Method, medium, and system for generating and ordering item offer themes
Generating themes for different item offers is described. An item listing system receives a request for a target item and generates themes for the target item by grouping offers based on their properties. The item listing system then determines a display order for the themes based on user behavior data. The item listing system then communicates the themes and display order to a client device from which the request was received, causing the client device to display an interface including at least a subset of the themes, arranged according to the display order. Themes including offers determined to be more appealing to the user of the client device are displayed more prominently relative to other themes. The item listing system is further configured to dynamically modify the display order in real-time based on offer changes, such that the interface continuously provides correct information describing available offers for the target item.
1. A method comprising:
receiving, from a client device, a request for a particular target item listed by a listing platform;
generating themes for the particular target item, wherein each of the themes includes only one item and multiple offers for the one item that share at least one common property;
caching historic theme scores, wherein the historic theme scores are generated based on only previous user interactions with the one item on the listing platform;
ranking the themes based on the cached historic theme scores and real-time inventory and price data associated with only the offers for the one item; and
causing display of the themes in a user interface of the listing platform according to the ranking.
2. The method of claim 1 , wherein a first machine learning model generates the historic theme scores based on the previous user interactions with the one item on the listing platform.
3. The method of claim 2 , wherein the first machine learning model generates the historic theme scores offline.
4. The method of claim 1 , wherein a second machine learning model ranks the themes based on the cached historic theme scores and the real-time inventory and price data associated with the offers for the one item.
5. The method of claim 1 , wherein the causing display includes determining a display order for a top ranked subset of the themes and arranging the top ranked subset of the themes in the user interface according to the display order.
6. The method of claim 5 , further comprising:
dynamically updating the ranking based on a change in the real-time inventory and price data associated with the offers for the one item; and
causing display of a new top ranked subset of the themes in the user interface according to the updated ranking, the new top ranked subset including at least one theme that was not included in the top ranked subset.
7. The method of claim 5 , further comprising:
dynamically updating the ranking based on a change in the real-time inventory and price data associated with the offers for the one item; and
determining a new display order for the top ranked subset of the themes based on the updated ranking and arranging the top ranked subset of the themes according to the new display order.
8. The method of claim 5 , wherein the causing display includes:
automatically causing display of a top ranked theme in the user interface along with one or more offers that are within the top ranked theme; and
causing display of other themes of the top ranked subset in the user interface without offers that are within the other themes.
9. The method of claim 5 , wherein the top ranked subset of the themes includes at least two of: a new condition theme, a standard shipping theme, a used condition theme, or a fastest available shipping theme.
10. A client device comprising:
a processing system; and
a computer-readable storage medium storing instructions that, responsive to execution by the client device, cause the processing system to:
communicate a request for a particular target item listed by a listing platform;
receive a ranking of themes generated for the particular target item, wherein each of the themes includes only one item and multiple offers for the one item that share at least one common property, the themes having been ranked based on real-time inventory data and real-time price data associated with only the offers for the one item and cached historic theme scores, wherein the cached historic theme scores are generated based on only previous user interactions with the one item on the listing platform; and
display the themes in a user interface of the listing platform according to the ranking.
11. The client device of claim 10 , wherein a first machine learning model generates the cached historic theme scores offline based on the previous user interactions with the one item on the listing platform.
12. The client device of claim 10 , wherein a second machine learning model ranks the themes based on the cached historic theme scores, the real-time inventory data, and the real-time price data associated with the offers for the one item.
13. The client device of claim 10 , wherein, to display the themes, the processing system is configured to receive a display order for a top ranked subset of the themes and arrange the top ranked subset of the themes in the user interface according to the display order.
14. The client device of claim 13 , wherein, to arrange the top ranked subset of the themes, the processing system is configured to display the top ranked subset of the themes in the user interface from left to right according to the display order.
15. The client device of claim 13 , wherein, to display the themes, the processing system is configured to:
automatically display a top ranked theme in the user interface along with one or more offers that are within the top ranked theme; and
display other themes of the top ranked subset in the user interface without offers that are within the other themes.
16. One or more non-transitory computer-readable storage media having instructions stored there on, that responsive to execution by a processor, cause the processor to perform operations comprising:
receiving a request for a particular target item listed by a listing platform;
generating themes for the particular target item, wherein each of the themes includes only one item and multiple offers for the one item that share at least one common property;
ranking the themes based on real-time inventory data and real-time price data associated with only the offers for the one item and cached historic theme scores, the cached historic theme scores generated based on only previous user interactions with the one item on the listing platform; and
causing display of the themes in a user interface of the listing platform according to the ranking.
17. The one or more non-transitory computer-readable storage media of claim 16 , wherein the ranking the themes includes ranking the themes in two phases, wherein a first phase comprises generating the historic theme scores offline and caching the historic theme scores and a second phase comprises ranking the themes based on the real-time inventory data, the real-time price data, and the cached historic theme scores.
18. The one or more non-transitory computer-readable storage media of claim 17 , wherein a first machine learning model generates the historic theme scores during the first phase, and a second machine learning model generates ranks the themes during the second phase.
19. The one or more non-transitory computer-readable storage media of claim 16 , wherein the causing display includes determining a display order for a top ranked subset of the themes and arranging the top ranked subset of the themes in the user interface according to the display order.
20. The one or more non-transitory computer-readable storage media of claim 16 , the operations further comprising:
dynamically updating the ranking based on a change in the real-time inventory and price data associated with the offers for the one item; and
causing display of a new top ranked subset of the themes in the user interface according to the updated ranking, the new top ranked subset including at least one theme that was not included in the top ranked subset.