{ "id": 5690, "title": "Questions about search results", "url": "/competitions/expedia-personalized-sort/discussion/5690", "author_name": "", "post_date": "2013-09-07T17:06:46.307Z", "votes": null, "comment_count": 3, "views": 2466, "content": "

It seems that some search results are missing for each individual search? See for example srch_id = 6. There are 5 results for this search, and the given hotel positions are[ 4, 6, 2, 1, 3]. Is there a reason why the hotel in position #5 is missing?

\n

 

\n

Additionally, can you shed some light on why some searches have more results than others? For instance, srch_id=6 has 5 results, while srch_id=1 has 28 results. Is there a default number of results shown, and more results means that the visitor clicked on some "Next" button?

\n

 

\n

Thank you very much; this is a really interesting competition :-)

", "messages": [ { "id": "30434", "postDate": "09/07/2013 17:06:46", "content": "

It seems that some search results are missing for each individual search? See for example srch_id = 6. There are 5 results for this search, and the given hotel positions are[ 4, 6, 2, 1, 3]. Is there a reason why the hotel in position #5 is missing?

\n

 

\n

Additionally, can you shed some light on why some searches have more results than others? For instance, srch_id=6 has 5 results, while srch_id=1 has 28 results. Is there a default number of results shown, and more results means that the visitor clicked on some "Next" button?

\n

 

\n

Thank you very much; this is a really interesting competition :-)

", "rawMarkdown": "", "votes": null }, { "id": "30441", "postDate": "09/07/2013 18:52:14", "content": "

Hi,

\n

1. In most cases a missing item corresponds to an "opaque inventory item" (http://en.wikipedia.org/wiki/Opaque_travel_inventory) which in our context means that during search we do not know the corresponding hotel. They usually appear close to positions 5, 10, 15, etc. (see e.g. http://www.expedia.com/Hotel-Search#destination=London+%28and+vicinity%29%2C+England%2C+United+Kingdom&startDate=12%2F06%2F2013&endDate=12%2F08%2F2013&adults=2&star=0)

\n

2. The cardinalities of searches differ because: (i) some markets have more hotels than others, (ii) occasionally markets get "compressed" for specific check-in and check-out dates which means that only a limited number hotels have free rooms (and only those are displayed), (iii) filters are used (e.g. only 5-stars hotels, or hotels in a specific neighborhood).

\n

 

\n

3. The released dataset only contains page 1 impressions. In other words, clicks on a "Next" button are not considered here.

\n

 

\n

4. You are welcome :-)

\n

 

\n

Adam

\n

 

\n

 

\n

 

\n

 

", "rawMarkdown": "", "votes": null }, { "id": "30458", "postDate": "09/07/2013 22:24:19", "content": "

Thanks a lot for your answers.

\n

I have another one regarding the "comp" variables in the dataset.  There seem to be some inconsistencies in these. For example, take {srch_id=1, prop_id=30184}. For this row, comp3_rate = 0 (so Expedia has the same price as Competitor 3), but comp3_rate_percent_diff = 7 (so there is a 7% difference in price with Competitor 3). Can you explain this case?

\n

There are also many cases where compX_rate=0, but then compX_rate_percent_diff=NULL. So it's not clear whether there is competitive data or not for this competitor.

\n

Thanks again!

", "rawMarkdown": "", "votes": null }, { "id": "30484", "postDate": "09/08/2013 15:02:44", "content": "

Here are some comments on the "comp" variables:

\n

1. The data which was used to construct the "comp" variables is volatile and not always correct.

\n

2. This data is also sparse which means that depending on the competitor, hotel, check-in and check-out dates, the data is gathered more or less frequently.

\n

3. compx_rate = 0 if the difference between rates is small (i.e. the rates do not necessarily have to be the same).

\n

4. As a result of 2., often we cannot match exactly hotel, search, check-in and check-out dates. In these cases, we match over aggregated scores. The inconsistencies you described are most probably a result of this aggregation and/or point 3.

\n

5. "compX_rate=0, but then compX_rate_percent_diff=NULL" is correct; in such cases you can assume the rates are similar.

\n

 

\n

Cheers,

\n

Adam

\n

 

", "rawMarkdown": "", "votes": null } ], "comments": [ { "id": 30441, "author_name": "adamwoz", "author_url": "", "post_date": "09/07/2013 18:52:14", "content": "

Hi,

\n

1. In most cases a missing item corresponds to an "opaque inventory item" (http://en.wikipedia.org/wiki/Opaque_travel_inventory) which in our context means that during search we do not know the corresponding hotel. They usually appear close to positions 5, 10, 15, etc. (see e.g. http://www.expedia.com/Hotel-Search#destination=London+%28and+vicinity%29%2C+England%2C+United+Kingdom&startDate=12%2F06%2F2013&endDate=12%2F08%2F2013&adults=2&star=0)

\n

2. The cardinalities of searches differ because: (i) some markets have more hotels than others, (ii) occasionally markets get "compressed" for specific check-in and check-out dates which means that only a limited number hotels have free rooms (and only those are displayed), (iii) filters are used (e.g. only 5-stars hotels, or hotels in a specific neighborhood).

\n

 

\n

3. The released dataset only contains page 1 impressions. In other words, clicks on a "Next" button are not considered here.

\n

 

\n

4. You are welcome :-)

\n

 

\n

Adam

\n

 

\n

 

\n

 

\n

 

", "votes": null, "replies": [] }, { "id": 30458, "author_name": "martin22", "author_url": "", "post_date": "09/07/2013 22:24:19", "content": "

Thanks a lot for your answers.

\n

I have another one regarding the "comp" variables in the dataset.  There seem to be some inconsistencies in these. For example, take {srch_id=1, prop_id=30184}. For this row, comp3_rate = 0 (so Expedia has the same price as Competitor 3), but comp3_rate_percent_diff = 7 (so there is a 7% difference in price with Competitor 3). Can you explain this case?

\n

There are also many cases where compX_rate=0, but then compX_rate_percent_diff=NULL. So it's not clear whether there is competitive data or not for this competitor.

\n

Thanks again!

", "votes": null, "replies": [] }, { "id": 30484, "author_name": "adamwoz", "author_url": "", "post_date": "09/08/2013 15:02:44", "content": "

Here are some comments on the "comp" variables:

\n

1. The data which was used to construct the "comp" variables is volatile and not always correct.

\n

2. This data is also sparse which means that depending on the competitor, hotel, check-in and check-out dates, the data is gathered more or less frequently.

\n

3. compx_rate = 0 if the difference between rates is small (i.e. the rates do not necessarily have to be the same).

\n

4. As a result of 2., often we cannot match exactly hotel, search, check-in and check-out dates. In these cases, we match over aggregated scores. The inconsistencies you described are most probably a result of this aggregation and/or point 3.

\n

5. "compX_rate=0, but then compX_rate_percent_diff=NULL" is correct; in such cases you can assume the rates are similar.

\n

 

\n

Cheers,

\n

Adam

\n

 

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "30434": "", "30441": "", "30458": "", "30484": "" }, "source": "meta" }