Flowchart
Which quasi-experimental design fits?
A decision flowchart for picking a quasi-experimental design when you cannot run an A/B test. A threshold rule points to regression discontinuity, a group you did not choose points to difference-in-differences, and a change everyone got at once points to interrupted time series. Every path ends with a confounder sweep, which Koji can run as short AI interviews with the treated group.
Where this appears
What's in this flowchart
Start: You cannot randomise who gets the change
- Was it assigned by a threshold or rule?
- No: go to 2.
- Yes: Use regression discontinuity. Compare units just below the cutoff with those just above.
- Did it land on a group you did not choose?
- No: go to 3.
- Yes: Use difference-in-differences. Subtract the change in an untreated comparison group.
- Did everyone get it at one clear moment?
- Yes: go to 4.
- No: It is only a before-and-after. Not a quasi-experiment. Treat the result as weak evidence.
- Use interrupted time series. Project the pre-intervention trend forward as the counterfactual.
- Finish: Ask the treated group what else changed. A confounder sweep catches events no dashboard holds.
Use this image
Free to use in your docs, decks and posts under CC BY 4.0, as long as you credit Koji with a link to this page. To embed it, paste this into your page:
html
<a href="https://www.koji.so/visuals/which-quasi-experimental-design-flowchart"><img src="https://www.koji.so/visuals/which-quasi-experimental-design-flowchart/which-quasi-experimental-design-flowchart.png" alt="Flowchart for choosing a quasi-experimental design: three questions on how treatment was assigned, leading to one of three designs." width="1200" height="1426" /></a>
<p>Source: <a href="https://www.koji.so/visuals/which-quasi-experimental-design-flowchart">Which quasi-experimental design fits?</a>, Koji</p>
