Spill and Recapture¶
spill¶
This table shows what passengers chose when compared with the offer they would have preferred without availability restrictions. It groups the outcomes by the carrier and fare class of that first-choice offer.
Passengers may buy their preferred offer, choose another fare or itinerary, move to a different carrier, or make no purchase. The columns count those outcomes, helping you see where a carrier keeps business and where it loses it. The preferred offer is determined by the choice model’s generalized cost, which accounts for the factors passengers consider; it is not necessarily the offer with the lowest ticket price.
Access this table as summary.spill.
Index¶
The rows have a simple numbered index. The fields that identify each observation are included as ordinary columns in the table below.
Columns¶
Field |
Type |
Meaning |
|---|---|---|
|
string |
Carrier of the first-choice offer. |
|
string |
First-choice fare class label (the collector uses |
|
integer |
Number of choice observations; counts events, not party members. |
|
integer |
Passengers choosing the same path and fare class. |
|
integer |
Passengers choosing the same path and a different fare class. |
|
integer |
Passengers choosing a different path, same fare class, same carrier. |
|
integer |
Passengers choosing a different path, same fare class, different carrier. |
|
integer |
Passengers choosing a different path, different fare class, same carrier. |
|
integer |
Passengers choosing a different path, different fare class, different carrier. |
|
integer |
Passengers making no purchase. |
In these column names, P means path, F means fare class, and A means airline or carrier. For
example, sameP_diffF counts passengers who stayed on their preferred itinerary but bought a
different fare class.
The outcome columns count passengers, while n_obs counts choice events. A party of several
passengers can make one choice, so the passenger counts need not add up to n_obs. These are the
simulation’s cumulative spill counters; the summary does not apply an additional filter to remove
burned samples.
After a MultiSimulation run, the counts from each trial are added together for each carrier and
first-choice class.
Activation¶
PassengerSim fills this table during a normal simulation run, without database recording or an entry
in outputs.reports. For meaningful results,
however, you need to set
simulation_controls.save_all_offers to
True before running. It is False by default, and without it closed offers are not kept, so they
cannot be counted as first choices. The table then reports observations as sameP_sameF, with the
other outcome columns, including no_go, left at zero. To turn the option on:
simulation_controls:
save_all_offers: true
Turning this option on can change the simulation’s results, so compare runs that use the same setting. There is no switch that turns off collecting this table. If you only want to leave it out of the returned summary, you can use selecting extracted tables to choose which tables to include.