Join us Partner with us

Partner with us

Every project starts with one decision your team makes again and again. We test a change on your data, then in your own operations, and hand over what worked as tools your team keeps.

Propose a project

From the first call to tools you keep

  1. Step 1

    Scope the decision

    We start with a conversation about one decision your team makes again and again: which job to offer a driver, when to mark a product down, when an AI agent should hand a customer to a person. Together we pin down who makes it today, what they know when they make it, and what better would look like in a number you already track.

    What we ask of you
    Someone who owns the decision, and an honest look at how it gets made today.
    What you get
    One page: the decision, the outcome we will measure and the data it needs.

  2. Step 2

    Pilot on your data

    We model the decision from the records you already keep and use the model to find where a change could help most. Before anything touches live operations, we check the idea against your own history, so the field test starts from changes that already hold up on paper.

    What we ask of you
    Historical data on the decision and its outcomes, shared under a data agreement your team is comfortable with.
    What you get
    A working model on your own data and a short list of changes worth testing.

  3. Step 3

    Field experiment

    Some workers, stores or customers get the change and a comparable group carries on as before, chosen at random. Because the groups differ only by chance, the gap between them is what the change did. We watch it as it runs, and we look at how people respond to it, not only at the headline number.

    What we ask of you
    A part of your operations to test in, and a contact who tells us when something in the field changes.
    What you get
    Measured results, reviewed with your team.

  4. Who should get this delivery?

    Offer it first to the driver four minutes away, who has taken most offers like it this week.From the dispatch tool your team keeps

    Step 4

    Tools you keep

    What worked becomes part of how your team decides. Sometimes that is guidance for managers, sometimes a better recommendation inside a tool people already use, sometimes software that runs on its own. We build it with the people who will use it, so it gets followed.

    What we ask of you
    The people who will use it day to day, for a few working sessions.
    What you get
    Guidance or software your team runs from then on, and what we learned along the way.

Bring these to the first call

Rough answers are fine. Working them out together is the first step.

  1. Which decision does your team make again and again?Which job to offer, which shift to staff, when to mark down, when to hand off to a person.
  2. Who or what makes it today?A manager, a fixed rule, a model, or some mix of the three.
  3. What already records it?The data that shows each decision and what happened after it.
  4. What would better look like?One number you already track and would be glad to move.

What we work on

Field experimentsRandomized tests of pay, incentives, scheduling and advice, run in your own operations.
Algorithm and incentive designDispatch, recommendations and pay schemes that people actually follow.
AI and decision toolsSoftware and AI agents that know when to act, ask or defer, measured before they ship.

Have a decision in mind?

Propose a project

Write to us at [email protected] with a few lines about the decision and the data behind it, and we will set up a first call.