{"id":9669,"date":"2026-03-13T14:36:00","date_gmt":"2026-03-13T14:36:00","guid":{"rendered":"https:\/\/pros.com\/learn\/?p=9669"},"modified":"2026-03-13T15:40:38","modified_gmt":"2026-03-13T15:40:38","slug":"airlines-keep-demand-forecasts-accurate-turbulent-times","status":"publish","type":"post","link":"https:\/\/pros.com\/learn\/blog\/airlines-keep-demand-forecasts-accurate-turbulent-times\/","title":{"rendered":"How Airlines Keep Demand Forecasts Accurate in Turbulent Times"},"content":{"rendered":"
<\/div>

Geopolitical conflicts. Pandemic waves. Volcanic ash clouds. Severe weather. When large\u2011scale disruptions hit, airline schedules can change overnight. Flights are cancelled, passengers rebook in droves, and booking patterns stop resembling anything close to normal.<\/p>\n

If revenue management (RM) allows these crisis\u2011era booking patterns to flow straight into the forecast, the system learns from data that does not reflect genuine demand. The effects can linger long after operations stabilize, blurring demand predictions and negatively impacting RM decisions.<\/p>\n

This blog explores why disruptions create \u201cbad data\u201d for RM and how airlines can prevent crisis\u2011driven patterns from distorting future forecasts using advanced tools in their systems. With a practical and reliable disruption\u2011response playbook, RM analysts can keep forecasts accurate through uncertainty and position their networks to rebound quickly when demand normalizes.<\/p>\n

Why Disruptions Create \u201cBad Data\u201d for Revenue Management<\/h2>\n

Two things are typically observed during major disruptions: flights get cancelled at scale, or booking patterns shift dramatically. Even if the disruption itself lasts only days or weeks, its effects on demand can linger much longer, requiring revenue management teams to actively shape how these abnormalities influence future forecasts.<\/p>\n

When mass cancellations occur, they generate rebookings that spill into future departure dates. If left unmanaged, the system may interpret these shifts as genuine demand rather than operational intervention. Similarly, zero demand that appears simply because service wasn\u2019t operating can create misleading signals.<\/p>\n

Disruptions can also surface in the form of abnormal demand, even when flights continue operating. Bookings may plunge for weeks due to suppressed demand or intentional restrictions, and recovery may be slow and uneven. These dips, followed by partial rebounds, do not represent true steady-state demand. If the dips are allowed to flow unfiltered into the model, the RM system may overfit to crisis conditions and carry distorted expectations well into the recovery period, leading to poor RM decision support.<\/p>\n

RM analysts need forecast-protection tools that automatically recognize the lack of scheduled service, filter out rebooking effects, and protect forecasts from absorbing irregular patterns to ensure confidence once the crisis passes.<\/p>\n

Reliable Forecast\u2011Protection Tools to Manage Crisis\u2011Era Data<\/h2>\n

When disruptions distort booking patterns due to widespread cancellations, suppressed demand, or uneven early-recovery behavior, RM analysts need to take a proactive role in shaping what the forecasting engine learns. Modern RM practice offers a set of tools and approaches that help analysts keep forecasts grounded in true demand rather than crisis noise, so your demand expectations remain stable, reliable, and aligned with real market behavior.<\/p>\n

Manage Data Inputs with Censoring<\/h3>\n

Censoring is the primary method for excluding periods that shouldn\u2019t influence the forecast model. This method allows teams to cleanly remove weeks or months affected by major incidents, service suspensions, or severe demand drops. By treating these observations as missing rather than legitimate zeros or valid signals, the system avoids embedding patterns that don\u2019t represent steady-state demand.<\/p>\n

RM analysts should apply censoring when crisis-era observations don\u2019t reflect the demand they expect to return. These observations can include periods of extended cancellations or minimal service, unusually low bookings, and early recovery phases when demand is still stabilizing. There is no fixed limit on how long censoring should remain in place; its duration should align with the commercial reality of the market. Many airlines choose to censor multi-month windows and then reassess as conditions evolve, ensuring the forecast continues to reflect healthy, representative demand patterns.<\/p>\n

\"Censoring<\/p>\n

Censoring in PROS Revenue Management<\/em><\/p>\n

Adjust Data with Influences<\/h3>\n

Once distorted periods are set aside, analysts can apply influences to finetune the remaining data, adjusting forecasts up or down when market recovery is expected or when booking behavior is evolving.<\/p>\n

\"Influences<\/p>\n

Influences in PROS Revenue Management<\/em><\/p>\n

Being deliberate and timely with these interventions ensures the system learns from healthy patterns, helps avoid long-lasting distortions, and supports more accurate pricing and inventory decisions as conditions stabilize.<\/p>\n

A Few Real-World Examples<\/h2>\n

Here are a few practical scenarios to demonstrate how these tools work in day\u2011to\u2011day airline operations:<\/p>\n