Introduction
Most test publishers work according to the waterfall model - they develop content for a new test, collect data, implement and launch. Then they do not touch the content until it’s time for the next version. Once the new version is out, it is treated as an entirely new product and the difference to the previous version may be very large. Results from the new version are often not comparable to the old version.
At Alva, we started out with a best-in-class personality test, powered by best practices in machine learning and modern test theory. We are now also introducing the agile model in the way we develop and iterate on our tests. Instead of waiting up to five years to launch an entirely new version of our personality test (which is common in the industry), we are continuously collecting data and introducing new questions. This way, we are making sure that our personality test stays ahead of the curve and performs even better than before.
What we do
Introduce new questions to reduce social desirability, cover the entire range of possible results and target work related behaviors.
Replace old questions based on customer feedback, our internal review of the content and psychometric analysis that shows performance below our standards.
Enrich the IRT model with the latest data.
Quality checks
The reliability of all facets and factors increase or remain at an excellent level according to European standards
The construct validity of all factors increase or remain at an excellent level according to European standards
The correlation between old and new scores is higher than 0.9
The average scores are not changed by more than 0.5 points on the standard scale (1-10)
More than 50% of users keep an identical score
Less than 5% of users change by more than +/- 1 point