New product forecasting checklist: data and assumptions to validate

Merchandising teams preparing demand forecasts for new products face a familiar challenge: which historical analogues to select, how much weight to assign each product attribute, and which manual adjustments to apply. A systematic checklist to validate these inputs before each season can reduce forecast error and align cross-functional teams on underlying assumptions. This article outlines the data points and decisions that warrant review, drawing on planning practices observed across the fashion industry in 2025 and 2026.

Which analogue products should be excluded from the training set?

The foundation of any product forecast is the set of historical analogues used to predict demand. Not all past products provide reliable signals. Analogues should be excluded if they experienced supply chain disruptions that artificially constrained sales, such as delayed shipments or stockouts that prevented full demand realisation. Products sold predominantly at markdown also distort demand signals, as the price elasticity observed during clearance rarely reflects full-price customer response.

External events that created abnormal demand should also trigger exclusion. Unseasonable weather in key markets can inflate or suppress sales for seasonal categories. Products launched during store closures or major competitive events may show atypical performance. The goal is to isolate analogues where observed sales reflect true underlying demand rather than supply constraints or external shocks.

How often should attribute weightings be recalibrated?

Product attributes such as fabric composition, silhouette, price point, and sustainability claims each carry a weighting in forecast models. These weightings reflect how much each characteristic influences customer choice. Consumer preferences shift over time, and weightings that were accurate six months ago may no longer hold. Quarterly recalibration is a practical cadence for most categories, with more frequent review for fast-moving segments like streetwear or activewear.

Price sensitivity has increased across income segments in recent seasons, particularly for discretionary categories. Sustainability attributes show mixed influence: some customer segments assign high value to recycled materials or verified supply chain transparency, while others prioritise price and fit above all else. Testing attribute weights against recent sell-through data helps identify which characteristics currently drive demand and which have diminished in importance.

What adjustments introduce the most forecast bias?

Human adjustments applied on top of statistical forecasts often reflect intuition about newness, trend momentum, or competitive activity. These adjustments can improve accuracy when they incorporate information not captured in historical data, but they can also introduce systematic bias. Common sources of bias include overestimating the halo effect of celebrity endorsements, underestimating cannibalisation from existing assortment, and applying blanket uplifts for trend categories without segmenting by customer cohort.

Every manual adjustment should be documented with a clear rationale and a named owner. This creates accountability and allows post-season analysis to identify which types of adjustments improved accuracy and which consistently erred in one direction. Adjustments for competitive launches require particularly careful validation: a competitor entering a category may shift share, but total category demand may not expand as much as initial excitement suggests.

Which data quality issues undermine forecast accuracy?

Forecast models depend on clean, consistent input data. Product attribute data often contains inconsistencies: the same fabric composition recorded differently across seasons, colour names that vary between buyers, or size scales that have changed over time. Standardising taxonomy and enforcing data entry validation at the point of product creation prevents these inconsistencies from degrading model performance.

Sales data requires equal scrutiny. Returns and cancellations should be netted out to reflect true demand. Store transfers can inflate sales in one location while deflating another, distorting location-based forecasts. Promotional sales should be flagged with the discount depth and duration, as this context is essential for separating price-driven demand from underlying preference.

How should cross-functional teams validate assumptions together?

Forecast accuracy improves when merchandising, buying, planning, and allocation teams review assumptions together before committing to buy quantities. A structured validation session should cover analogue selection rationale, key attribute weights, and any manual adjustments. Each function brings a different perspective: buyers understand product differentiation and supplier constraints, planners see financial targets and inventory flow, allocators know location-specific demand patterns.

This cross-functional review surfaces hidden assumptions that might otherwise go unchallenged. A buyer may flag that a proposed analogue product had a supplier quality issue that suppressed reorders. A planner may note that the forecast period includes a promotional event not reflected in the model. An allocator may identify regional preference differences that suggest segmenting the forecast by market. Documenting these discussions creates an audit trail for post-season learning.

What should the pre-forecast validation checklist include?

A practical pre-forecast checklist consolidates the validation steps into a repeatable process. The checklist should be completed before final buy quantities are confirmed and signed off by representatives from merchandising, planning, and buying.

• Confirm all analogue products meet inclusion criteria: no major stockouts, no abnormal promotions, no external disruptions during their selling period.
• Review attribute weightings against the most recent season's sell-through data and adjust if consumer preferences have shifted.
• Document every manual adjustment with rationale, magnitude, and owner, and flag any adjustment exceeding 20 per cent of the statistical forecast for additional review.
• Validate product attribute data for consistency: standardised fabric names, colour codes, and size scales across all new products.
• Check sales data for completeness: returns netted out, transfers reconciled, promotional flags applied with discount depth.
• Schedule cross-functional validation session at least two weeks before buy commitment deadline to allow time for model adjustments.
• Archive all assumptions and input data to enable post-season accuracy analysis and continuous improvement.

Forecasting new product demand remains as much art as science, but a disciplined validation process reduces the risk of preventable errors. By auditing analogue selection, recalibrating attribute weights, challenging manual adjustments, and ensuring data quality, merchandising teams can improve forecast accuracy and build confidence across the organisation in the demand signals that drive inventory investment.