Skip to content

← Back to courses

Running Behavioural Experiments

Find out whether your intervention actually worked.

81 min · 8 modules · 8 assessments
A campaign ran and the numbers moved, but was it the campaign? This course teaches you to set up a test that could fail, choose a design that rules out the usual explanations, run it for long enough, read the result honestly and turn it into a decision the business will act on.
Register to enrol →

Already have an account? Sign in

Watch the welcome video

What you will be able to do

Write a claim that can fail, with one primary outcome, guardrails and a baseline, sealed before launch.

Choose a design, from a matched comparison to a stepped wedge or cluster randomised trial, and say which rung it stands on.

Work out how many teams or sites you need and how long to run, and avoid stopping at the first good week.

Run a pilot that tests delivery, with fidelity, dose and kill criteria written down first.

Analyse a result with difference in differences and an interval, and know when to call a statistician.

Spot the threats to validity, run tests fairly and lawfully, and turn the result into scale, adjust or stop.

What's inside

  1. 01

    Welcome

    1 video · 7:49

  2. 02

    The Question and the Outcome

    Turn an intervention idea into a testable claim with one primary outcome, guardrails, a baseline and a prediction written before the data.

    1 video · 8:26

  3. 03

    Designs: From Before-and-After to Randomised

    Choose the design that gives you a real counterfactual: randomised where you can, clustered where people talk, stepped where you must roll out, matched where nothing else is possible.

    1 video · 8:23

  4. 04

    Sample Size, Power and How Long to Run

    Work out before you start whether the test can see the effect you care about, and write down when you will stop looking.

    1 video · 8:35

  5. 05

    Running a Pilot

    Run a pilot that answers a question about delivery, with sites chosen honestly, fidelity logged, and kill criteria agreed before it starts.

    1 video · 7:56

  6. 06

    Analysing the Result

    Compute the effect as a difference in differences, report it as an effect size with an interval, and read a null honestly.

    1 video · 8:52

  7. 07

    Threats to Validity

    Recognise the eight things that can produce a result instead of your intervention, say what each did to the estimate, and mitigate them in the design.

    1 video · 9:10

  8. 08

    Ethics, Fairness and Consent

    Design an experiment on colleagues that withholds no protection, blames nobody, tells people what they need to know, and treats every group fairly.

    1 video · 8:36

  9. 09

    From Result to Decision

    Turn an estimate into a decision: scale, adjust or stop; price it per unit of behaviour; set the maintenance dose; and report it on one page.

    1 video · 9:40

  10. 10

    Close

    1 video · 3:45


Assessments

  • ✓

    Module 1 quiz

    6 questions · pass mark 66%

  • ✓

    Module 2 quiz

    6 questions · pass mark 66%

  • ✓

    Module 3 quiz

    6 questions · pass mark 66%

  • ✓

    Module 4 quiz

    6 questions · pass mark 66%

  • ✓

    Module 5 quiz

    6 questions · pass mark 66%

  • ✓

    Module 6 quiz

    6 questions · pass mark 66%

  • ✓

    Module 7 quiz

    6 questions · pass mark 66%

  • ✓

    Module 8 quiz

    6 questions · pass mark 66%


Course workbook

Exercises and reflection prompts that follow the videos. Print it, or fill it in as you go.