How to use early sell-in data to trigger in-season production adjustments
Analysing the first 20 per cent of wholesale orders with statistical confidence intervals and ERP integration allows brands to adjust production volumes before fabric cutting begins.

KEY TAKEAWAYS Summary by the editors
- Early sell-in data from wholesale orders provides the first real demand signal of a season before retail sell-through data becomes available.
- Confidence intervals quantify forecast uncertainty and enable brands to plan production at conservative thresholds when order books are incomplete.
- Real-time ERP integration synchronises wholesale order data with production planning systems to accelerate decision-making before fabric is cut.
- Brands using this workflow have reduced overproduction whilst maintaining adequate stock for strong performers during the selling season.
Wholesale order books offer fashion brands an early-warning system for production planning, but most wait until sell-through data arrives from retail partners weeks or months into the season. The workflow described here analyses the first 20 per cent of wholesale bookings to generate statistically grounded production adjustments before factories begin cutting fabric, reducing both overproduction risk and stockout exposure.
Why does early sell-in data matter for production decisions?
Early sell-in data from key accounts is the first real demand signal of a season, arriving before retail sell-through becomes available. Sell-in is what a brand ships to a retailer, whilst sell-through is how much of that stock the retailer sold to shoppers, and strong sell-in with weak sell-through usually leads to smaller reorders, markdown requests or returns the next season. Analysing this initial data while the order window remains open allows merchandisers to adjust production commitments for styles showing weak demand and scale up volumes for early winners.
Wholesale accounted for 57 per cent of brands' total business in a 2025 market survey, reinforcing the channel's importance as a demand signal source. The transition to smaller collections has also changed the stakes: the share of brands producing full new collections has declined gradually, whilst those focusing on smaller capsule collections of new styles has risen from 19 per cent to 37 per cent over the past two years. With fewer styles carrying more revenue responsibility, early read accuracy becomes operationally critical.
How do you structure the 20 per cent threshold analysis?
The 20 per cent threshold represents the point at which order book patterns begin to stabilise enough for directional decisions without waiting for complete data. Set a calendar checkpoint when approximately 20 per cent of expected wholesale accounts have submitted orders, typically four to six weeks into the order window depending on your seasonal calendar and buyer behaviour.
At this checkpoint, calculate style-level order penetration: total units booked divided by the original production plan. Styles above 25 per cent of plan at the 20 per cent account threshold signal strong early demand. Styles below 10 per cent at the same checkpoint indicate weak reception and become candidates for production reduction. The middle range requires further monitoring before adjustment.
Segment analysis by account type and geography improves signal quality. If your strongest accounts or highest-volume regions show consistent patterns across multiple styles, those signals carry more weight than diffuse bookings from smaller or newer accounts. Track order velocity as well as absolute volume: styles that arrive in the first 10 per cent of accounts and continue in the second 10 per cent demonstrate momentum.
What statistical methods quantify uncertainty in partial order books?
Confidence intervals provide a mathematically grounded way to express forecast uncertainty when working with incomplete data. Probabilistic forecasting not only provides forecasts but also accompanies each with a probability and a confidence interval, offering an uncertainty estimate for every prediction, and this method acknowledges and quantifies the inherent uncertainty in forecasting, allowing businesses to make more informed decisions under varying conditions.
Analysing the spread of forecast distribution enables the construction of prediction intervals that reflect the confidence associated with each forecast, and these intervals offer valuable insights for operational decision-making tasks, such as inventory planning and production risk management. Bootstrapped predictions used 100 resampling iterations per product, and 90 per cent prediction intervals were derived from the resulting empirical distributions.
For fashion applications, bootstrap resampling works by taking your historical order book data from previous seasons, resampling it with replacement to create multiple simulated scenarios, and calculating demand forecasts for each scenario. The range of those forecasts becomes your confidence interval. The forecasting model outputs SKU-level demand predictions in a format that the production scheduling system can consume, typically as weekly production quantity recommendations with confidence intervals.
A practical approach: if your 20 per cent order book data suggests 5,000 units for a style but your 90 per cent confidence interval spans 3,500 to 7,200 units, plan initial production at the lower bound and reserve factory capacity for a second production run if bookings continue strong. For true new-to-world products, manufacturers should plan initial production at 40 to 50 per cent of the AI's upper confidence bound and use the first two weeks of sell-through data to recalibrate the remaining production allocation.
How does ERP integration accelerate the workflow?
ERP-connected wholesale platforms eliminate data entry duplication and ensure accurate inventory visibility during order placement. When wholesale orders flow directly into your ERP system, production planning teams can query real-time booking data without waiting for manual reports. ERP integration creates automatic data flow between your ordering portal and back-end systems, orders sync instantly without manual re-entry, inventory levels update in real time, and fulfilment begins faster.
