Experiment Template

Experiment Template (WIP)

• Status: WIP|Submitted|Approved|Rejected
• Deciders: Leadership Team|Team|Product
• Submitters:
• Date:

This template is not meant to create bureaucratic red tape, but rather to make suggestions about the sorts of information that is needed to decide if an experiment is worth running and to record the outcomes of the experiment. If it would take longer to write this up than to do the experiment, then this form may be overkill.

Learning/Goal

What is the one critical question you are trying to answer?

Assumptions

What do you already know? What biases do you have?

Data

What is measurable? Qualitative or quantitative?

Hypothesis/Fail Condition

Is there something falsifiable and specific to test? What would convince you your hypothesis is false?

Plan

How will you collect the data? Is the plan specific? Is it achievable? Link to any supporting documents?

Time Box

Is the plan timely? Can you get the data faster? Would less data be sufficient?

Expenditure Box

Is the plan frugal? Could you get the data for less? Would less data be sufficient?

Early Stop Condition

What would let you know right away that your plan won’t work?

Implementation

At a high level how will the experiment be implemented and what teams or process may be impacted?

Revert

After the experiment time box or in the event of an early stop condition, how is the experiment reverted?

Risks

Risks to the experiment itself

Are there any potential failure points that are already known? Is there a plan to mitigate these?

Unintended Consequences

How will this experiment potentially affect other people, projects, processes, etc?

Results & Insights

What happened? Why? Anything unexpected?

Next Steps

What changes should you make based on these results? Do you need more data or a new plan?

If the experiment:

1. Answer hypothesis (successfully)
2. Has a path forward for meaningful impact

Then do one or more of the following:

• Submit an decision record
• Share a report (this document or an expanded, more detailed version)

If not, determine if the experiment is concluded or may need to be performed again through a new experiment.