Through November 10, 2026, the loudest number in ecommerce was volume, and the quietest was incrementality. Singles Day is still the largest shopping event on the calendar, larger than Black Friday and Cyber Monday, and brands are still pouring budget into marketplaces and live commerce to ride it. The question the latest numbers raise is uncomfortable: how much of that revenue is genuinely new demand, and how much is a discount handed to a customer who would have bought anyway.
The number nobody reports
Sales across major platforms reached 1.02 trillion yuan in the three weeks to October 31, up nearly 2.4 percent year on year, on Syntun data reported this week. The point most campaign reviews skip is simpler than the headline: an ad platform earns whether the order came from a new customer or from an existing one who waited two weeks for a coupon. The platform revenue is real either way; the brand category growth is not.
This is not an argument against promotion. It is an argument about measurement. If a brand cannot separate new buyers from deferred buyers, it cannot tell whether the campaign bought growth or rented it for a fortnight. The distinction matters most in categories with short repurchase cycles, where a deferred purchase shows up as a hole in next month numbers.
Two baselines, two answers
A campaign can be judged against last month or against the same month a year earlier. In a category that is growing, both baselines flatter the campaign. In a category that is shrinking, both make it look worse than it was. The honest baseline is a holdout group that sees no promotion at all, and almost nobody runs one during peak season.
Agents are becoming the entry point
The channel itself is being rebuilt. Google Universal Cart is rolling out across Search and Gemini, consolidating offers from multiple retailers into a single checkout flow. Walmart has upgraded its Sparky assistant, and Amazon has added more than 100 shopping functions to Alexa. The TikTok Shop app and Meta shopping engagement numbers point in the same direction: social apps are turning into full commerce stacks rather than top of funnel traffic sources.
| Entry point | What it changes for a brand | What it hides |
|---|---|---|
| AI agent | Product data must be machine readable | Which brand was considered and dropped |
| Social commerce | Catalogue and checkout move inside the app | Attribution across platforms |
| Marketplace search | Bidding still decides placement | Whether the buyer was new |
When an agent assembles the cart, the brand loses the moment where a shopper compares two facings. What remains is structured data: price, availability, delivery promise, reviews. A brand that has not published those in a machine readable form is not in the consideration set at all, and it will never see that exclusion in a traffic report.
What the ad numbers actually show
Amazon advertising business reported roughly 19.6 billion dollars in the third quarter of 2026, up about 12 percent year over year. That growth sits on top of a marketplace where brands already pay for placement, fulfilment and now insurance. Each line item is defensible on its own; together they explain why a brand can grow revenue and shrink margin in the same quarter.
The practical response is to treat measurement as a budget line rather than an afterthought. A holdout test on one category, run for four weeks around the event, costs less than the discount depth most brands accept without testing. It also produces a number that survives the next planning cycle.
Where the discount actually lands
Discount depth and incrementality are inversely related in most categories. A ten percent cut may pull forward purchases from the following month. A thirty percent cut pulls forward purchases and recruits new buyers, but the new buyers often return to their previous brand once the price normalises. The useful question is not how deep the discount was, but how many of those buyers were still buying ninety days later.
Running a review that answers the question
None of this makes peak season less important. It makes the review more honest. A brand that knows how much of the event was incremental can decide next year whether to buy the same attention again, or spend the same money on retention.
FAQ
What does incremental mean in this context?
Revenue that would not have happened without the promotion. It excludes purchases pulled forward from later weeks and orders from buyers who would have paid full price.
Why do platforms not report it?
Their revenue does not depend on the answer. A deferred purchase and a new customer generate the same ad spend.
How do you measure it without a huge budget?
Pick one category, hold out a small share of the audience for four weeks, and compare repeat purchases at 90 days.
What changes when agents assemble the cart?
Structured product data becomes the entry ticket. Brands without machine readable price, availability and review data are excluded before any comparison happens.
