If you ask users, a search engine is the best way to find information and products on company and organisational websites. As mobile usage and speech recognition increase, the need for search functionality will continue to grow.
The search engine is a customer server
Relevant search results are key for business and the most relevant search results must be displayed at the top of the search results. At its best, a search engine is a customer service agent that:
– answers customer questions
– collects customer information
But this is not done by clicking your fingers. The enthusiasm for development is also dampened by the fact that Google does a reasonably good job of on-page search and companies do not use the search engine as an interface with the customer. Users provide information to Google and Google answers their questions. However, this does not necessarily serve the business or even the users in the best possible way.
The good, the bad and the ugly of a UX consultant
During my career as a UX consultant, I have analysed over a thousand usability test sessions. In these tests, users would prefer to use search functionality, but search rarely meets the user’s needs. There are happy exceptions, such as Motonet, whose search engine was developed in-house. Too often, however, it is assumed that search works with the default settings. Many search engines bought as a service search blindly on the basis of their default algorithms without understanding the specifics of the site: if you are looking for skates in a hockey store and the first skates are ranked number 12 in the search results, the machine-learning search engine has clearly learned nothing from the user clicks. In this case, users had been clicking for four years.
Search technologies are evolving rapidly
However, there have been positive developments, especially on the open source side, and quite meaningful strategies have been developed to produce relevant search results. Largely free platforms such as Solr and Elasticsearch have so far based their algorithms mainly on the frequency of words (would you buy a cookbook because of the frequency of the word cookbook in its title and text?) Today, these search engines can be made to interpret hundreds of page attributes and user signals. Users are understood despite spelling mistakes and the use of different word forms. Wikimedia and Bloomberg are active users and developers of Elasticsearch and Solr.
It pays to develop a search engine yourself
Measurement is a prerequisite for development. If the first search result gets an average of ten percent of clicks, its relevance needs to be adjusted to reach the twenty percent level. Resources need to be allocated in-house to develop the search engine, as the work requires a subtle understanding of one’s own customers and services.
The search engine can become an important part of your core business, so investing in its development internally is worthwhile.
Consultants are also needed, as collecting and using the data needed for machine learning requires expertise and there are not always ready-made tools.
Optimal results are usually achieved through trial and error. In the target mode, the UX of the on-site search engine is better than Google and the data related to business measurement and customer insight comes directly to your own use.
Scrooge Koski
The author is a UX consultant at Nextcon with 25 years of experience in UX design. He is passionate about goals and measuring their achievement and is always learning new things.