The science behind taste-based recommendations

Many recommendation engines use collaborative filtering: they look at what similar users bought or watched and suggest the same to you. That works well for films and everyday products, where tastes tend to cluster.

It works less well for personal decisions. Choosing a supplement, a trip or a gift for someone you love is not like choosing a film. "People like you also bought this" says little about why you are choosing, or what you care about most this time.

Flairmingo borrows a simple idea from the study of preferences: ask people to rate a few clear dimensions, rather than guess from their clicks. Each category has its own set of dimensions, chosen for the decisions that matter in that area.

The useful signal is the balance between those dimensions. Knowing that someone "cares about wellbeing" is not enough. Do they lean towards calm or energy? Sleep or strength? That balance is what we call your taste.

When you set your taste profile, each answer becomes a number on one dimension (for wellness: calm, energy, sleep, strength, mental wellbeing and looking younger). Together, those numbers describe your priorities.

Every item in our catalogue is scored on the same dimensions. That lets us treat recommendation as a comparison: which items have a pattern of strengths closest to your pattern of priorities?

At the core of the comparison is cosine similarity, a standard way to measure how closely two patterns point in the same direction. It looks at the balance of values rather than their size.

So someone who rates everything highly and someone who rates the same things more modestly get similar rankings, because the balance of what they care about is the same. If you care about sleep and calm equally, an item strong in both ranks well, whether you said so boldly or gently.

The figure you see, such as "88% fit", starts from this comparison. Flairmingo blends in a few other signals and scales the result so it is easy to read, then marks the best fit as your Top find.

A ranked list is a good start, but real decisions often involve a few things together. Flairmingo can group items into routines in wellness, and into plans in Chapters for life's big moments such as Christmas, birthdays and weddings.

A group is ranked by how well its parts fit your taste together, so the suggestion reflects the whole, not just its best single item.

Your first taste profile is a starting point. As you save ideas or adjust your priorities with the Trait Tuner, Flairmingo uses those signals to refine your own rankings.

The aim is simple: fewer irrelevant results, less scrolling and less mental load each time you come back.