AI Credits Definition

15 Million Incoming Events

–

✓

✓

–

25 Million Trigger Activations

–

✓

✓

–

1 Million Email Messages

–

–

✓

–

10 Million Mobile Push Messages

–

–

✓

–

40 Million Predictions - ML Models

–

✓

✓

✓

100 Million Predictions - RFM Model

–

✓

✓

✓

* Measured and incurred monthly.

Calculation Examples:

Customer consumes 550 Conversations = .92 Credits

Customer consumes 300 Conversations + 500,000 Messages = 1.00 Credit

Conversations

A Conversation is the metric of usage of Treasure Data's Treasure Studio, Treasure Code, and agents created or used from the AI Agent Foundry.

A conversation is an average of five back-and-forth's between two entities, which may be a human and an agent, or an agent and another agent, leveraging one of our standard Large Language Models as a basis. Shorter sessions will be metered as a fraction of a Conversation, measured to two decimal places. For longer sessions, the metering rate (i.e., the ratio of back-and-forth's to a conversations) will decrease below 5.0 if and to the extent the number of back-and-forth's exceeds 5 (for example, a session with 10 back-and-forth's may be metered as 2.87 Conversations).

Conversations are metered on all Production Instances.

Real Time Profile

A Real Time (RT) Profile is a collection of Core ICDP attributes and RT Attributes related to a single individual held in the Real Time Decision Engine.

Each RT Profile may have up to a combined two hundred (200) Core ICDP attributes and RT Attributes, with a maximum size of 500 bytes per profile.

Each RT Profile is counted as one (1) RT Profile, regardless of whether it is a Known RT Profile (has Personally Identifiable Information) or an Unknown RT Profile (no PII).

RT Profile quantities are measured on a monthly basis to determine Credit consumption. The number of RT Profiles calculated for a given month will be the 4th highest number of RT Profiles held in all Production Instances counted on a daily basis during the month.

Personalization Calls

A Personalization Call is a call to the Personalization API which, using a Real Time (RT) user identifier, returns a payload of up to 20 RT Attributes and Core ICDP attributes which are stored in the RT Decision Engine.

The number of Personalization Calls for a given period will be the total number of Personalization Calls recorded in the RT Decision Engine(s) in all Production Instances for that period.

Incoming Events

An Incoming Event is an incoming streaming event that is ingested into the Real Time (RT) Decision Engine.

The number of Incoming Events is the total number of Incoming Events recorded in the RT Decision Engine(s) in all Production Instances for a given period.

Trigger Activations

A Trigger Activation is an outgoing streaming event that is generated by the Real Time (RT) Decision Engine, which can result in a call to downstream streaming connections for an action to be taken.

The number of Trigger Activations is the total number of outgoing events recorded in the RT Decision Engine(s) in all Production Instances for a given period.

Email Messages

Email Messages measures the number of messages sent by a customer using Engage Studio in all Production Instances within a given period.

Mobile Push Messages

Mobile Push Messages measures the number of messages sent by a customer using Engage Studio in all Production Instances within a given period.

Predictions

Predictions are generated as a result of training and prediction runs of Profiles through a machine learning model. The models include but are not limited to Next Best Product (NBP), Next Best Action (NBA), Recency-Frequency-Monetary (RFM), and Customer Lifetime Value (CLTV).

The number of Predictions are recorded as the number of outputs from a prediction run for all Production Instances for a given period.

AutoML Tier 1/2/3

AutoML automates machine learning (ML) process steps such as data processing, feature engineering, model selection and hyper-parameter tuning to reduce time to deploy ML models. It provides users access to the AutoML capabilities from Treasure Workflows.

AutoML is available in the following options:

AutoML Tier1:

Up to 256GB RAM

Maximum runtime hours: 500 Unit Hours

4 AutoML tasks can be run concurrently

5 TD users can run AutoML tasks

AutoML Tier2:

Up to 384GB

Maximum runtime hours: 1000 Unit Hours

8 AutoML tasks can be run concurrently

15 TD users can run AutoML tasks

AutoML Tier3:

Up to 512GB RAM

Maximum runtime hours: 2000 Unit Hours

16 AutoML tasks can be run concurrently

Unlimited TD users can run AutoML tasks