Pull daily traffic, conversion rate, average order value, and revenue from Thanksgiving through Cyber Monday and the weeks around it. Compare each against last year and against your October baseline. Adobe projects Thanksgiving at $6.9 billion, Black Friday at $12.9 billion, and Cyber Monday above $15 billion in U.S. online sales, so check whether your peak followed the market’s shape or shifted. Look especially for days where one metric moved and the others didn’t, such as traffic up while conversion stayed flat.

Site performance and incidents

Pull error logs, checkout failure rates, page speed by day, and your peak log if you kept one. Look for when problems started, how long they lasted, and what it took to fix them. Then decide which fixes belong before next peak, and which safeguards were missing, such as alerts, a rollback plan, or a named on-call owner.

Promotions

Pull results by offer: orders, discount cost, margin after discount, and how often customers stacked offers. Adobe expects discounts of up to 30% during Cyber Week, so deep offers will be common. Look for promotions that drove volume but little profit, and for discounts that behaved unexpectedly at checkout. Decide which offers to repeat, change, or drop.

Fulfillment and shipping costs

Pull delivery times, delay rates, and your carrier invoices. Holiday surcharges deserve a close look. For standard shippers, UPS and FedEx set a residential Ground surcharge of $0.50 per package, rising to $0.75 (UPS) or $0.80 (FedEx) at peak. For shippers above 20,000 packages a week, both carriers tier the charge by volume against a June 2026 baseline. Compare what you planned for against what you were billed, and ask whether your free-shipping threshold held up.

Customer support

Pull ticket volume by topic, response times, and the most common complaints. Look for questions that a clearer product page, policy, or confirmation email could have prevented. Decide which answers belong on the site instead of in a queue.

AI-referred traffic

Adobe projected that AI traffic to U.S. retail sites would grow 130% year over year this season, with the biggest jumps on Thanksgiving (159%), Black Friday (95%), and Cyber Monday (88%). Pull sessions, conversion, order value, and landing pages for AI-referred visits, and compare them with your other channels. If you couldn’t isolate that traffic, treat that as a finding and fix the measurement before next season.

Save returns and refunds for the January review

A December sales number is not final until the returns arrive. NRF and Happy Returns estimate that $849.9 billion in merchandise will be returned in 2025, about 15.8% of annual sales, and that 9% of returns are fraudulent. Online, an estimated 19.3% of sales come back.

By mid-January, you can see the full picture. Pull these numbers:

• Return rate by SKU and variant, with the reason codes customers chose
• Refund timing, meaning how long refunds took and when you issued them relative to package receipt
• Abuse and fraud signals, such as repeat non-delivery claims, return weights that don’t match the item, and the same product returned across several accounts
• Net margin by order cohort, after discounts, shipping, and returns

Then look for patterns you can act on. A product with a high fit-related return rate may need better sizing guidance. A cluster of “not as described” returns points to content. Repeat claimants point to a policy or verification gap.

The questions to answer are practical. Which policies helped, and which cost you? Did refunds go out faster than your team could verify returns? And which return reasons could be prevented upstream, on the product page or in the confirmation email? Those answers feed next year’s merchandising and policy decisions directly.

Turn findings into decisions

A review is only worth the time if it ends with choices. Sort every finding into one of three buckets:

• Fix now. High stakes, likely to repeat, and quick to address. These get done in the first weeks of the new year.
• Plan. Worth doing, but larger. These go on the Q1 or Q2 roadmap with an owner and a date.
• Set aside. Real but low-impact, or too costly for the benefit. Write it down and revisit after next peak.

To place each finding, ask three questions. How much revenue or margin is at stake? How likely is it to happen again next peak? How much effort or risk does the fix carry? A finding that scores high on the first two and low on the third goes to the top of the list.

Then record each decision in a simple log with four parts: what you saw, what you decided, who owns it, and when you will check whether it worked. Here is a hypothetical entry:

• Seen: checkout errors doubled on Black Friday evening.
• Decision: add error alerting and a documented rollback step.
• Owner: engineering lead.
• Check: end of January, with a test that confirms alerts fire within minutes.

Keep the priority list short, ideally five to seven items. Fifteen priorities means none of them are priorities. “Set aside” is also a decision, and naming it keeps the team from relitigating the same idea every January.

A reusable agenda and prep list

Here is a 90-minute agenda you can reuse for both holiday season review dates. It is our suggested format, so adjust the timing to your team.

1. Set the frame (5 minutes). The owner states the goal and what is out of scope.
2. Walk through the numbers (25 minutes). Each area lead brings three numbers and one surprise.
3. Review incidents and near-misses (15 minutes). Work from the peak log, not from memory.
4. Name what worked (10 minutes). Identify what to protect and repeat, not only what broke.
5. Sort findings into buckets (25 minutes). Fix now, plan, or set aside.
6. Assign owners and dates (10 minutes). No item leaves the room without both.

Most of the work happens before the meeting. Complete this prep list before peak starts, while there is still time:

• Book both review dates and name one owner for each.
• Capture your October baseline for traffic, conversion, and order value.
• Start a shared peak log and tell the team where to add to it.
• Set up a way to isolate AI-referred sessions in your analytics.
• Ask each area lead to bring three numbers and one surprise.
• Decide where decisions will be recorded so they don’t live in someone’s notes.

The simplest version of all of this

If you do nothing else, do three things.

1. Book the two review dates now.
2. Keep a shared log while peak is happening.
3. And end each review with a short list of decisions, each with an owner and a date.

Peak season will generate more data than any other stretch of your year, but data only helps if someone reads it and acts on it. A scheduled, structured review is how that happens. If your numbers are scattered across systems and a review feels hard to run, that is the kind of problem our Reporting & Analytics Sprint is built for.