Skip to main content
This guide takes you from a question, “which of these outputs is better?”, to an evidence-backed answer. You need a HeyBee account and a tablet or desktop for the setup steps.
1

Create an evaluation

Go to Evaluations and select New evaluation. Give it a name, pick the output type (text, image, audio, or video), and choose Which option wins? to compare candidates head to head.Keep the shared input as prompt unless your data uses a different field. Outputs that share the same prompt are the ones compared against each other.
2

Add outputs

Open the Outputs tab and add at least two outputs for each prompt. Paste a table, upload a CSV, drop media files, or import a Hugging Face dataset. HeyBee validates everything up front and shows exactly what to fix before collection can start.
3

Start collection and share a link

On Setup & collect, select Start collection once the readiness checks pass. Then create a voting link under Voting links with New link. The URL is copied to your clipboard, ready to share. Voters do not need an account.
4

Vote, or let others vote

Anyone with the link sees two outputs side by side and picks the better one. Each decided comparison costs one credit; ties and skips are free. The Free plan includes 100 credits every month.
5

Read the result

Open Results. The verdict card shows the leader, the confidence score, and a concrete next step, for example “Collect 12 more votes before finalizing.” When the signal is strong enough, select Complete evaluation to lock in the answer.
Prefer code? The Python SDK and CLI run this same workflow end to end.

​
Where things live

Every evaluation has the same tabs:
TabWhat it is for
Setup & collectConfiguration, readiness checks, voting links, and collection controls
OutputsYour uploaded or imported outputs (called Configurations in settings evaluations)
ResultsThe verdict, rankings, and per-criterion outcomes
ReliabilityIndividual votes, voter quality, and written feedback
Phones are great for voting, monitoring collection, reviewing results, and managing your account. Creating evaluations and importing data need a tablet or desktop.