Search Optimization Strategies
The strategies below handle the following in the search strings:
-
Different cases (i.e., case insensitive)
-
Stemming - reducing inflected (or sometimes derived) words to their stem (i.e. removes characters such as 'ing', 's', 'est')
-
Common misspellings
-
Word order (For example, Hanks, Tom vs Tom Hanks or duck horn vs horn duck)
-
Punctuation and other (usually) non searchable characters (these are ignored/stripped)
Example (Stemming and Case Insensitive):
If customer A searches for "shoes" and subsequently purchases product A, then
-
When customer B searches for "Shoes", the system will return product A.
-
When customer C searches for "Shoe", the system will return product A.
The strategies below will not handle synonym expansion.
Strategy |
Strategy ID |
Sample Message |
Description |
Instrumentation Page Types |
Required Parameters |
---|---|---|---|---|---|
SolrRecentSearchToPurchase |
44 |
"Products Related To Your Recent Search: ''" |
Products most frequently purchased, given that the customer has searched the same terms that the customer has recently searched. Multi-word search terms are treated as one word. |
Search |
User |
SolrSearchToProduct |
42 |
"People Who Searched 'Keyword/Phrase> Ultimately Bought" |
Products most frequently purchased by customers that have searched the current terms. Multi-word search terms are treated as one word. |
Search |
Search term |
SolrSearchToCategory |
87 |
"People Who Searched '$STRATEGY_HINT$' Bought These Items From Other Categories" |
Top selling products from categories most often purchased by customers searching one or more keywords in current search. |
Search |
Search term |
SolrSearchToPurchasePercent |
43 |
"What Was Ultimately Purchased By People Who Searched for ''?" |
Products most frequently purchased (in percentage form) by customers who searched the same terms the customer has recently searched. Multi-word search terms are treated as one word. Requires layout that exposes conversion %. |
Search |
Search term |
SolrSearchToView |
98 |
"People Who Searched For '' Also Browsed" |
Based off the current search context, products are generated based on products which have been ultimately viewed by customers who have searched for one or more of the keywords that make up the current search. Multi-word search terms are treated as one word. |
Search |
Search term |
SolrSearchToViewCategory |
99 |
"People Who Searched For '' Also Browsed These Categories" |
Keys off of the current search context and finds categories which have been browsed by customers who have searched for one or more of the keywords that make up the current search. It then aggregates all the purchased products by the categories that they belong to and then recommends those categories. Multi-word search terms are treated as one word. |
Search |
Search term |
SearchedPurchasePercent |
29 |
"Items Related To Your Current Search: '$STRATEGY_HINT$' " |
Products purchased by customers that searched one or more keywords that make up current search. Requires layout that exposes conversion % |
Search |
Search term |
RecentSearch |
10 |
"Items Related To Your Recent Search: '$STRATEGY_HINT$' " |
Products purchased by customers that searched one or more keywords that make up a recent search (most often, the last one) |
Search |
User |
MultiSolrRecentSearchToPurchase |
60 |
"Items Related To Your Recent Search: '$STRATEGY_HINT$' " |
Products purchased by customers that searched or more keywords that make up multiple recent searches (most often, the last one). Multi-word search terms are treated as one word. MultiItem strategies do not require multiple seeds in order to be eligible, they are simply able to use multiple seeds when available. |
Search |
User |
RelatedToInternalSearch |
12 |
"Items Related To Your Recent Search: '$STRATEGY_HINT$' " |
Keys off of keywords from the search and correlates them with purchases. When the products are retrieved, products which match the most keywords will have their scores added together. When the products are written to the model, they are written for each term. The strategy will pull the products out and accumulate the scores for those products that match the most keywords. This strategy sees each individual word in a multi-word search as individual words. |
Search |
Search term |