The National Retail Federation has published its 2026 holiday sales forecast alongside seasonal hiring expectations, and within hours the numbers were circulating in summaries, newsletters and assistant answers. What determines which version of those numbers a shopper actually sees is not the press release's design or its headline. It is whether the underlying figures are published in a form a machine can parse, attribute and compare. Forecast season is therefore a useful test of a capability most retail brands have not built: machine-readable evidence.
A Forecast Only Travels If It Is Parseable
An answer engine assembling a response about holiday spending needs three things from a source: a number, a definition, and a time reference. A figure without a definition cannot be compared to another figure, and a figure without a time reference cannot be placed in a trend. Most forecast documents contain all three in prose, which means the engine has to extract them, and extraction is where attribution gets lost.
Publishing the same figures in a structured table with explicit labels solves most of the problem. It does not require new data, only a second presentation of data that already exists. The organizations that do this consistently are quoted far more often than those that publish only a PDF, not because their numbers are better, but because their numbers are usable without interpretation.
What Parseable Means in Practice
Parseable means a table with named columns, units in the header, a stated period, and a footnote defining the base. It means avoiding ranges without endpoints and avoiding adjectives in numeric fields. It also means keeping the same figure identical across the press release, the report and the landing page, because inconsistent duplicates cause engines to drop the number rather than choose between versions.
What AI Answers Quote From Holiday Data
Assistants tend to quote three types of holiday data: total spend growth, the channel split between online and store, and hiring or labor expectations. Each answers a question a shopper or an analyst actually asks. Notably absent from most answers are brand-level claims, because brand claims usually lack a comparable base. This is the asymmetry retail marketers underestimate: the category-level number gets quoted, the brand-level assertion gets skipped.
AI-referred traffic to retail sites has doubled year over year, which means the answer layer is no longer a marginal channel. When an assistant quotes a category growth figure, the brands that appear alongside that figure inherit credibility they did not earn directly. That is why publishing a defensible number is a brand-building act, not just a research function.
The Brand-Level Gap
Brands can close part of the gap by publishing their own data with a clear method. A retailer that reports its own basket composition, with sample size and period stated, creates a citable fact that no competitor can replicate. The requirement is methodological honesty: a small sample is fine if it is labeled as small, because engines penalize ambiguity far more than they penalize scale.
Three Gaps Between a PDF and a Citation
The first gap is structure. A PDF is a visual layout, not a data structure, and extraction from it is lossy. The second gap is granularity. Forecasts are often published at a level of aggregation that cannot answer the specific question a shopper asks, such as what a particular category will do. The third gap is freshness, because a forecast published in October and never updated looks stale by December even when its annual figure remains valid.
Each gap has a cheap remedy. Structure is solved by an HTML table with proper headers. Granularity is solved by publishing a category breakdown alongside the total. Freshness is solved by dating every figure and adding a short note when the underlying assumption changes. None of these require new research, and all of them increase the chance of being cited.
Why Freshness Compounds
Freshness compounds because engines weight recency when several sources carry the same figure. A brand that republishes its data with an updated timestamp, and explains what changed, accumulates citation history. A brand that publishes once and moves on is eventually treated as background rather than as a source, even if its original number was the most accurate one available.
How to Publish Numbers Machines Can Use
Start with a single canonical page per dataset, and make every other mention point to it. Give each figure a stable identifier so that citations can be traced. State the method in one paragraph directly beneath the table rather than in an appendix. Add a machine-readable block that repeats the table's key values, so that parsers do not have to infer them from prose.
The next step is consistency across surfaces. The same figure should appear identically in the press release, the investor deck, the landing page and the sales collateral. This is an operational discipline more than a technical one, and it usually requires a single owner for the number, which most organizations do not have. Assigning that owner is the highest-leverage change available.
Stable Identifiers and Versioning
Versioning matters as much as identifiers. When a forecast is revised, the old version should remain accessible with a clear note that it has been superseded, rather than being silently overwritten. Engines that encounter a revision with history can describe the change, which produces a richer citation. Engines that encounter a silent overwrite often avoid the figure entirely.
What to Check Before November
Before the peak season, audit four things. First, whether the brand's key category figures exist as structured tables anywhere on the owned domain. Second, whether those figures are consistent across the site. Third, whether each figure carries a period and a definition. Fourth, whether the brand appears in assistant answers for the ten questions its customers actually ask. That last check is the only one that measures outcome rather than readiness.
Younger shoppers research earlier and rely more on assistant-led discovery, which means the answer layer is forming impressions before a store visit ever happens. A brand that is absent from those answers is not merely losing a click; it is absent from the shortlist that the shopper brings into the store. The cost of that absence is invisible in attribution reports and very visible in conversion rates.
Meta and other platforms are now shipping merchant-facing agents that answer product questions on a brand's behalf, which raises the stakes further. When an agent speaks for a brand, the brand's own structured data becomes the source of truth or the source of error. Publishing numbers that machines can parse is no longer a search optimization tactic; it is the mechanism by which a brand controls its own description.
