How should an e-commerce search engine ideally search?

43 Views Asked by At

If I am searching for washing machine.
Should it search for -

  1. Only washing machine
  2. Washing + machine + washing machine, showing the washing machine results first

Or is there another better logic?

2

There are 2 best solutions below

0
mkysoft On

The result can be change by side: customer vs seller. You want to sell more profitable product as a seller: washing machine with dryer. You want to find just washing machine as a buyer.

You need to cover some basic concept for searching product:

  • Exact match
  • Partial word/sentence match
  • Synonyms
  • Root of the word
  • Ranking
  • Ordering
  • Manipulating the result

You can check this blog post. Also you can find other blog post about searching product for commerce.

0
Martin Ackermann On

I think, you should definitely apply the second approach.

In your given example it may look indeed questionable what to search for. However if you do exact search in a real world project you would lower the customer experience dramatically. Customers don't know your products or data, they don't know what you expect them to search for. Some customers may come with silly long search terms, or searching in sentences.

Maybe one customer searches for "fast washing machine", next one for "silent washing machine", a third comes with "fast and silent waching machine".

In such cases you still want to show search results - that's why you search for words, rather then phrases. A simple paradigm is then, that more matches equals more relevance of the result.

However, you also have to "finetune" your search engine, e.g. feeding it stopwords (= words to ignore, like "and", "or", etc.), synonyms and apply "fuzziness" (e.g. my last example contained "waching" instead of "washing").

Actually, you always want to show results.

Since this question is tagged with "SAP Commerce Cloud" - I suggest to use the standard SOLR integration, which covers most of use cases, and can be tuned very fine granular.

Feed it with data, stopwords and synonyms. Configure the query building and use default boost-mechanisms. Use adaptive search extension for the personalization.

From my point of view, this is already very strong - it usually takes care of the aspects you are asking for.