Starting and tuning a project

Use Toloka to post tasks and find performers. When the assignments are completed, you can analyze the results. This section describes the general procedure for starting and setting up your project with tasks.

Tip. First, take a look at the list of examples in Solutions to popular tasks if you haven't done so yet. Perhaps sample instructions are already available for your task type.

You can contact Toloka partners if you need help with your project as a whole or with individual steps like task markup.

Before you start

Make sure you're registered as a requester.

If you have a complex project, register in the sandbox and create a project there. It lets you test the settings of your project as a performer and then transfer them to the Toloka production version and put them into use by real performers. This way you can avoid making mistakes and spending money if it turns out that your task is not functioning.

Workflow

  1. Think through the solution architecture. Perhaps you should create a number of small interrelated projects instead of one project. This will simplify a task and hence reduce its cost and improve the quality of results.
  2. Create a project. Project settings define the task interface and the input and output data.
  3. Create a task pool in the project. Pool settings define all other settings, such as pricing, rules for displaying tasks to performers, and quality control.
  4. Add tasks to the pool. To do this, you need a TSV file with task data and correct responses to control and training tasks.
    Tip. First, send a small trial pool (10–100 tasks) for completion.
  5. Add a training pool. Using training pools, performers can practice before getting started and you can assign the main tasks only to those performers who have passed the training.
  6. Start the task pool so that performers can see your tasks.

    After starting the pool, watch for incoming messages from performers during the first hour. Usually performers report problems with tasks quickly.

  7. Process the results. If the response quality is poor, try to find out why. You may need to edit the instructions or add more tasks to the training.

    Check the responses if you chose to review assignments when configuring the pool.