At Coveo Impact 2019, Vincent Bernard from Coveo presented about the use of Coveo Cloud to maximize the value of relevance on a website.
3 Main Ingredients for a Good Search Experience
Vincent described three main ingredients or best practices for web site search that all work together:
- Machine Learning Query Suggest – not just type ahead, but AI powered suggestions as you type
- Automatic Relevance Tuning – brings the most relevant results to the top of the list
- Content Recommendations – show appropriate content to the user without them asking for it.
At least 75% of Coveo sites used at least one of these ingredients, but only 25% use all of them.
Keyword Problem
When humans communicate, the use a lot of non-verbal signs, but search engines don’t understand these items. When we search for just keywords, we don’t take advantage of the “non-verbal” cues. On the other hand, if you can combine keywords with other clues about the user, you have a better chance of returning more relevant content.
Promise of AI-Powered Search
That’s the promise of AI within search systems. AI powered search systems can discern intent based on the user’s context, their behavior and what they type into search boxes. With machine learning, the system can look at what other people with similar context, similar behavior and similar keywords searches found important. These results are likely to be the same for the current user, so the system brings high relevance to the results.
Taking Advantage
To take advantage of the Coveo Platform, you should follow these steps:
- Get trained and really understand how to use the platform
- Set as a goal in phase 1 to take advantage of Machine Learning and other components, not some later phase
- Take advantages of connectors – Coveo doesn’t charge for the volume in the index or the connectors use, but only for the number of queries.
- Gather and take advantage of the user information available to you whether the user is known or anonymous
- Track the data provided by the system through clicks and events