Introduction to Octy AI

Mission Statement

To create affordable, analytical and comprehensive machine learning toolchains that are simple to integrate with your current system.

What's new

Here at Octy AI we provide a wide range of solutions that are integrated end-to-end.
We reduce your cost and the risk of failed internal projects. We have new and innovative solutions that are simple to integrate with your current systems. Our intelligent back-end system performs the ‘heavy lifting’ by training data aggregation/shaping, model-training, model-deployment and model-hosting for each algorithm available in it’s toolchain.
In terms of a machine learning implementation, your systems only have two responsibilities; providing raw input data and making model predictions.
Studies show that it is 5 to 25 times more costly to acquire a new customer than it is to retain an existing one. Our solution to this is the Retention Toolchain. You get all the benefits associated with machine learning implementations in a customer retention context without the costs and risks of creating an internal project.
The Retention Toolchain is comprised of multiple pipelines that seamlessly fit together. From customer segmentation and item recommendations to natural language generation, At Octy, we have done all the work for you. We have researched the best possible implementations to maximize the true power of your data, saving you valuable time that you can spend on more important tasks.
For more information on the Retention Toolchain, visit our API reference ↗️  page.

Retention toolchain concepts

To help you understand the value that you get with the Octy platform, we have outlined the key concepts that the Retention Toolchain delivers. These concepts are explained in further detail, later in this documentation.
Profiles: At the centre of Octy's toolchain is the customer profile. A profile is an anonymised representation of each customer. Events and segmentation data are recorded against each customer profile. Octy then further populates each profile with model predictions and segment tags. This gives you a clearer picture of every customer on a granular level, allowing you to understand their behaviours and needs. Now you can make predictions and item recommendations tailored to each customer. Profiles can be created for customers who have not conducted a purchase yet, by setting their 'has_charged' attribute to 'false'.
Items: Structured representations of the products your business sells. Item data is used in various training data sets for different models within the Octy platform.
Events: Events are referenced in two contexts throughout the Octy platform:
  • Custom Event-Type: user-defined event types, allowing you to track instances of this type.
  • Event instances: data is used both for training datasets and plays a fundamental role in profile segmentation.
Segments: A group of profiles based on common characteristics and behaviours. Allowing you to hyper-target your outreach, content and messaging. Reducing unnecessary advertisement costs. There are two primary segment types: Live-segments and Past-segments. Also, segment tags are populated in each profile that meets any segment definition criteria. Octy defines a segment as established criteria of both profile attributes and a sequence of predefined events.
Churn prediction: Models are trained with a combination of profile and event data to produce accurate churn likelihood predictions against/for each profile. Once you have set the relevant algorithm configurations and there is adequate data to perform training, this process will run automatically and at regular intervals. Churn likelihood predictions are populated in each profile that exists at the time of training.
RFM analysis: The three dimensions are Recency, Frequency and Monetary value. A score is assigned to each dimension and recorded in the customer profile. Again, once you have set the relevant algorithm configuration and there is adequate data to perform analysis, this process will run automatically and at regular intervals.
Messaging: A basic but robust form of natural language generation. Message content templates can be tailored for each segment. You can create message templates and generate unique content at scale in seconds.
Recommendations: Items are recommended based on customers’ previous item interactions and the item interactions of similar profiles. Item recommendations can be retrieved directly via the API, or populated into messaging content. Recommendation model training jobs are conducted automatically, at regular intervals, to maintain relevant item recommendations.


Currently, the Octy platform has two key tools:
API (The Retention Toolchain Application Programming Interface)
CLI (Octy's Command Line Interface, used for configurations and resource management)
The CLI is a limited scope wrapper for the API, allowing you to quickly update configurations or manage resources without a third party API client. The exact functionality of CLI and commands can be found in the API Reference. Each Request that can be performed using a CLI command is marked with this 🖥️ icon, with an example command.
The CLI tool can be downloaded from here Downloads section ↗️
You can view the full CLI documentation here ↗️

Where to start

There are a few tasks you must complete in order to get everything up and running.
Go to the Getting Started 🔗 section of this documentation to get started.


Getting Started
Creating Resources