For clients

Working with us

We work best with machine-learning teams who care what is in their data and where it came from. Most engagements start small and last years: our longest has run since 2016.

Clients in 2025
46
First-pass acceptance across all projects
98.7%
Average length of a client relationship
4.2 years
Hours a day our centres are labelling
21

How an engagement runs

  1. Conversation

    Tell us what the model must do, what the data is, how much there is and when you need it. We will tell you honestly whether we are the right people.

  2. Provenance

    Before we quote, we ask in writing how the data was collected, from whom, and on what basis you hold it. This is article 4 of our ethics charter and is not negotiable.

  3. Pilot

    A paid pilot of 2,000 to 10,000 items, usually two weeks, with a calibration round and a quality report you can check against your own gold set.

  4. Production

    A dedicated team in one centre or both, with a named project lead in Toronto and a team lead in each centre.

  5. Reporting

    Weekly quality reports and a monthly acceptance statement, from which we invoice.

Pricing

We quote per accepted item, by task type, after the pilot. Items you return are corrected at our cost, so the price you pay is the price of correct data. For evaluation and some text work we quote per judgment or per hour. Restricted-room work and content tiers S2 and S3 carry a surcharge that pays for the rooms and the wellbeing provision.

Invoices are raised on the fifth working day of each month for the items accepted in the month before, in US or Canadian dollars, on thirty days’ terms.

Tell us about your project

The task, the volume, the timescale, and where the data comes from.

Page last reviewed 12 October 2026