The business runs as a sequence of collection launches, media works mainly in the first window of each one, and budgets have been reduced enough that every allocation now matters. Spend commits per collection before any sell-through data exists, while the read that would say whether the last one worked arrives many months later, well after the next launch decision has been made. Every allocation becomes an irreversible bet, graded afterward against numbers nobody could see when the call was made.
A pre-launch read built from the brand's own history of prior collections and categories informs the bet before it is placed rather than judging it only after. The evidence moves to before the commitment, where it can still change the decision. The grading stops being on bets nobody had a way to inform.
Every planning cycle someone asks what the demand looks like for each product category, and the answer comes from forecasting search volume by hand, because no function in the business actually does this formally. The category split that comes out of that process is treated as fixed once it is written down, even though it is movable and has no justified data behind it. Now external retail-media advertisers are asking the same question of those category splits, and intuition gets defended as if it were analysis to a second audience with its own money on the line.
A category demand forecast built from patterns in the business's own sales and media history replaces one assembled by hand under deadline. A split stands ready to defend as reasoned rather than felt. And one answer works for internal planning and external retail-media conversations alike.
Category demand forecasting is estimating how much revenue a product category or collection will generate before it launches, using patterns from comparable past launches rather than waiting for sell-through data. The output feeds a category-level budget split, made before spend commits rather than justified after the fact.
By comparing the new category or collection against past launches with similar characteristics, using the sales and media pattern those launches actually produced. The read draws on the business's own history, comparable past launches and categories in its own data, rather than on a number set by feel.
How much revenue a category is expected to generate, based on comparable history, is the forecast. How that expected revenue translates into a media spend split across categories is the budget, and the two are related but distinct decisions: a category can be forecast to perform well without automatically justifying a large budget increase.
Retail and publishing brands whose business runs as a sequence of category or collection launches, where budget has to be committed before any sales data exists for that specific launch. It is most useful where enough comparable prior launches exist to build a credible pattern, rather than for a genuinely first-of-its-kind category.
When the category or collection is genuinely new with no comparable history to draw on, when the decision needed is at SKU or individual-product level rather than category level, or when the forecast is being used to set a live price rather than a media budget. In those cases the honest output is a wide range or a recommendation to wait.