Black Friday falls on 27 November this year, which gives you a little over three weeks. Cyber Monday lands on 30 November.

Adobe's forecast, published at the end of September, puts the US online holiday season at 275.1 billion dollars, up 6.7% on last year. Cyber Week alone is projected at 47.5 billion, with Black Friday at 12.9 billion and Cyber Monday at 15.1 billion. The full forecast is worth reading if you sell online.

The useful part is not the headline. It is that two things have changed since last year in ways that affect what you should fix in the next three weeks, and both of them are measurable.

---

What the forecast actually says to prepare for

• A record 63% of peak-day spend is expected on mobile.

  Up from 61.6% last year, with the full season at 57.4%. Your phone experience is not a secondary channel during Cyber Week. It is the channel.



• Traffic from AI tools to US retail sites is forecast to rise 130% year on year.

  Adobe expects the largest jumps on Thanksgiving at 159%, Black Friday at 95% and Cyber Monday at 88%. Last year that traffic grew 670% on Cyber Monday from a small base.



• Buy now pay later is forecast at 21.3 billion dollars across November and December.

  Mobile accounts for 82% of it, which tells you where the basket-size decisions are being made.



• Cyber Week is 17.3% of the entire season's online spend.

  Five days carrying roughly a sixth of two months. Whatever breaks will break then.



• Black Friday is growing faster than Cyber Monday, at 9.2% against 6.2%.

  The single-day peak is still Cyber Monday, but the Friday is where the growth is.

One caveat worth stating: these are forecasts from one analytics provider measuring US online spend, published in September. Treat them as the shape of the season rather than a promise, and note that the AI traffic figure is measuring people clicking through from AI chat and browser tools, not AI agents buying things on their own.

---

1. Load test this week, not in the last one

Traffic concentrated into five days finds every weakness you have been tolerating. The time to discover that your checkout falls over at four times normal traffic is now, while there is still room to fix it.

Run a realistic load test against a staging environment: homepage, category pages, product pages, add to cart, and the full checkout, not just the front page. Watch your database and your payment provider's sandbox, since those are the usual choke points rather than the web server. If your hosting plan has a burst limit, find out what it is before you meet it on a Friday morning.

2. Fix the mobile checkout, in that order of priority

With 63% of peak spend expected on phones, a desktop-first checkout is a decision to lose the majority of your opportunity.

Test it on a mid-range Android device on mobile data, not on a flagship on wifi. Check that the numeric keypad appears for card and postcode fields, that the pay button is reachable without scrolling past a sticky banner, that address autocomplete works, and that wallet payment options appear at the top rather than buried under a card form.

3. Verify every tracking tag before the spend starts

Broken conversion tracking during Cyber Week is expensive twice: you misjudge which campaigns worked, and your bidding algorithms optimise against bad data for the most important week of the year.

Fire a test purchase and confirm it appears in every platform you will be spending on. Check that purchase values and currencies are passing correctly, not just that an event registered. Confirm your consent banner is not silently blocking tags for a share of visitors. Do this now, because the first week of a sale is the worst possible time to discover a measurement gap.

4. Decide the offer mechanics, then test them in the cart

Most discount problems are not strategy problems. They are stacking problems discovered by a customer at 2am.

Decide whether codes combine, whether discounts apply before or after shipping thresholds, what happens to a bundle when one item sells out, and how a return is valued when the item was bought in a two-for-one. Then actually run each case in the cart. Write down the answers so whoever is handling support on the day does not have to invent a policy.

5. Make your offers legible to the tools people are shopping with

This is the newest item on the list and the one most stores will skip. With AI traffic up 130%, a meaningful share of shoppers will ask an assistant for the best deal on something before they ever reach a store.

That means your product pages need the facts in text rather than in a promotional image: current price, original price, what the discount is, availability, delivery estimate, and the terms. Keep your structured data accurate and matching what the page displays, especially price and stock, since a sale price that lives only in a banner image is invisible to anything reading the page. A deal nobody can parse is a deal nobody recommends.

6. Add buy now pay later only if the basket justifies it

At a forecast 21.3 billion dollars, this is no longer a fringe payment option. It is also not free, and the fees are typically higher than card processing.

If your average order value is high enough that the instalment framing changes the purchase decision, it earns its place, and 82% of that volume being mobile tells you where to display it. If you sell twelve dollar items, it adds a logo and a cost. Decide on the numbers rather than on what competitors are doing.

7. Publish your delivery cutoffs where people are deciding

Shipping deadlines are the most searched piece of information on any ecommerce site in December, and they are usually buried.

Get the cutoff dates from your carriers, publish them on the product page and in the cart rather than only on a policy page, and update them as they pass. Stating a date that a customer can trust prevents both abandonment and the support queue that follows a missed Christmas.

8. Freeze the code, and know how to roll back

The single highest-return decision in this entire list costs nothing and requires no work. Stop deploying changes a week before Thanksgiving.

Every retailer with a bad Black Friday story has a deployment in it. Agree a freeze date, document what the emergency exception process is, make sure one named person can roll back a release at 3am, and test that the rollback actually works. Then do not touch anything.

---

The discount nobody needed to give

Here is the part that does not fit on a campaign plan. For most independent stores, the biggest money lost over Cyber Week is not lost to insufficient discounting or a weak campaign. It is lost to a checkout that takes nine taps on a phone, a tracking tag that stopped firing in October, and a server that got slow at exactly the hour the traffic arrived. None of those have a creative brief attached, so they get attention last.

There is a harder commercial point too, and it deserves saying plainly even though it argues against the season. Deep annual discounting trains your customers to wait. If a meaningful share of your December revenue is people who would have bought in November at full price, the sale moved the revenue rather than creating it, and it did so at a lower margin. Some businesses need the volume, the cash flow or the new customers badly enough that this is still the right trade. Many run the sale because last year they ran the sale.

Worth doing the arithmetic at least once: what percentage of your Cyber Week orders came from customers who had never bought before. If it is high, the discount is buying you something durable. If it is low, you gave your existing customers money and called it a campaign.

Either way the infrastructure work above pays for itself, because it applies to every order you take for the rest of the year rather than only the discounted ones.

---

Three checks to run today

• Buy something from your own store on a phone, on mobile data.

  Count the taps from product page to confirmation. Anything above six on a returning-customer path is worth a conversation.



• Place a test order and confirm it appears in every ad platform you will spend on.

  Check the value and currency, not just that something fired.



• Ask an AI assistant for the best price on one of your own bestsellers.

  See whether you appear, whether the price it reports is correct, and whether your competitor's page explained itself better than yours did.

An hour of work, three weeks of runway, and the results will tell you whether your next three weeks go on campaign creative or on the checkout.

Need your store load tested, tracked and mobile-ready before the 27th? Talk to Spark while there is still time to fix what the test finds.