In the PRIMO project, the focus has been on travel cost variation for long-distance trips. All modes – car, train, air, and bus – have been considered, even though it has been diffcult to find data for long distance bus. The PRIMO project has consisted of three parts: First, descriptive data has been explored, which showed considerable contemporary travel cost variation especially for air travel, but also for train and car travel (Deliverable 1). Second, possible methodologies for how cost variation can be considered to a larger extent in demand models have been explored (Deliverable 2). Third, recommendations for short- and long-term improvements of Sampers have been given (this Deliverable 3). One of the most important findings within the Primo project is that business trip reimbursements and their variation could have substantial effects on model results. Business trip reimbursements are not accounted for at all in the current Sampers long-distance model. There are at least three reasons why this shortcoming is a candidate for some of the problems with the long-distance model discussed in this deliverable: first, the level of reimbursement and it variation is quite large; second, not modelling this cost reduction leads to it being implicitly accounted for in the cost parameter common to all modes; third, business trips is a trip purpose with high values of travel time which will have large effects on model results. The recommendations set out in this deliverable would benefit from a clearer context regarding the development of Sampers long-distance model. Therefore, a study set out to investigate and prioritize among the other problem issues listed in this deliverable would be beneficial. Some of the choices that need to be made about the development of the long-distance model depend on the policy to be evaluated. For example, if high-speed rail is the main policy to be evaluated, this will impact the balance between segmentation and choice of fare level described above; for this policy it may well be appropriate to have both fare-related segmentation and a choice model between fast/expensive high-speed trains and slower but cheaper conventional trains.