New Insights in the Customer’s Resistance to Interchange Between Trains

New Insights in the Customer’s Resistance to Interchange Between Trains


Bart De Keizer, NS, Marco Kouwenhoven, Significance, Freek Hofker, ProRail


Improved analysis of existing SP data shows customer resistance to interchanges is highly differentiated and underestimated. A comparison with actual passenger data shows these effects are real and important to include in demand forecasts.


Train passengers experience an interchange between trains in their trip as a nuisance. To model this resistance, a penalty is usually added in the calculation of the perceived journey time. For many years, NS and ProRail used a fixed penalty of 10 minutes per interchange for rail trips in the Netherlands. This penalty was based on expert-judgement.

A stated preference survey in 2011 [De Keizer et al, 2012, customer resistance to interchanges, ETC Glasgow] demonstrates that customers experience a much higher penalty than 10 minutes. The penalty also strongly differs with characteristics of their transfer, like transfer time, frequency of the connecting service and whether the transfer is cross platform or not.

Recently, the SP data have been re-analyzed, based on the recommendations of an audit on the 2011 work. This new analysis shows that the average penalty is 24 minutes, which is more than twice as high as the current value. However, under certain optimal circumstances, the penalty can be lower than 10 minutes.

The outcomes of the new analysis have been compared with real-world data. This comparison showed that using the new values led to a much better fit with reality. For example, growth rates of various direct connections to Schiphol Airport were 30-100% higher in reality than had been predicted with the 10 minutes penalty . Applying the new differentiated penalty led to a forecast with a maximum deviation of only 15% compared with the real-world figures . Therefore, NS and ProRail have decided to implement the results of this study into their standard calculations of perceived journey time that are used for demand forecasts.

This paper will discuss the re-analysis and compare the findings with earlier results and findings from international literature. It will also show the comparison between the model outcomes with the real-world observations and discuss its application in demand forecasting models.


Association for European Transport