Where does the holiday number actually get decided? Adobe's forecast puts US online spending at $275.1 billion for November 1 to December 31, up 6.7%, with $47.5 billion across Cyber Week, $15.1 billion on Cyber Monday and $12.9 billion on Black Friday. The same forecast assigns 57.4% of the season to mobile, which means the ceiling on the quarter is set on a screen roughly six inches wide. Up to November 10, 2026, the most useful way to read those numbers is not as a demand forecast but as a description of where a store can lose money.
The forecast is a mobile number
| Forecast item | Figure | What it constrains |
|---|---|---|
| Season total (Nov 1 to Dec 31) | $275.1 billion, +6.7% | Total demand, not your share of it |
| Cyber Week | $47.5 billion, +7.4% | Five days carry the year's concentration |
| Cyber Monday | $15.1 billion | Peak load on checkout and payment |
| Mobile share | 57.4% of the season | Layout, speed and wallet behaviour |
The mobile share is the part that changes work. When more than half the season runs on a phone, a friction that barely registers on a desktop becomes a majority problem: a wallet button that fails to load, a keyboard that covers the postcode field, a product page whose specification table needs horizontal scrolling. None of those show up in a revenue report, and all of them show up in conversion.
Adobe also points to AI traffic growth of 130%, a channel most stores cannot yet isolate in analytics. Assistants send shoppers who arrive with a narrower question and a shorter patience, and they land on whichever page the answer cited. That makes page clarity a referral issue, not only a conversion issue.
What a phone changes about a product page
A product page written for a browsing session does not survive a phone held in one hand. The specification table has to be readable without zoom, the delivery promise has to be stated before the add-to-cart button rather than in a tab, and the return terms have to be visible without a click. These are structural changes, and they are cheaper to make in November than in December.
The second structural item is payment. Every wallet a store offers is a path that can fail independently, so the test is not whether checkout works, but whether it works once per payment method on a real handset with a real card. Teams that test one method and assume the rest are testing their own optimism.
What the discount depth cannot fix
A deeper discount raises intent but does not repair a broken step. If the leak is at payment or at address entry, extra depth simply moves more shoppers into the same wall, and the cost of that depth is booked whether or not the order completes. Discount decisions and checkout decisions should therefore be reviewed in the same meeting, in that order reversed: fix the path, then price it.
Assistant traffic has no dashboard
AI referral traffic is the newest channel with the least instrumentation. A shopper who asks an assistant for a comparison and clicks a citation arrives on a page chosen by a model, often deep in the catalogue, frequently without a campaign parameter. If the store cannot see that traffic, it cannot tell whether the page converted or the page confused.
The practical move is to treat page clarity as an acquisition cost. A page that states price, delivery window, return terms and specification in plain text is both easier for a shopper to trust and easier for a retrieval system to quote. The same edit serves two channels, which is rare enough to be worth scheduling first.
Tracking gaps land in the same five days
Measurement is the quieter risk. A tracking gap costs the same five days as peak spend, and it is usually invisible until reconciliation. The order of work that survives a deadline is prove then fix: place real test orders, compare platform-reported conversions with backend orders, and use the size of the gap to separate a capture problem from a deduplication problem, a consent problem or a click-identifier problem.
Freeze, do not rebuild
Between November 26 and 30 there is no time to rebuild attribution and trust the result. Freezing changes after a final passed test, and monitoring backend orders against platform conversions daily, is the only discipline that fits the window. A new server-side sender stacked on an old one during Cyber Five produces exactly the double counting the rebuild was meant to remove.
Consumer demand this season is both price-sensitive and time-compressed, which adds a second constraint. A survey of 1,501 consumers put average expected holiday spend at $1,018, 5.5% less than a year earlier, and in Walmart's own customer panel 43.8% named affording the holidays as their top stressor while 70.2% said lower prices would make the season easier. Demand that is price-sensitive and time-compressed rewards the store with the shortest path from citation to checkout, not the deepest discount.
The ten listings that carry the week
With a fortnight left, the highest-return hour is spent on the ten listings that will take most of the weekend's traffic: their stock position, their price after discount, their page clarity and the advertising behind them. Getting those ten right beats anything new that could still be added, because a new element cannot be tested twice before the peak and an untested element is a risk with no upside.
Everything else follows the same logic. The forecast tells you the size of the window; the phone tells you where the window is; and the tracking tells you whether you can see through it. Only the second and third are within a store's control this month.
FAQ
Why does the mobile share matter more than the total forecast?
Because 57.4% of the season runs on a phone, so a mobile defect is a majority defect. Small frictions that are invisible on a desktop, such as a wallet that fails to load or a field hidden by the keyboard, cost more than any desktop issue.
How should AI referral traffic be handled this month?
As an acquisition channel with a page-clarity cost. State price, delivery window, return terms and specifications in plain text so both a shopper and a retrieval system can use the page, and try to isolate assistant referrals in analytics rather than folding them into direct traffic.
What is the safest tracking work two weeks before peak?
Prove, then fix. Place real test orders and compare platform conversions with backend orders to locate the gap, fix one primary conversion per platform, then freeze changes after a final passed test and monitor daily.
Which ten listings deserve the last hour of preparation?
The ten that will take most of the weekend's traffic. Check their stock position, their price after discount, their page clarity and the advertising behind them; an untested new element is a risk without upside at this point.
