How Snap Works

Get Going in 4 Easy Steps
Explore the Snap Prediction model catalog. Understand the data dictionary, what the models predicts, sample data and underlying algorithms
Experience the models 'live'. Train the model, score your data and evaluate performance
Like the model but need to make some changes? Customize the model to fit your requirements
With a single click, deploy the model on the Snap Analytx service and start using the model right away!

Save Money with Snap

No Long Term Commitments
With Snap you pay as you go, and you only pay for what you use.
Flexible Subscriptions
You can subscribe to Snap models on a month-by-month basis. You can unsubscribe at any time.
No Added Costs
There are no software license costs or annual contracts.

Choice - Flexibility - Options

Choice of Tools and Languages
Snap supports most of the popular Predictive Analytic tools, languages such as SAS, SPSS, Matlab, R, Python, Java and Mahout. You are not locked into a proprietary tool or language.
Cloud or On-site
You have the choice of deploying your models (without any changes) either on-site or in the cloud.
Pre-built or Built in-house
You can deploy in-house developed custom models on the Snap service to take advantage of the hassle-free end-to-end model lifecycle management. So, whether you are buying from the Snap marketplace or building your own models, you can still offload model management to Snap.

End-to-end Lifecycle Management of Models

Rapid Deployment
With Snap's cloud service you can quickly deploy your models into production for use by all authorized users worldwide. Models that require no customization can be deployed with a single click.
Hassle-free Provisioning
With Snap, you do not have worry about any of the things necessary for deploying and administering models in production, such as provisioning hardware, scaling on demand, security, tracking performance, retraining the models etc. All of it is taken care of for you.
Perfomance Monitoring and Correction
You can define the metrics for acceptable performance of your models. As the performance of even the best designed models deteriorate overtime, due to evolving business requirements and changing data, Snap will continously monitor the performance of models to ensure they continue to meet requirements.
What a Model Seeker is Saying
Tarek Najm
Founder and CEO, Parllay Inc
Open marketplace for predictive models is a good idea
Whitepapers about Predictive Analytics Marketplaces
When will the Snap service be publicly available?

The service is currently available only as a customer preview. We are working with a limited number of early adopter customers who are helping us fine tune the service. During the preview period, the service is available for free ($0.0/per month) for customers accepted into the preview program.
To join the preview program contact
We will soon be announcing the date for general availability of the service.

What is the process for purchasing a subscription to a model in the catalog?

After you find a model in the catalog that meets your requirement, please contact us ( or call 425-261-1155, ext Sales ) to purchase a subscription to the model. After the transaction is complete, we will provision the required cloud computing resources, deploy the model and provide you with the access code so you use the model through web service APIs. You can start using the model right away.

What usage rights do I get when I subscribe to a model?

As an active subscriber, you will have the right to use the models that you have subscribed to, by invoking the associated web service methods (APIs). The monthly subscription fees includes a certain number of free web service calls and the amount of data that can be submitted to the model for training or scoring. Additional fees may apply for additional calls and additional data. The details will be specified in the pricing and licensing terms and conditions for each model.

Can I purchase the model outright if I do not want to subscribe?

Yes, you can purchase the model from the author. The author will specify the price of the model along with licensing terms and conditions. Purchased models can still be deployed anywhere you choose, includinh on the Snap's cloud to take advantage of Snaps's end-to-end lifecycle management services.

I found a model that I like, but I want to further customize it to suit my specific business requirements. What is the process for customizing a model?

To customize a model click on the "Customize this Model" button for the model to contact us.

We will schedule a meeting with you within 24 hours to understand your requirements, including your performance and security requirements, the service level agreement and usage patterns.

We will then get back to you within 5 (five) working days with the one-time customization fees, monthly subscription fees and the estimated time to deliver the completed model. The time to customize a model depends on the extent and nature of customization and can take up to 2 weeks. The one-time customization fees is non-refundable.

