How to improve project performance

The first step to improving your project’s performance is to measure it, which you can learn how to do on our How to monitor project progress and quality page.

Once you have an idea of your project’s performance, you can take actions to improve it. There are 3 ways you can do this:

  • Improve instructions and task design
  • Choose a different pricing plan
  • Use your own labelers

Super.AI will also make adjustments behind the scenes to optimise performance. You can find out more about this on our Quality assurance page.

Improve instructions and task design

Below you can see how the super.AI system enables a feedback loop that allows you to make improvements to your project’s performance.

The super.AI feedback loop, showing task instructions, super.AI, output review, and communication with labelersThe super.AI feedback loop, showing task instructions, super.AI, output review, and communication with labelers

The super.AI feedback loop, showing task instructions, super.AI, output review, and communication with labelers

Within the super.AI system, you are able to affect quality through the following processes:

You can use what you learn during the review process to make modifications to your instruction set and provide feedback to the human labelers on your project to address common mistakes and edge cases.

Choose a different pricing plan

Every project you create with super.AI comes with a custom pricing plan.

You can view and change your project’s pricing plan at any time by opening your project in the super.AI dashboard and navigating to the Settings page.

If you're not sure what plan suits your project best, talk it over with our team.

Use your own labelers

Sometimes projects require specialized expertise or tightened privacy and security that a third-party crowd can’t provide. For these cases, super.AI provides the option to use your own labelers that you invite to join your project. These two pages will give you the information you need to decide which option is best and configure super.AI for your own labelers:


Did this page help you?