Humanising product search and making it truly personal
Personalisation is a macro-trend that’s driving real changes in how businesses work – from marketing to e-commerce. When using marketplaces or e-commerce sites, it is clear that users would prefer a more personalised experience – tailored to their needs and wants. And, of course, platforms desire the resulting improvement in user propensity to buy driven by a more productive and satisfying experience.
Getting personalisation right requires a new product search approach
The approach that many platforms have taken to personalisation seems quite odd. Aside from attempts to provide ‘recommendations’ (frequently bought together, customers who bought this also bought that, etc), which have had mixed success, platforms have focused largely on reducing cart abandonment, but as a side effect have created a ‘panicked’ shopping experience.
Here at Traverz, we believe that the core of the user experience is the product search process itself. As we noted in an earlier article, that process is generally based on an on-off filter bar or single-shot search bar that simply does a poor job of representing the ‘fuzzyness’ and multi-step nature of user search requirements. It also by its very nature forces users to express their interests in a very shallow way.
True personalisation requires us to completely reimagine the core search process, and unleash ourselves from the shackles of the on/off binary filtering approach.
The joy of Preferences
Searching is a journey of learning and discovery. You’re looking at specific examples, learning about your requirements and making detailed decisions along the way.
How might we let the user communicate those requirements?
Firstly, let’s give the user an option to indicate Like, Dislike, and potentially stronger or weaker gradations of these. The standard binary filter typically only lets a user express a ‘like’ option, so this instantly gives them a great deal more flexibility.
Next, we enable some form of explicit ranking of those preferences, so the user can indicate which elements are the most important – for example that for them, the clothing colour is less important that the brand. This enables sophisticated simultaneous balancing of multiple product features.
Using these user-provided preferences and importances, we can then create a product ranking. This results in an ordered product list that fully reflects the user’s desire to balance between different wants and needs, where the highest ranking products are the most preferred.
This is far from being a pure user benefit – imagine if, as a platform, you had access to this level of preference information for your users? Instead of largely guessing at user needs, you are instead able to both better serve individual users by basing results on previously expressed preferences, as well as understanding the larger picture of what users generally look for in your products.
Modernising conversational recommendation for today’s product search
This approach to registering user ‘preferences’ and using a utility-style ranking calculation is not new – it is closely related to the ideas behind an approach to product recommendation called ‘conversational recommendation’. This was not widely adopted because it led to a user experience marred by annoying sequences of question forms attempting to extract preferences, and the ‘conversational’ element required significant user patience.
New technologies such as Artificial Intelligence (AI), Natural Language Understanding (NLU) & chatbots present the opportunity to change all that. Through natural language communication, and a pro-active agent-based approach, we can encourage a fluid and dynamic conversation with the user – ensuring that search becomes a multi-step journey and that we are able to bring the benefits of assistance and insights into it.
To provide a truly fluid search experience we need to go one step beyond the concept of a smart agent – and give the user the ability to provide feedback everywhere and anywhere in the search process. This enables the user to independently and in a very natural way, provide preference information without intrusion into their search process.
As we’ve already discussed in an earlier article, at Traverz we have developed exactly such mechanisms – the Traverz agent, and the Traverz feedback mechanism that we name “in-place feedback”.
Revolution not evolution – bringing together preferences, feedback, and agent for a human search experience
So now we have all the components required to bring together to deliver a complete, modern and far more human search experience for users. We have a system of recording multi-dimensional feature preferences as well as the ability to rank those preferences in relation to each other. We have a mechanism for users to pro-actively provide preference feedback as they experience the search results. And we have a smart agent which enables natural language communication, provides intelligent insights, and asks relevant questions – to facilitate the search as a journey.
This is what we call Search-by-Traverz.
For the user, this provides a personal, productive and enjoyable search experience. They are free to browse at leisure but can also take advantage of all feedback and agent mechanisms when desired.
For the platform, this steps up the level of personalisation and intelligence provided to the user by an order of magnitude. Users have greater trust in the choice they are making, and are therefore more likely to convert to a customer – and are more likely to return. In addition, the preference information collected during the user journey is extremely valuable not only for serving this user in their next and future searches, but also can help the marketplace make better recommendation and return more accurate results for other users.
To find out more about the creation of a more human, personalised search experience,
ask about our white paper.