Over the years many consumer services providers in ecommerce, talent search, food delivery, streaming entertainment, social media, etc. have started providing users with personalized search results.
But almost all of the personalization efforts have been focused on using any derived understanding that brands gain from the user and the content/ services they consume.
In technical terms, this means that search personalization is considered to provide different, but relevant, results for the same query based on the user and context.
This effectively puts the user ‘query’ to one side, and takes it as given. In practice, the ‘query’ is any information that the user has provided during the search journey.
By defining search personalization this way, there is a real danger that when brands try to enhance personalization they focus fully on external/observed/indirect user data. They have stopped seeing user interaction as a powerful source of ‘personalization’. This is exactly what we see happening in the industry right now. Almost 100% of the focus in personalization and recommendation is on predicting user needs, without any actual direct involvement from the user.
So why don’t we consider user interaction to be a key mechanism to improve personalization? There are two objections to consulting the user:
- The first objection to exploring user interaction is that it is unnecessary.
This assumes that given sufficient data (from third-, second- or first-party sources) and enough machine learning power, brands can ‘predict’ what the user wants — maybe even before the user realizes it.
For impulse buys, it would seem plausible that brands can meet the needs of a decent proportion of users this way (although today’s user responses to recommendations show that platforms are far from having a crystal ball!). But how big is “a decent proportion”? And can we afford to just focus on that group, shrugging our shoulders at the experience of the remainder?
This has been Amazon’s initial strategy to gaining market dominance — serving up a predicted product set that meets a meaningful proportion of users’ needs — but now that this method of dominance is nearly exhausted, they are looking at how to better meet the needs of the remainder. Privacy concerns are also gradually reducing the amount of user data that is available, thereby reducing prediction accuracy.
When it comes to considered purchases, the search journey is an iterative learning process and one-shot predictions are often a recipe for frustration. User engagement then becomes an absolute necessity.
- The second objection is that user interaction creates friction, i.e. users “might not like it.”
True. But the benefits gained from knowing what the user wants, rather than guessing, are huge.
Users appreciate being able to convey their needs and reactions, and the one-shot personalized recommendation is a very hit-and-miss affair. A hit is an instant success; a miss is an instant failure that all too often leads to abandonment of the search. The key is that users should be “able to” convey needs and wants — as opposed to being “required to.”
When users are able to engage, they can allow us to deliver a 10X improvement in personalization, as well as recover from incorrect or unhelpful assumptions. On top of this, brands can still add an additional layer of personalization based on observed data about the user and the context of their query. But when users are unable to engage, the ability to personalize is severely constrained by very limited knowledge about them and their current intent.
Moving Personalization to the Next Phase
Why are brands limiting themselves to simply predicting what users want when they have the opportunity to know? It’s time for brands to rethink their approach to how users search for products on their websites and move personalization to its logical next phase — where we escape making predictions and instead focus on truly understanding consumer intent.
The good news is that the route to achieving the next phase of personalization does not involve a radical overhaul of existing methods/approaches.
Brands need to remember that people are at the heart of all search queries. Yes, they have the technology to deliver various algorithms to help with online shopping experiences, but without interaction they simply cannot continue to guess what the consumer wants. It is time to engage them, to understand and know what they want.
Twan Vollebregt is Co-founder and CEO of Traverz. He is a serial entrepreneur and executive leader whose experience is focused on innovation. He has led and exited two previous tech startups whose products are still very much alive for the acquiring companies today. In 2014, Vollebregt won the prestigious Risk Magazine Innovation of the Year award for his work relating to a web platform product. He holds a PhD in Operational Research.