findsi: find-as-i

A next generation recommendation-engine, designed to supersede the "people who bought this, also bought"-breed.


Findsi takes the anonymized individual as its starting-point. Results are always personalized while privacy is guaranteed.

Findsi considers the complete anonymized user set present as the individual’s context. Therefore, results are more accurate.

Findsi is a semantic tool. Nevertheless, it is language agnostic.

Findsi is only accessible through an API. You decide how to display the results. We only empower your decisions.

How it works

Despite the inert complexity of the fundamental algorithms and technology, Findsi is built around relatively simple concepts.

  • Step 1. You collect user (Actor) action logs and send them to findsi.
  • Step 2. Findsi anonymizes the logs before they reach the Findsi database
  • Step 3. Findsi returns to you a top-N recommendations.

The below diagram visualizes the described process:

The recommendations can take the following forms:

  • Similar Actors
    A ranked list of Actors that are deemed similar to the specified Actor
  • Similar Targets
    A ranked list of Targets (items) that are deemed similar to the specified Target for the also specified Actor
  • Recommended Targets
    Based upon Similar Targets, nut now Targets known to the specified Actor are filtered out

What features are available to you as a consumer, as well as the amount of date returned, depends on the pricing plan you choose.

Collaborate in isolation

Whilst Findsi constantly attempts to merge tacit knowledge from all consumers, it does guarantee complete isolation. At first glance these goals seem to contradict, but due to a cleverly layered data structure it is actually possible. As a consumer, you communicate with Findsi about your Actors and Targets based on two concepts, an Identifier, which is specific to you, and a resource, which is common to everyone. Some examples:

  • In an online book store, you identify a product as (Product\12345689). To the general public that same book is known as (ISBN\999-9-99-99999-9)
  • In a mobile game you allow users to post their scores to a social network. You identify the player as (Player\987654321), whereas to the general public he is known as (SocialNetwork\@handle)
  • In a corporate intranet, you identify your users based on an entry in a directory service (DirectoryService\GUID). That same user is known to the general public as (Email\ but also by his handle on a social network (SocialNetwork\@handle)

From this kind of information a layered structure emerges, where a consumer requests results from Findsi based on private identifiers, but where our algorithms work on the public resources. The results, send back to the consumer, will only contain the private identifiers, never the public resources.

By identifying Actors and Targets with public Resources, eventually data overlap will inevitably occur. These joint data sets result in much richer data to be analyzed by our algorithms. Simply put: Behavior of an individual, registered by one consumers application might affect the results returned by Findsi to another consumer’s application. You will still only receive entities that are identified by you, but their selection and ranking will be more precise. NB None of your data will be explicitly shared with other consumers nor will they receive recommendations linking to your entities.

Privacy and Security

Findsi guarantees privacy of profiled subjects by scrambling the Resources. in other words, our data stores will never contain any plain text Resource representations. As Identifiers are specific, and thus only useful to, a certain consumer, they are stored without any form of encryption. This does not mean that you could not encrypt your own Identifiers, before you send them over to Findsi. It does imply that if you do, you are responsible for the inherent key management and decryption.

All communication with the Findsi API is done over an encrypted HTTPS connection. Each request to our API is related to a consumer, based on an embedded API-key.

Getting started

To get started with Findsi, you need the following:

  • An API-key
  • An SDK, or the expertise to communicate with a RESTful HTTP endpoint

If you do not have an API key yet, contact us and we’ll set you up with one.

There is a growing set of language specific SDK’s available on GitHub. If for your technology stack none is available, feel free to contact us, so we can assess the possibilities to create one.

Before you get started, be sure to read through our developer documentation first.


Findsi is still in beta. Before charging anything for the use of our service, we want to make sure that there are no unpleasant bugs, and that the algorithm performs as expected, even under high loads. There’s also, a backlog of potential new features, which we'll only implement if there's actual interest. Be sure to vote!

The beta period extends until 2015-01-01 and everyone is welcome to join in free of charge. Participating during the beta phase incurs no obligation whatsoever when the period expires. We hope you'll stay with us of course! Read more


As we are still in beta we have not yet settled on actual prices. We do have a, still evolving, idea of what plans we will be offering after the beta phase.

  Free Bronze1 Silver Gold  
Logs/month unlimited unlimited unlimited unlimited  
Requests/month t.b.d. t.b.d. t.b.d. t.b.d Excess requests will be charged
Excess Requests fee n/a2 €0,004 €0,002 €0,001 Charged per month
Similar Targets max 3 max 5 max 5 max 10 Ranked, similar Targets, possibly known to the Actor
Recommended Targets n/a n/a max 5 max 10 Ranked, similar Targets, unknown to the Actor
Similar Actors n/a n/a n/a max 10 Ranked, similar Actors
Filtering n/a n/a Targets Actors, Targets Filter out certain Target or Actor types
Help & Support docs docs, e-mail docs, e-mail, personal3 docs, e-mail, personal3  
  €0 €50/month Contact us Contact us  

1This plan is also available, at no charge, for selected non-profit organizations
2Request will result in an error once the limit is exceeded
3Personal support while planning integration (or upgrades)

If you recognize the value in our services, and are considering to permanently integrate Findsi into you own application, feel free to contact us, so we can discuss ballpark figures. Somewhere in Q3 of 2014, we will publish our final pricing plan. All beta testers will receive a notification when this happens.

© 2014 Findsi B.V.