Teachometry
Run

Work with usEvaluation first. Selected collaborations.

See how your AI tutor actually teaches.

Teachometry builds tools to understand and improve AI teaching quality. We also collaborate with teams on evaluation, technical research, advisory work, and selected engineering projects.

Discuss a project How we work

A paper pattern piece for the front of a bodice, drafted in pencil on white dot-and-cross pattern paper, lies on periwinkle wool twill. A yellow tailor’s tape measure runs from its brass end along the hem, off the paper and up in a half twist. Three glass-headed pins hold the side seam and two lie loose on the paper; the five heads are ruby, citrine, emerald, sapphire and amethyst.
Cover · Made to measure: a paper pattern pinned to wool, a tailor’s tape laid across it, five pins in the colours of the five dimensions. An evaluation is cut to the product in front of it.

What we do

Evaluation first.

AI education evaluation is the core of our work. Research and advisory grow out of it, and engineering projects are taken on selectively.

Core

AI education evaluation

For teams building AI tutors and AI education products, education technology companies, and research groups.

  1. 01

    Teaching-quality evaluation and independent review

    How your AI tutor behaves with learners, observed across structured cases. We can also act as an independent reviewer of its teaching.

  2. 02

    Custom scenarios and evaluation criteria

    Scenarios built around your subject, your learners and your product, with the criteria agreed before any results are seen.

  3. 03

    Diagnosis, evidence reports and regression tests

    Where the teaching goes wrong, shown case by case with the evidence, then kept as tests you can run again after every change.

  4. 04

    Product-quality research and evaluation partnerships

    Longer studies of the quality of AI education products, carried out together with your team.

Collaboration

Research & advisory

AI system evaluation, technical research, product-quality advice, prototype validation, research collaboration, and ongoing advisory work.

Supplementary

Selected engineering work

We take on a small number of remote software projects: API integrations, AI tools, automation and technical delivery. This work helps fund the research, so we only accept what leaves room for it.

Not sure where your question fits? Bring it as it is. Working out the scope together is part of the work.

How we work

From a first conversation to delivery.

Every engagement is scoped with you and agreed in writing before work starts.

  1. 01

    Discuss

    You tell us what you are building and what you need to know. We ask questions until the problem is clear.

  2. 02

    Scope

    Together we settle the scope, the deliverables, the timeline and the fee.

  3. 03

    Agreement

    The terms are put in writing and agreed by both sides before any work begins.

  4. 04

    Delivery

    We carry out the work as agreed and hand over what was promised, with the evidence behind it.

Forms of engagement

By project
One defined piece of work with a fixed scope.
In milestones
Larger work in stages, each one agreed and delivered in turn.
Ongoing advisory
Regular advice over a longer period.

Which of these we can offer depends on the work and on our capacity at the time; not every form is always available.

We don’t publish fixed prices, and nothing is sold online. Fees are agreed for each piece of work.

Independence and boundaries

Fees pay for the work,
not for the findings.

These rules apply to every engagement, whoever it is with.

  1. § 1

    Findings are not for sale.

    Independent research and public results are never tied to sponsorship or consulting fees. An evaluation reports what the evidence shows, whoever pays for it.

  2. § 2

    Two roles, kept apart.

    Evaluating a product independently and helping to build it are different roles, and they can conflict. We say up front which role we are in, and if we have helped build something, we say so before we evaluate it.

  3. § 3

    TutorBench stays whole.

    An engagement does not transfer the intellectual property of TutorBench as a whole. What each project delivers, and who owns it, is set out in the agreement.

  4. § 4

    Honest about the instrument.

    TutorBench is a Developer Preview. Its results are not yet calibrated, and it cannot show that learners learn more. No report of ours will claim otherwise.

Read the methodology and its limits

Discuss a project

Tell us what you are building.

contact@teachometry.com

Discuss a project Other ways to reach us

A useful first email covers

  • what your product does, and who its learners are
  • the question you want answered
  • any timeline or budget you already have in mind
  • constraints such as data, privacy, or who will see the results

Please don’t send real learner data or private transcripts in a first email. How data is handled is agreed before any of it is shared.