Previous
¶ Data Review and Processing
¶ What it is
Following a round of experimentation, there is usually a Data Review, whereby the team looks at all the experiment data generated, processes that data, and turns it into the form that is ultimately helpful in scoring the experiment based on the agreed criteria.
¶ Why it is useful
Raw experiment data is typically not in the right state for experiment evaluation. It will typically only be in an intermediate form, such as raw interview scripts or raw analytics data, which needs to be reviewed and processed to be helpful.
¶ When to use it
Whenever you conduct experiments, you will need to review and process the data that results from those experiments.
¶
¶ How is it done
- Following an experimentation sprint, set aside time for processing and review.
- Open the data room for your experimentation cycle, and go experiment by experiment, reviewing the target evaluation and the current state of the data.
- Process the data as needed to make it useful
- Interview protocols might need to be mined
- You may need to run algorithms on data sets to derive a target variable to evaluate
- Statistical analysis of survey results will help get the real insights
- Review and continue until the data you have is helpful for evaluation.
¶ Do's & Don't
Do's
- Review data before processing to ensure it is of the right quality.
- Sense-check processing outcomes to ensure they make sense and you are on track to getting valuable results.
Don't
- Don’t run statistics or algorithms on data sets that are too small and will only give you a false sense of security in aggregation and processing.
¶ Tools needed
- Experiment data
- Processing tools
¶
¶ Example