Configure your ERP to generate production planning alerts when style-level bookings cross predefined thresholds. For example, set an alert when any style reaches 30 per cent of planned units at the 20 per cent account checkpoint, flagging it for immediate production increase review. Similarly, trigger alerts for styles below 8 per cent booking at the same checkpoint, prompting production reduction analysis.
Link wholesale order data to production scheduling modules within the ERP. A unit can be traced from the material on a vendor PO, through cutting and sewing, into a warehouse or 3PL, out to a wholesale order, dropship, or DTC shipment, and back again if it is returned, with landed cost attached the whole way. This end-to-end visibility allows planning teams to calculate the financial impact of production adjustments before committing to changes.
What production adjustments can you make at this stage?
The adjustments available depend on your supply chain lead times and contractual commitments with manufacturing partners. The most flexible lever is fabric procurement: if you have not yet ordered fabric for the full production run, reduce or increase fabric orders to match revised demand forecasts. Many brands split fabric orders into tranches, committing to 60 per cent upfront and reserving decisions on the remaining 40 per cent until early order book data arrives.
For fabric already ordered, adjust cut-and-sew work orders. Cancel or reduce cutting for weak styles before fabric enters the production line. Scale up cutting for strong styles if you have fabric inventory or can secure additional yardage quickly. Factories start with tight, test-sized batches, and if a jacket moves fast in Paris and London in week one, production scales up immediately, whilst if it sits in Seoul, it gets cut.
Colour and size mix adjustments offer another lever. Early order data may reveal that buyers favour certain colours or size runs. Shift production allocation within a style to match these preferences, maintaining total unit count but improving sell-through probability by aligning production with demand signals.
| Production lever | Decision window | Typical lead time required | Risk mitigation |
|---|---|---|---|
| Fabric procurement reduction | Before fabric order placement | 8 to 12 weeks before production | Eliminates excess material cost |
| Fabric procurement increase | Before fabric order placement | 8 to 12 weeks before production | Secures material for strong styles |
| Cut-and-sew volume reduction | After fabric order, before cutting | 4 to 6 weeks before shipment | Reduces finished goods overstock |
| Cut-and-sew volume increase | After fabric order, before cutting | 4 to 6 weeks before shipment | Prevents stockouts on winners |
| Colour and size mix shift | During cutting | 2 to 4 weeks before shipment | Improves sell-through within plan |
How do you validate results and refine the workflow?
Track two metrics to validate this workflow: forecast accuracy at the 20 per cent checkpoint and final-season sell-through rates by style. Calculate forecast accuracy by comparing your production decisions at 20 per cent order completion to final wholesale bookings. If your early forecasts consistently undershoot or overshoot by more than 15 per cent, recalibrate your confidence intervals or threshold triggers.
Monitor sell-through rates for styles where you made production adjustments versus styles where you held to the original plan. A good sell-through rate for apparel ranges from 65 to 85 per cent, with rates under 65 per cent needing review and under 40 per cent considered critical. Styles adjusted based on early order data should demonstrate higher full-price sell-through and lower markdown rates than the control group.
Document decision rules and thresholds in a playbook that merchandising and planning teams can reference each season. Specify which account segments trigger production reviews, what confidence interval percentiles guide production decisions, and how ERP alerts route to responsible teams. Brands expanded data-driven merchandising systems that track sell-through rates and regional demand changes in real time, allowing faster production adjustments. This workflow becomes more effective as you accumulate seasonal data and refine your statistical models.
Frequently asked questions
When in the wholesale calendar should you analyse the first 20 per cent of orders?
Analyse orders when approximately 20 per cent of expected wholesale accounts have submitted bookings, typically four to six weeks into the order window. This timing balances the need for early signal detection against statistical reliability, as patterns begin to stabilise once key accounts have placed orders.
What confidence interval threshold works for production planning decisions?
A 90 per cent confidence interval provides a practical balance between statistical rigour and operational utility. Plan initial production at the lower bound of this interval and reserve factory capacity for additional runs if bookings continue strong, particularly for new or unproven styles where demand uncertainty is highest.
Which ERP modules must integrate with wholesale order data for this workflow?
Integrate wholesale order entry with production planning, materials management, and inventory modules. This connection allows production teams to query real-time booking data, generate alerts when styles cross threshold triggers, and calculate the financial impact of production adjustments before committing to changes.
Sources
- just-style: Wholesale resurgence, AI to lead fashion wholesale in 2026
- Tommaso Maria Ricci: AI systems that correlate early wholesale performance with in-season sell-through
- Springer: Deep learning for new fashion product demand prediction
- AIMS360: Sell-Through Rate Calculator, Formula and Apparel Benchmarks
- Reports and Data: Clothing Market brands expanded data-driven merchandising systems