After the customization is complete, you can purchase a subscription to the model described above. We will provision the required cloud computing resources, deploy the model and provide you with the access code so you use the model through web service APIs. You can start using the model right away.

What is involved in customizing a model? Who actually does the customization?

You can get a model customized to meet your exact business requirements and your data. Customization can range from tweaking the model parameters to reimplementing the model with (for example) a different algorithm. Generally customizing a pre-built model can be substantially cheaper and quicker than building a model from scratch.

Customization can be done by the model author. However, if the model author is not interested or unable to do the work, Snap's in-house team of data scientists can do it for you. Your customized model is exclusively for your use and will not be shared with anyone else.

The cost to customize a model depends on the type of customization and varies by model. Customizatio is usually a one time cost that is different from the monthly subscription fees.

What if I cannot find a pre-built model in the catalog that meets my specifications?

We have established partnerships with select companies who have domain-specific expertise and experience creating prediction models. We will work with them to source the model for you. Here are the steps to get a custom model. We will be able to provide an approximate availability date for the model usually within two weeks. Please submit your model request using the Snap Prediction Model Request form and we will get in touch with you. The process is similar to customizing a model.

Will I have access to the source code to the predictions models I subscribe to?

You can get the source code only if the model author consents.

What tools do I need to use the service? Do I need tools such as SAS or SPSS?

You do not need any data mining or predictive analytics tools to use the service. All you need to do to use the service is be able to invoke the associated web service methods. You can also user the browser interface to interact with the models.

How do you ensure the security of my data? How do you prevent other customers from viewing my systems?

Each Snap deployment operates within a unique sandbox identified by a unique secret key. All interfaces to Snap must provide the secret key. The data from one customer is used to train an instance of the model, and the model parameters that fit the customer's training data are unique to the customer. Similarly, the web service call for run-time scoring must also include the secret key. Snap will securely encrypt all customer data that is saved persistently. For additional security, Snap will include libraries that a model integrator can use to convert any sensitive predictor variables to anonymous.

Are the models deployed on your own servers in your data centers?

No, we use Amazon Web Services and Windows Azure.

Do you offer any service level agreements (SLAs)?

Yes, please scroll down for more details.


Service Level Agreement

The Snap Analytx service level agreement (SLA) is subject to the terms of the Snap Analytx agreement between Snap Analytx, Inc. and the customer.

Snap Analytx will use all reasonable efforts to make the Snap prediction service available with an annual uptime percentage (defined below) of at least 99.95% during the service year. In the event that Snap prediction service does not meet the annual uptime percentage commitment, you will be eligible to receive a service credit as described below.


"Service year" is the preceding 365 days from the date of an SLA claim. "Annual uptime percentage" is calculated by subtracting from 100% the percentage of 5 minute periods during the service year in which the Snap Analytx prediction service was in the state of "service unavailable". If you have been using Snap Analytx for less than 365 days, your service year is still the preceding 365 days, but any days prior to your use of the service will be deemed to have had 100% availability.

Any downtime occurring prior to a successful service credit claim cannot be used for future claims. Annual uptime percentage measurements exclude downtime resulting directly or indirectly from any Snap Analytx SLA exclusion (defined below).

"Service unavailable" means that the prediction service is unreachable to you or does not return the predicted result within the response time specified in the Snap Analytx agreement.

The "eligible credit period" is a single month, and refers to the monthly billing cycle in which the most recent service unavailable event included in the SLA claim occurred.

A "service credit" is a dollar credit that we may credit back to an eligible Snap Analytx account.

Snap Analytx SLA Exclusions

The service commitment does not apply to any unavailability, suspension or termination of Snap Analytx, or any other Snap Analytx performance issues:

If availability is impacted by factors other than those explicitly listed in this agreement, we may issue a service credit, considering such factors in our sole discretion.

Request Prediction Model
First Name
Last Name
Company / Organization
Your Industry / Vertical
Please describe what your organization does?

Are you using any of these tools today?
Briefly describe your business problem and what you are trying to predict?
By when do you need the model?