Today, Vantage launched machine learning powered cost forecasting for all users. Vantage Cost Reports now include forecasts that are based off of their AWS account's historical cost patterns. As a result, customers can see what their expected total AWS bill will be for the current month in total as well as on a per-AWS service. Lastly, as forecasts are generated for each Cost Report, users can also forecast for an individual AWS Member Account, AWS Tag or specific region. From speaking with customers we find that individuals and small startups enjoy seeing forecasts for their entire AWS account and per-AWS service whereas larger organizations with more complex spend may want to see it broken down by AWS Tag or Member AWS Account.
Prior to this launch, Vantage made basic cost predictions based on current accrued cost during the month. However, since spend can be spiky, especially at the start of a month, and dependent on actual AWS service behavior patterns, this resulted in inaccurate and often higher-than-expected forecasts.
Now, users can see accurate forecasts generated by a machine learning model that incorporates their account's historical costs. The forecast is updated daily as new cost patterns are synced and represented in each Cost Report accordingly. This forecast will account for first of the month spikes, service-dependent changes and common seasonality charges. The machine learning model that powers these forecasts will also adapt as new patterns emerge to provide more accurate forecasts over time.
At the time of this blog post, all users have access to forecasting.
What is being launched today?
Today, Vantage is launching AWS Cost Forecasts: the ability for users to get visual and numerical predictions for what their end-of-month expected bill will be. Users can see forecasts in two parts: (1) for the total amount of costs for a Cost Report and (2) on a per-AWS service basis.
What is a Cost Report?
A Cost Report is a set of constraints to segment portions of your infrastructure. Vantage creates a default Cost Report for each user, but users can also create new Cost Reports for things like seeing costs per AWS Tag, per region, or per AWS Member account. You can read more about Cost Reports here.
How does Cost Forecasting work?
Cost Forecasts leverage machine-learning models that are trained by looking at historical and current daily accrued costs. The models look at daily incurred costs for each AWS service that aggregates into a total cost. Forecasts will generate a prediction of what these costs will be with a particular range (that can vary based off of the model's confidence) with a median forecast that is also included.
How accurate are Cost Forecasts?
Depending on the particular AWS service and behavior, forecasts can be incredibly accurate. That being said, the models are constantly learning and can potentially vary from reality depending on how erratic the behavior is in your AWS account. From running an early beta of forecasting, we haven't had any users mention any negative experiences with forecasting in terms of accuracy.
Will forecasts be included in weekly email reports?
Not at this time but we are working on including forecasts into weekly email reports in the near future.
I created a new Cost Report, when will the forecast generate?
Depending on the complexity of your AWS account data, a forecast will be generated typically within 10 minutes. All forecasts are updated daily as new cost data is retrieved.
Is Forecasting data available through the API?
Not yet, but we are looking to add support for this soon.
How do I access Forecasting?
Assuming you have an AWS account added to your Vantage account, forecasting should be live for your account at the time that this blog post has been published.
What does the gray line represent on the forecast?
The gray line shows actual accrued costs for your account during the previous period. So for example, if you're looking at a forecast for June, we will show you your account's actual accrued costs from May with the gray line to see how things are trending.
I have feedback or suggestions for forecasting, how can I give it to you?
Please email email@example.com or join our Community Slack Group.