Update a list of topics of interest and mark them as mandatory, required or desirable. Mandatory topic: Everyone in the team should be familiar with it. There must be provided a list of literature that covers that topic in minimal content. Required topics are those that are a must for the team, but not a must for each in the team. Every required topic should be covered by at least by two team members. The more we distribute the required topics the larger and deeper is our team knowledge base. Desired topics are those that are not widely used in MaxInspect but may be applied in future.

Try to perform bias-variance error decompostion when possible

Singling out sources of inaccuracy in metrological tests; for every source of inaccuracy there should be a test that is able to precisely identify if that source is critical. The tests that are now used by metrology to identify if device is "good" or "bad" but don't explain what exatly is "bad" should be reviewed and replaced with more descriptive tests (like left-to-right reproducibility tests on ferrule)

Documentation: All the specific algorithms that we develop should be structured and documented. We may create a requirement that the corresponding bugzilla bug must not be closed until the algorithm is not described in Wiki.

Systematization first: most critical problems should be investigated first and there is no need to go much deeper (if it's time consuming) than the problems we have.

All modeling scripts should be publicly available, easy-to-parametrize and easy-to-use.

Implement automatic tests on modeled data to be able to test the algorithms on a wide set of inputs and check their limits.

All algorithmic problems and their possible reasons should be listed and their criticalness estimated. It will help to find the gaps in our knowledge base and algorithms and plan corresponding research.

One of the simplest models is pertubated real data, it can be used for experiments