ABOUT THE PROJECT

We love movies very much. And, like most of you, we often encounter the following problem:

What to watch when you don't know what you want?

There are different ways to solve this problem: reviews from friends and acquaintances, recommendations from mailing lists, advertising. They all have a common drawback - they do not take into account your individual tastes and mood.

Even your closest friends may not share your tastes, and recommendation lists are subject to subjectivity. This also applies to awards and film festivals, which often focus on popular cultural trends rather than the actual quality and relevance of films.

But even if you have compiled a rough list of films to watch, it will be quite difficult for you to predict whether you will really like a particular film or not, only based on its description, reviews and trailers.

Another way is to focus on your favorite directors and actors, but they delight you with new works no more than once or twice a year…

We offer a different approach - using an independent recommendation service.

Why the real recommendation service should be independent?

The idea of personalized recommendations is nothing new. Now all content providers offer their recommendations. But they have two big drawbacks:

1. They are limited to the provider's directory. Tey are not at all interested in offering films that he cannot show (sell), and thereby send users to other resources.

2. Their task is not to satisfy users, but to increase service revenue. The priority in recommendations is always films that are profitable for the provider to offer for some reason. For example. these are new items, which are not included in the subscription and are paid separately, or productions of provider himself, does not require royalties to the content producer.

And recommendation services, embedded into the content providers applications, always play a purely supporting role. Therefore they are not very functional.

We tried to make a service tailored specifically for receiving recommendations, as easy as possible to use on the one hand, but with unrivaled filtering and sorting mechanisms on the other.

And it seems to us that we succeeded.

How it works

The basic logic of the FilmAdviser is the same as that of all recommendation services. We collect user ratings of films from various sources and ask the user of our service to rate the films he has seen. This allows you to create his individual preferences profile and, based on the ratings of other users, predict how much he will like this or that film. After this, a list of films with the highest predicted ratings is compiled, and these films are included in the recommendations.

However, even with very accurate forecasts, choosing a movie to watch "here and now" does not come down to choosing films with the highest predicted ratings. Among these films are many that the user has already seen, or at least knows about them, and films that he does not want to watch now (not in the mood).

In addition, an accurate forecast requires the user to provide a large number of ratings, which is difficult to achieve in reality.

We came up with and implemented a new approach to user interaction with a recommendation service, similar to a regular TV. But we suggest not single button for switching programs, but number of them.

We offer films to the user one by one, and the user, in turn, can either skip them, or rate them, or add them to the wishlist. when a user skips a suggested movie, he can specify the reason for the refusal.

During this process, we collect much more information than a classic recommendation service. Because the very fact of missing the film is already significant, especially if the user specifies the reason (for example, he doesn't like this genre). Using this information coupled with ratings, we quickly find what is really interesting to the user.

In addition, FilmAdviser has a unique feature - it can filter films by subtle properties: kindness\hardness, fantastic\realism, dramatic\lightheartedness, etc.

With two button presses, the user can say: "What kind of abstruse trouble is this? I want it to be more dynamic and simpler"

We have set ourselves an ambitious task – to make sure that anyone, even those who are not familiar with our service, can find what they want to watch within 10 minutes. And we can proudly say that we solved it!

P.S. To learn more about the capabilities of the service, we recommend reading Questions and Answers

P.P.S. We are waiting for your comments and suggestions at team@kinonavigator.ru

Successful choosing with FilmAdviser!

Our team

	Andrey Akopyants Concept, architecture, product management
	Alexander Razorenov Programming
	Vladimir Bobrikov Recommendation system, usability