The fastest and most accurate suggestion engine

Tamber is the best way to put recommendations in your app. No big data headaches. No '90s algorithms. No tedious tutorials. Just deliciously tasteful, real time personalization.

Integration easy as π.

Find our client libraries on GitHub.

Discover recommendations that feel tasteful and informed, not robotic

Pull head-scratchingly tasteful recommendations for your users in real time, and start solving discovery in your app.

Node.js
Ruby
Python
Go
Objective-C
Java
curl

tamber.discover.recommended({
	user: "user_rlox8k927z7p"
}, function(err, result){
	//handle result
});

recs = tamber.Discover.recommended(
    user='user_rlox8k927z7p'
)

d = Tamber::Discover.recommended(
  :user => 'user_rlox8k927z7p'
)

recs, info, err := discover.Recommended(&tamber.DiscoverParams{
	User: "user_rlox8k927z7p",
})

$ curl https://api.tamber.com/v1/discover/recommended \
	-u :SbWYPBNdARfIDa0IIO9L \
	-d user=user_rlox8k927z7p

TMBDiscoverParams *params = [TMBDiscoverParams discoverRecommendations:nil];
[[Tamber client] discoverRecommendations:params responseCompletion:^(TMBDiscoverResponse *object, NSHTTPURLResponse *response, NSString *errorMessage) {
    if(errorMessage){
        // Handle error
    } else {
        object.discoveries // Array of recommendations
    }
}];

HashMap discoverParams = new HashMap();
discoverParams.put("user", "user_rlox8k927z7p");
JSONObject resp = new JSONObject();

try {
    resp = tamber.discover.recommended(discoverParams);
} catch (TamberException e) {
    System.out.println(String.format("%s=%s", e.getClass().getName(), e.getMessage()));
}

[
  {
    "item": "item_83jx4c57r2ru",
    "score": 0.94321,
    "popularity": 0.34590,
    "hotness": 0.67234,
    "item_created": 1481167989
  },
  ...up to 500 items per page...
]

The new standard in recommendation accuracy.

Tamber makes mainstream recommendation engines feel like the 90s.

Accuracy (f1-score)

Run Time (hours)

Accuracy comparison of f1-scores of engines cross-validated on the popular Movie Lens datasets. Read our white paper for a deeper dive.

Occam's recommendation engine

Instead of clunky, and cycle-intensive algorithms, Tamber focuses on the metrics that matter and uses lightweight, heavily optimized models. It's single-minded design that does what it says on the tin - and it’s so fast it runs in 100% real time. Never make a user wait for recommendations to generate. Forget the neolithic practice of batch processing. Stop giving users yesterday’s content.

The popularity feedback loop sucked in high school. It still does now.

Search only works when users already know what they want, while mainstream personalization techniques like Alternating Least Squares can only recommend what's already popular. This results in obvious and robotic suggestions, and creates a popularity feedback loop that amplifies the problem.

Sift out the hidden gems

Tamber solves this underlying issue by learning to understand crowd behavior. Accurately predicting the broader movements of groups (and the niches and sub-cultures they represent) allows Tamber to see past the popularity feedback loop that sinks mainstream techniques. This bigger picture means users get highly accurate recommendations that don't incorrectly favor the already popular, or the old news.

Start showing users what they want, before they want it. Get your free API key and get setup in minutes.

Want to try our demoware? Our demos are exclusively customized for your needs, so you can experience the accuracy and speed of Tamber first hand.