I spent a large portion of my e-commerce search life explaining why searching for „t-shirt“ and „tschirts“ sometimes yields totally different results in terms of result size, product selection, sorting and visual display. The major part of my job (as may apply for other „Customer Excellence“ or „Site Search“consultants out there) was to teach e-commerce professionals how to tweak their search engine to produce exactly the results they wanted.
Things like adding synonyms, defining manual search rewrite rules or even more advanced curated search results. Or (for the real, real professionals) configuring the internal settings of the search engine (fuzziness, field weight, product ranking rules).
Usually, all of these efforts resulted in numerous quite well-optimized search result pages that converted nicely. But they completely ignored Long Tail! Even worse: Only in very rare cases optimized or curated search results were tested properly to make sure to have the best possible result set. Even if they were tested — things changed within a matter of weeks or days and test results turned old.
The enemy of a good search result may be a new marketing campaign, new viral hype or (in most cases) a simple index update with new products.
I’ve seen a lot of customers that have tried to solve these issues by adding almost endless lists of “synonyms” in order to outsmart the search engine whilst completely ignoring the definition of a synonym. Synonyms represent terms with the same meaning but different spelling. A good example for a pair of synonyms is notebook and laptop. In this case, you might need to add a synonym because the word similarity is extremely low while the semantic similarity is quite high.
However introducing synonyms for handling spelling errors, over-/understemming, decomposition problems is something different and represent nothing more than a non-scalable short-term solution to a much bigger problem. As it produces other problems like unmanageable lists of thousands of synonyms, very complex queries with sometimes transitive matches especially when you have to deal with multi-term synonyms.
Finding a totally new approach
When we at CXP decided to create a new kind of search appliance, we didn’t want to invent yet another search engine. Nor did we want to create even better tweaks for existing search engines.
Instead, we wanted to create something that really helps search engines to finally deliver the best result available for each and every search request.
The best result is a result that really fits the customer‘s need.
But how could a search engine know what a customer wants? Very easy: They are telling us what they want by typing letters in a search box (or by commanding Siri, Cortana, Alexa, … to do so).
Once we know what our customer want, we can automatically deliver the result that performed best for others who wanted the same by infusing the unbeatable combination of data and AI. It is as simple as that.
Now this is exactly what CXP Search|Hub does:
- Understanding customer intent.
- Find the search request that best served other customers with the same intent
- Ask the internal search engine to deliver a matching result set
- Observe the customers behavior and feed it back into the Search|Hub Analytics
Facing the same challenges as Site Search Consultant. Give www.searchub.io a try and make your search fly.