Amazon Personalize pricing

Amazon Personalize uses machine learning to automatically segment your users based on their affinity for different products, categories, brands, and more to create more effective marketing campaigns.

The following pricing applies when using the following user segmentation recipes:

• aws-item-affinity
• aws-item-attribute
  

You are charged per GB of data uploaded to Amazon Personalize. This includes real-time data streamed to Amazon Personalize and batch data uploaded via Amazon Simple Storage Service (S3).

Data ingestion costs: $0.05 per GB

You are charged for the training hours used to train a custom model with your data. A training hour represents 1 hour of compute capacity using 4v CPUs and 8 GiB memory. Amazon Personalize automatically chooses the best instance types to create your solutions, and the instance may exceed the baseline specifications in order to complete the job more quickly. Amazon Personalize calculates training hours based on the instance used relative to the baseline instance. The number of training hours charged may be higher than the time elapsed on the clock during training.

Training costs: $0.24 per training hour

You are charged per item or item attribute query based on the number of users* in the dataset processed by Amazon Personalize.

*The number of users (identified with a ‘user_id’) is calculated as the unique users in the union of your ‘Users’ and ‘Interactions’ datasets.

A retailer uses batch segmentation to generate lists of users that are likely to be interested in SMS and in-app messaging campaigns about particular products that are on sale. The retailer has 500,000 users that will be considered for the campaigns. It uses 10GB of data and consumes 50 training-hours. The retailer trains the solution once per month and runs segmentation on 10 products every two weeks.

The bill for the month for using Amazon Personalize will be:

• Data processing and storage charge = 10 GB * $0.05/GB = $0.50
• Solution training charge = 50 training-hours * $0.24/training-hour = $12
• Batch segment generation charge, first 100,000 users = 100,000 users * $0.016/1,000 users * 10 queries * 2 times per month = $32
• Batch segment generation charge, next 400,000 users = 400,000 users * $0.008/1,000 users * 10 queries * 2 times per month = $64

Total cost = $0.50 + $12 + $32 + $64 = $108.50

A media company uses batch segmentation to identify users that would be interested in streaming movies based on attributes of the movies, such as genre, lead actor/actress, and awards won. The company uses the lists of users generated to drive push notifications and email marketing campaigns. The service has 20 million users that are considered for each campaign. The company uses 650 GB of data, and training for each recipe requires 1,800 training-hours. It trains the two solutions once per month and runs segmentation on 25 different movie attributes every week.

The bill for the month for using Amazon Personalize will be:

• Data processing and storage = 650 GB * $0.05/GB =$32.50
• Solution training charge = 2 solutions * 1,800 training-hours * $0.24/training-hour = $864
• Inference charge, first 100,000 users = 100,000 users * $0.016/1,000 users*25 queries*4 times per month = $160
• Batch segment generation charge, next 900,000 users = 900,000 users*$0.008/1,000 users*25 queries*4 times per month = $720
• Batch segment generation charge, next 9 million users = 9,000,000 users*$0.004/1,000 users*25 queries*4 times per month = $3,600
• Batch segment generation charge, next 10 million users = 10,000,000 users*$0.001/1,000 users*25 queries*4 times per month = $1,000

Total cost = $32.50 + $864 + $160 + $720 + $3,600 + $1,000 = $6,376.50

Amazon Personalize makes it easy for you to build a wide array of personalization experiences, including specific product recommendations, personalized product re-ranking and customized direct marketing. Recommendations can be served in real-time to quickly respond to changing user intent or in batch.

The following pricing applies when using the following recipes:

• user-personalization
• popularity-count
• Personalized-Ranking
• Similar-Items
• SIMS
• HRNN (legacy)
• HRNN-Metadata (legacy)
• HRNN-Coldstart (legacy)
  

You are charged per GB of data uploaded to Amazon Personalize. This includes real-time data streamed to Amazon Personalize and batch data uploaded via Amazon Simple Storage Service (S3).

Data ingestion costs: $0.05 per GB

When creating a custom solution, you are charged for the training hours used to train a custom model with your data. A training hour represents 1 hour of compute capacity using 4v CPUs and 8 GiB memory. Amazon Personalize automatically chooses the best instance types to create your solutions, and the instance may exceed the baseline specifications in order to complete the job more quickly. Personalize calculates training hours based on the instance used relative to the baseline instance. The number of training hours charged may be higher than the time elapsed on the clock during training.

Training costs: $0.24 per training hour

Real-time recommendations
You are charged for the requests processed by Amazon Personalize. Real-time recommendations usage is measured in transactions per second (TPS). You specify the minimum throughput you need. If the throughput of your requested recommendations exceeds the minimum provisioned TPS, Amazon Personalize auto-scales to serve your additional requests and returns to your minimum provisioned TPS when your traffic reduces. The actual TPS served is calculated as the average requests per second within a five-minute window.

For real-time recommendations, you are charged for throughput capacity per hour in units of TPS-hour (rounded up to the nearest hour). This is calculated by taking the maximum between the actual TPS served and the minimum provisioned TPS, for each five-minute increment in the hour, multiplied by the total time that requests are processed. Your usage is aggregated for the month and billed according to the usage tiers.

TPS-hour = maximum of (actual TPS served, minimum provisioned TPS) * (5/60 minutes)

Batch recommendations
You are charged for the number of users requested when using “USER_PERSONALIZATION” and “PERSONALIZED_RANKING” recipes and for items requested when using “RELATED_ITEMS” recipe for a batch inference job, regardless of the number of results requested.

*Recommendations generated for each user when using a ‘solution’ based on “User-personalization” recipe type are counted as one recommendation regardless of number of results (items) requested per user for batch pricing. Similarly, you only pay for number of users processed for “Personalized-ranking” regardless of number of items re-ranked per user and for number of items processed when using “Related_items” recipe regardless of number of similar items per item requested.

A media company uses custom recommendations to serve recommendations at different points in their users’ journey. They upload 200 GB of data in the month, and train a single solution once per day with each training taking 20 mins to complete and consuming 10 training hours per training. Further, the customer uses inference capacity of 10 TPS for 720 hours for the month for generating real-time recommendations.

The bill for the month for using Amazon Personalize will be:

• Data processing and storage charge = 200 GB * $0.05 per GB = $10
• Solution training charge = 300 training hours * $0.24 per training hour = $72
• Campaign charge (real-time inference) = 10 * 720 * $0.20/ TPS-hour = $1440

Total cost = $10 + $72 + $1440 = $1552

For simplicity, let us assume the same amount of data upload and training as Example 1. Thus for:

• Data processing and storage charge = 200 GB * $0.05 per GB = $10
• Solution training charge = 300 training hours * $0.24 per training hour = $72

But, this time, we’ll vary the volume of requests throughout the day.

Inference usage and charge: In following table, we walk through a variable traffic scenario and calculate the TPS-hours consumed in a day of usage:

A retailer uses custom batch recommendations to generate item recommendations, for use in personalized emails. They use 10 GB of data, training consumes 50 training-hours, and a batch inference job is used to generate recommendations for 1 million users with 10 item recommendations for each user.

In this case the charges for using Personalize will be:

• Data processing and storage charge = 10 GB * $0.05/GB =$0.50
• Solution training charge = 50 training-hours * $0.24/training-hour = $12
• Inference charge = 1 million users * $0.067/1,000 recommendations = $67

Total cost = $0.50 + $12 + $67 = $79.50