If the Experiment settings panel is not open, click in the experiments controls.

The Cost to serve experiment considers only one period out of the available periods from the Periods table. The starting date of the experiment corresponds to the date specified in the Start column (in the Periods table) of the selected period, and the ending date corresponds to the date specified in the End column (in the Periods table) of the selected period.

• Solution strategy — depending on the solution type the experiment will either use the existing Network optimization result or trigger the Network optimization experiment execution to later use its result:
  • Optimization + Cost to serve — the Network optimization experiment will be executed within the Cost to serve experiment. The obtained result will not be available in the Network optimization experiment. It will be instantly used in the Cost to serve experiment, which will be started once the result is ready. Note that not all periods will be processed.
  • Cost to serve calculation only — the Cost to serve experiment will use the data from the existing Network optimization experiment. The result to work with is defined in the Result parameter. The measurements units will also be taken from the selected result.
• Result — [available, if Cost to serve calculation only is selected] select the required result from the list of existing Network optimization results. This result will be used in the Cost to serve experiment run.
• Experiment duration — the period of time that will be processed by the experiment. The list of available periods may differ depending on the selected Solution strategy:
  • Optimization + Cost to serve — the periods are taken from the Periods table.
  • Cost to serve calculation only — the periods are taken from the obtained Network optimization result.

  If the Optimization + Cost to serve option is selected, the Network optimization experiment will consider all scenario periods. The result will offer a complete set of data, but the Cost to serve experiment will use data only from the selected Experiment duration period for its run.

• Ignore straight routes — [available if Optimization + Cost to serve is selected as Solution strategy] [use it only if Straight is disabled in the Paths table] defines if the experiment considers straight routes or ignores them.

  By straight routes in this case we mean the actual routes (roads) that are treated as straight routes because their data couldn't be obtained (e.g. the road does not exist, the data on this route failed to download, etc.). Hence, a straight line connecting two objects. This parameter defines if the experiment should consider such straight routes.

• Select demand variation type — [available if Optimization + Cost to serve is selected as Solution strategy] specifies the type of acceptable demand deviation to consider during the experiment:
  • Exact demand — [selected by default] — no deviation is allowed. The down and up penalties will be applied instantly on violating the specified demand.
  • 100% — 105% — deviation of up to 5% is allowed. The up penalty will be applied only if the violation exceeds the 5% threshold.
  • 95% — 100% — deviation of up to — 5% is allowed. The down penalty will be applied only if the violation is below the — 5% threshold.
• Select search type for N best solutions — [available if Optimization + Cost to serve is selected as Solution strategy] defines the objective of the experiment run. You may find the number of solutions satisfying 1 site as well a number of sites.
  • Find N best — allows you to define the number of best solutions for the optimal number of sites. Having found the best solution, the solver blocks it, excludes its flows, and starts searching again for the second best solution. The solver continues searching until the required number of best solutions is found. In other words, the solver solves a number of separate tasks and provides the best result for each of them.
  • Solution pool — allows you to define multiple solutions to a mixed integer programming (MIP). The solver will find the required number of best solutions during a single search and return all of the possible solutions.
• Number of best solutions to find — [available if Optimization + Cost to serve is selected as Solution strategy] specify the number of solutions that you need to find.
• Optimization time limit — [available if Optimization + Cost to serve is selected as Solution strategy] sets the maximum time you would like to allot to defining one solution.
• Relative MIP gap — [available if Optimization + Cost to serve is selected as Solution strategy] sets a relative tolerance on the gap between the best found solution and the best possible solution. The solver will stop as soon as it finds the solution within the specified percent (e.g. 5%, or 0.05 when specifying the MIP gap).
• Number of threads to use — [available if Optimization + Cost to serve is selected as Solution strategy] the number of tasks that can be run in parallel.
• Problem definition type — [available if Optimization + Cost to serve is selected as Solution strategy] define what to formulate the site selection problem (sites with the Consider inclusion type) for the solver with:
  • Use Big M — [selected by default] it is a big enough constraint used in a mixed integer programming (MIP) to model if...then. If its value is too big, result might be inaccurate (e.g. the second and the third best results are better than the first one).
  • Indicator — use it if the results acquired using Big M are inaccurate.
• Finances stats unit — [available if Optimization + Cost to serve is selected as Solution strategy] the monetary unit that will be used in the statistics.
• Product stats unit — [available if Optimization + Cost to serve is selected as Solution strategy] the product measurement unit (also includes units defined in the Measurement Units table), in which the Flows Amount will be displayed.
• Distance stats unit — [available if Optimization + Cost to serve is selected as Solution strategy] the distance measurement unit that will be used in the statistics.

How can we improve this article?

Send